OUR WORLD - Perspectives on Artificial Intelligence
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
January 2021<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> WOR LD<br />
<str<strong>on</strong>g>Perspectives</str<strong>on</strong>g> <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong>
A NEW EUROPE PUBLICATION<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g><br />
<str<strong>on</strong>g>Perspectives</str<strong>on</strong>g> <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong><br />
Foreword by Adina Vălean<br />
European Commissi<strong>on</strong>er for Transport<br />
&<br />
Kyriakos Pierrakakis<br />
Greek Minister of Digital Governance<br />
In partnership with
The Cover<br />
EDITOR & PUBLISHING PARTNER<br />
Jerry Zagoritis<br />
ASSOCIATE EDITOR’S<br />
Nicholas Waller<br />
Ariti-Marina Alamanou<br />
EDITION COORDINATORS<br />
Vassilis Nanis<br />
LAYOUT & DESIGN<br />
Suman Haque<br />
LOGISTICS & OPERATIONS<br />
Irini Panagopoulou<br />
CONTRIBUTORS:<br />
Julia Anders<strong>on</strong><br />
Anna-Michelle Asimakopoulou<br />
Adam Bielan<br />
Iban Garcia del Blanco<br />
Sybrand Brekelmans<br />
Angeliki Dedopoulou<br />
Alexandra Geese<br />
Eva Kaili<br />
Hannes Kirchbaumer<br />
K<strong>on</strong>stantinos Kyranakis<br />
Abraham Liu<br />
Luis Neves<br />
Georgios Petropoulos<br />
Kyriakos Pierrakakis<br />
Dragos Tudorache<br />
Adina Valean<br />
Axel Voss<br />
Cover title: <str<strong>on</strong>g>Perspectives</str<strong>on</strong>g><br />
Illustrati<strong>on</strong> & C<strong>on</strong>cept: Lavrentis Choraitis<br />
Images: iStock by Getty Images & Shutterstock<br />
A publicati<strong>on</strong> by:<br />
ISSN number: 2593 - 4163<br />
Published by Brussels News Agency SPRL<br />
NEW EUROPE Tel. +32 2 5390039<br />
Av. de Tervuren/Tervurenlaan 96,<br />
1040 Brussels, Belgium<br />
Fax +32 2 5390339<br />
info@ourworld.co<br />
© 2021 Our World all rights reserved. No part of this publicati<strong>on</strong> may be reproduced, stored in a retrieval system, or transmitted in any form by any<br />
means, electr<strong>on</strong>ic or otherwise, without express permissi<strong>on</strong>. The Publishers accept no liability for third party views published, nor damage caused by<br />
reading, viewing or using our c<strong>on</strong>tent. All informati<strong>on</strong> is correct at the time of going to print, we accept no liabilities for c<strong>on</strong>sequent changes.
Editor’s Letter<br />
A<br />
few m<strong>on</strong>ths back, while preparing a workshop <strong>on</strong> gender stereotypes and biases, I came<br />
across research by the World Ec<strong>on</strong>omic Forum suggesting a significant gender<br />
gap am<strong>on</strong>g AI professi<strong>on</strong>als, with <strong>on</strong>ly 22% of AI professi<strong>on</strong>als globally being female,<br />
compared to 78% being male. Perhaps I was expecting it, but I still found it not just disturbing;<br />
more than that. I thought it’s dangerous and I’ll explain why right away.<br />
By Jerry Zagoritis<br />
Editor and<br />
Publishing Partner<br />
of Our World,<br />
and the CEO of<br />
Campaign Lab<br />
An algorithm that has been coded by humans, may reflect biases of the developers. If the<br />
data collected and used for the development of AI is gender biased, the results will also be<br />
biased, leading as a c<strong>on</strong>sequence to biased decisi<strong>on</strong>s in a number of fields, including for<br />
example hiring decisi<strong>on</strong>s.<br />
And so I’d like to use this editorial as an opportunity to raise the attenti<strong>on</strong> of the business<br />
leaders and policy makers hosted in our pages to this simple fact: to avoid the replicati<strong>on</strong> of<br />
stereotypes and biases, and to enhance AI’s efficacy, we must urgently promote more diversity,<br />
including the participati<strong>on</strong> of more women, in the development of AI.<br />
Back to O/W. In the development of this editi<strong>on</strong> we invited authors who share our visi<strong>on</strong> that<br />
our world needs inspirati<strong>on</strong>, passi<strong>on</strong> and boldness for a better tomorrow.<br />
This is why we invited the Transport Commissi<strong>on</strong>er, Adina Valean, and the Greek Minister<br />
for Digital, Kyriakos Pierrakakis, to foreword this special editi<strong>on</strong>. Mr Pierrakakis has inspired<br />
change and has helped his (our) country take unprecedented digital leaps forward, while Ms.<br />
Valean is no stranger to the readers of this magazine - she has repeatedly c<strong>on</strong>tributed her<br />
always inspiring thoughts to O/W editi<strong>on</strong>s over the years.<br />
The topic of <strong>Artificial</strong> <strong>Intelligence</strong> can be divisive by nature; as are most things novel. With that<br />
in mind we are excited to be hosting representatives of the European Parliament’s Special<br />
Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age (AIDA) from all political groups, except<br />
<strong>on</strong>e* – including the AIDA Committee Chair and the European Parliament Rapporteur. EU<br />
Policy makers who, next to important business leaders, nati<strong>on</strong>al policymakers and thinkers,<br />
c<strong>on</strong>tribute to what we see as an exciting debate <strong>on</strong> the present and future of our world.<br />
Happy Holidays and a Happy New Year!<br />
*We made sure to invite Members of European Parliament from all political groups in order to foster<br />
a fair and balanced debate. Unfortunately, we didn’t receive a resp<strong>on</strong>se from the GUE/NGL (European<br />
United Left) Member of AIDA, although our team called and followed up by email repeatedly.
Foreword<br />
By Kyriakos Pierrakakis<br />
Greek Minister of Digital<br />
Governance<br />
The dream to create an artificial being that could think and act just like a human<br />
was c<strong>on</strong>ceived in the myths of ancient Greece and has travelled for more than<br />
2,000 years to the science-ficti<strong>on</strong> literature of modern times. But each passing<br />
day the dream of <strong>Artificial</strong> <strong>Intelligence</strong> (AI) gets closer to realisati<strong>on</strong> as technology<br />
advances from all over the world suggest.<br />
Even though we still lack flying cars, according to the visi<strong>on</strong>s of 20th century’s futurists,<br />
but we are close to having completely aut<strong>on</strong>omous vehicles that can safely transport<br />
people and goods regardless of the weather c<strong>on</strong>diti<strong>on</strong>s. AI technologies are quickly<br />
integrated into medical procedures, manufacturing, banking, trade, art, and numerous<br />
other everyday activities and it is just a matter of time before seeing it expanding further.<br />
When it comes to AI, Europe is at a pole positi<strong>on</strong> as it has a str<strong>on</strong>g track record in<br />
research and in development. Currently, <strong>on</strong>e out of four industrial robots is made in<br />
Europe, so we have an advantage towards our global competitors. We, as Europeans,<br />
can further expand our lead doing what we know best – cooperate and address together<br />
the challenges that could emerge in the future.<br />
But with no doubt, AI is a great power and great power entails great resp<strong>on</strong>sibilities.<br />
European Commissi<strong>on</strong> President Ursula v<strong>on</strong> der Leyen rightly pointed out in her remarks<br />
<strong>on</strong> Shaping Europe’s Digital Future last February that “<strong>Artificial</strong> <strong>Intelligence</strong> must serve<br />
people and therefore, <strong>Artificial</strong> <strong>Intelligence</strong> must always comply with people’s rights”.<br />
We, as policy-makers, are called to strike the right balance and ensure that<br />
technological evoluti<strong>on</strong> goes hand-in-hand with upholding the rights of people. We<br />
are called to ensure that AI is used to benefit all citizens within the European Uni<strong>on</strong>. We<br />
must also ensure that the disrupti<strong>on</strong> caused by the evoluti<strong>on</strong> of technology, does not<br />
lead to exclusi<strong>on</strong> or enhance inequalities and does not questi<strong>on</strong> well-founded values<br />
such as the protecti<strong>on</strong> of privacy. We want AI and all other technologies to empower<br />
people, provide inspirati<strong>on</strong>, and serve the needs of our societies. AI poses the challenge<br />
of addressing biases and ensuring that the great achievements of technology benefit<br />
every<strong>on</strong>e regardless of residence, gender, and pers<strong>on</strong>al beliefs. Trust must be built for<br />
people to use AI in their daily activities and this is where policy-makers are called to work<br />
together with engineers and all societal stakeholders and ensure that AI is an inclusive<br />
technology of inclusi<strong>on</strong> that will not leave any<strong>on</strong>e behind.
Women in Energy, Climate and Sustainability (WECS) is a<br />
public foundati<strong>on</strong> established to promote gender equality<br />
as an enabler of the transiti<strong>on</strong> towards a climate neutral<br />
ec<strong>on</strong>omy in Europe and worldwide.<br />
Our objective is to facilitate gender diversity and women<br />
empowerment in the fields of energy, climate and<br />
sustainability.<br />
info@wecsfoundati<strong>on</strong>.org<br />
wecsfoundati<strong>on</strong>.org
Foreword<br />
By Adina Vălean<br />
European Commissi<strong>on</strong>er for<br />
Transport<br />
W<br />
i ill <strong>Artificial</strong> <strong>Intelligence</strong> mean<br />
driverless cars criss-crossing<br />
our towns and cities, with<br />
passengers relaxing in the back seat? It<br />
could mean this, <strong>on</strong>e day in the future. But<br />
what most people d<strong>on</strong>’t realise, is that AI,<br />
while still in its infancy, is already tip-toeing<br />
its way into our transport network – just less<br />
visibly than the arrival of aut<strong>on</strong>omous cars<br />
would be. AI has the potential to transform<br />
our travelling experience, enabling a<br />
smoother and more sustainable journey<br />
for drivers, passengers and freight, and –<br />
crucially – making mobility even safer.<br />
AI is already making a difference. It is<br />
part of a new algorithm used to provide<br />
expected time of arrival for rail freight<br />
trains, and has significantly improved<br />
performance over time through machinelearning.<br />
Accurate arrival time predicti<strong>on</strong>s<br />
are important for the seamless integrati<strong>on</strong><br />
of rail freight with road and maritime<br />
intermodal services. The technology was<br />
developed through the EU-supported<br />
ELETA project.<br />
In aviati<strong>on</strong>, applicati<strong>on</strong>s are being<br />
deployed to automate flight plan<br />
correcti<strong>on</strong>s, improve passenger transfers<br />
and predict traffic. In shipping, AI is helping<br />
c<strong>on</strong>tainer terminals and carriers alike to<br />
determine the best sequence for docking<br />
and loading/unloading. It does this by<br />
randomly trying many soluti<strong>on</strong>s and<br />
comparing the merits of each soluti<strong>on</strong>.<br />
It then starts applying its knowledge to<br />
each new scenario it tests so they get<br />
progressively better.<br />
The transport industry has clearly<br />
seen the potential, and is investing heavily.<br />
Driverless trains, cars able to detect animals<br />
<strong>on</strong> the road, and vessels able to advise <strong>on</strong><br />
the most fuel-efficient way to operate al<strong>on</strong>g<br />
a specific shipping route are the focus of<br />
recent trials around Europe. Companies<br />
are investing, testing and manifesting<br />
their interest. For road transport, we are<br />
still far from aut<strong>on</strong>omous cars. But we are<br />
already <strong>on</strong> a path to cooperative c<strong>on</strong>nected<br />
automated mobility (CCAM). Particularly<br />
in talk of cars, the relati<strong>on</strong>ship between<br />
AI and aut<strong>on</strong>omy is often misunderstood.<br />
The former does not necessarily mean the<br />
latter. Instead, AI can be used to provide<br />
different levels of support to a car driver:<br />
from advice <strong>on</strong> the best route to take given<br />
current traffic c<strong>on</strong>diti<strong>on</strong>s, to full aut<strong>on</strong>omy.<br />
While AI may appear to be a silver bullet<br />
that solves all of our transport challenges,<br />
it is nothing without data. From traffic<br />
flows to maintenance requirements, data<br />
is the fuel that AI needs to perform. It is in<br />
analysing big data faster than any human<br />
can that AI technologies can enable a<br />
vehicle to understand the world around<br />
it and make the right decisi<strong>on</strong> at the right<br />
time. It is AI combined with data that can<br />
cut journey times, c<strong>on</strong>gesti<strong>on</strong>, unnecessary<br />
polluti<strong>on</strong>, downtime for maintenance and<br />
ultimately, frustrati<strong>on</strong> due to time lost.<br />
The next piece in the puzzle is therefore<br />
sharing this data. Transport authorities,<br />
infrastructure managers, transport<br />
operators, manufacturers, service<br />
providers, platforms, final users and n<strong>on</strong>transport<br />
stakeholders are already doing<br />
so to improve operati<strong>on</strong>s and services,<br />
but this is a sensitive topic for many in the<br />
industry, so we want to develop a comm<strong>on</strong><br />
European mobility data space.<br />
Not all hurdles to AI are technological.<br />
The European Commissi<strong>on</strong>’s 2018 AI<br />
strategy raises the need to prepare for<br />
socio-ec<strong>on</strong>omic change; and to ensure an<br />
appropriate ethical and legal framework.<br />
A Communicati<strong>on</strong> <strong>on</strong> AI’s trustworthiness<br />
followed, designed to increase public<br />
acceptance while providing ethical<br />
guidelines for organisati<strong>on</strong>s pursuing AI.<br />
C<strong>on</strong>cerns range from safety and jobs to<br />
privacy and liability, and it is important<br />
that we recognise, c<strong>on</strong>sider and investigate<br />
each and every <strong>on</strong>e of these anxieties now.<br />
Technological progress will be worthless if<br />
our citizens are not <strong>on</strong> board.<br />
Our job in the Commissi<strong>on</strong> is to<br />
adapt our regulatory framework so that<br />
it supports innovati<strong>on</strong>, while at the same<br />
time ensuring respect for fundamental<br />
values and rights.<br />
The European Pillar of Social Rights,<br />
which calls for decent working c<strong>on</strong>diti<strong>on</strong>s,<br />
social protecti<strong>on</strong> and equal opportunities<br />
for all, is the Commissi<strong>on</strong>’s first resp<strong>on</strong>se<br />
to the challenges posed by new forms of<br />
work.<br />
We are also promoting skills<br />
development, as skills dictate employability.<br />
The New Skills Agenda for Europe also<br />
targets smoother transiti<strong>on</strong>s and reskilling<br />
for people changing jobs or careers.<br />
To improve understanding of the<br />
current transformati<strong>on</strong> process in the<br />
transport sector, the Commissi<strong>on</strong> is also<br />
funding research projects <strong>on</strong> the effects of<br />
automati<strong>on</strong> <strong>on</strong> the labour force, working<br />
c<strong>on</strong>diti<strong>on</strong>s, and skills requirements. And we<br />
are engaging with stakeholders, including<br />
our social partners. The results of a study<br />
analysing the situati<strong>on</strong> across transport<br />
modes and recommending future steps<br />
are due at the end of this year.<br />
Al<strong>on</strong>gside social c<strong>on</strong>cerns are those<br />
from industry: will the current safety<br />
legislative framework work for potential<br />
new safety risks? Will certificati<strong>on</strong> processes<br />
need to change to take account of machine<br />
learning? Will some traditi<strong>on</strong>al players see<br />
their role change? Will new alliances see<br />
others missing the AI boat?<br />
While I take these c<strong>on</strong>cerns seriously,<br />
I do not see them as blocking factors. It<br />
is clear that all stakeholders will need to<br />
collaborate much more than they have<br />
d<strong>on</strong>e, and to find their place in new value<br />
chains. Together, we will find the answers<br />
to these questi<strong>on</strong>s.<br />
Ultimately, the world is changing.<br />
Today, technology and innovati<strong>on</strong> are<br />
already challenging our habits and dayto-day<br />
lives and I welcome this! But it is<br />
not <strong>on</strong>ly technology that is changing. Our<br />
citizens will no l<strong>on</strong>ger accept the status<br />
quo. They no l<strong>on</strong>ger accept emissi<strong>on</strong>s or<br />
c<strong>on</strong>gesti<strong>on</strong>. They want change. And so do<br />
I. I want a modern, accessible, safe and<br />
c<strong>on</strong>nected transport system that is both<br />
sustainable – through decarb<strong>on</strong>isati<strong>on</strong> –<br />
and smart – through digitalisati<strong>on</strong>. The will<br />
and momentum for change is there, and<br />
AI will be a driver in making that change<br />
happen.
CONTENTS<br />
88<br />
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Editor’s Letter...............................................................................5<br />
By Jerry Zagoritis<br />
Editor and Publishing Partner of Our World, and the CEO of Campaign Lab<br />
Foreword......................................................................................7<br />
By Kyriakos Pierrakakis<br />
Greek Minister of Digital Governance<br />
Foreword......................................................................................9<br />
By Adina Vălean<br />
European Commissi<strong>on</strong>er for Transport<br />
AI bey<strong>on</strong>d tomorrow: transforming the state to cope with the<br />
challenges of tomorrow............................................................... 11<br />
By Ioan-Dragoş Tudorache<br />
Member of the European Parliament from Romania (Renew Europe), Chair of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
Politicizing the 5G debate is not good for Europe........................... 14<br />
By Abraham Liu<br />
Huawei Chief Representative to the EU Instituti<strong>on</strong>s<br />
We need to invest in our own digital sovereignty .......................... 16<br />
By Axel Voss<br />
Member of the European Parliament from Germany (EPP), Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
Digital Rights in the Age of Super <strong>Intelligence</strong>................................ 20<br />
By Eva Kaili<br />
Member of the European Parliament from Greece (S&D), Chair of the Panel for the Future of<br />
Science and Technology (STOA), Member of the Committee <strong>on</strong> Industry, Research and Energy,<br />
Substitute Member of the Special Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
<strong>Artificial</strong> <strong>Intelligence</strong>: a geopolitical and ec<strong>on</strong>omic game changer.... 24<br />
By Adam Bielan<br />
Member of the European Parliament from Poland (ECR), Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
Is “Open Strategic Aut<strong>on</strong>omy” … “artificial” or “intelligent”?.......... 26<br />
By Anna-Michelle Asimakopoulou<br />
Member of the European Parliament from Greece (EPP), Vice-Chair of the Committee<br />
<strong>on</strong> Internati<strong>on</strong>al Trade, Member of the Special Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a<br />
Digital Age<br />
An <strong>Artificial</strong> <strong>Intelligence</strong> For Human Well-Being............................. 28<br />
By Iban Garcia del Blanco<br />
Member of the European Parliament form Spain (S&D), Vice-Chair of the Committee <strong>on</strong><br />
Legal Affairs, Member of the Special Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
AI in the EU: Balancing Benefit & C<strong>on</strong>trol....................................... 30<br />
By Angeliki Dedopoulou<br />
Senior Manager for EU Public Affairs with Huawei Technologies<br />
Str<strong>on</strong>g rules for digital opini<strong>on</strong>-forming marketplaces.................... 32<br />
By Alexandra Geese<br />
Member of the European Parliament from Germany (Greens-European Free Alliance),<br />
Member of the Special Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age<br />
Teach AI in schools. Win the tech race in 2030............................... 34<br />
By K<strong>on</strong>stantinos Kyranakis<br />
Member of the Hellenic Parliament, former President of the Youth of the European<br />
People’s Party<br />
AI and Machine Learning in Digital Campaigning – will bots bring us<br />
closer than ever or will they ruin our system? ............................... 36<br />
By Hannes Kirchbaumer<br />
Founder and CEO of KIRCHBAUMER.COM, His main focus lies <strong>on</strong> digital marketing<br />
c<strong>on</strong>sulting and political campaigning<br />
<strong>Artificial</strong> intelligence’s great impact <strong>on</strong> low and<br />
middle-skilled jobs...................................................................... 38<br />
By Sybrand Brekelmans & Georgios Petropoulos<br />
Sybrand Brekelmans: Research assistant at Bruegel.<br />
Georgios Petropoulos: Marie Curie Skłodowska Research Fellow at MIT and Bruegel and<br />
post-doctoral fellow at the MIT Initiative <strong>on</strong> the Digital Ec<strong>on</strong>omy<br />
Global cooperati<strong>on</strong> is the <strong>on</strong>ly way forward.................................. 40<br />
By Luis Neves<br />
CΕΟ of the Global Enabling Sustainability Initiative (GeSI)<br />
The dynamics of data accumulati<strong>on</strong>.............................................. 42<br />
By Julia Anders<strong>on</strong><br />
Research Analyst, Bruegel
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
AI bey<strong>on</strong>d tomorrow: transforming the state<br />
to cope with the challenges of tomorrow<br />
By Ioan-Dragoş Tudorache<br />
Ioan-Dragoş<br />
Tudorache<br />
Member of the<br />
European Parliament<br />
from Romania (Renew<br />
Europe), Chair of the<br />
Special Committee <strong>on</strong><br />
<strong>Artificial</strong> <strong>Intelligence</strong> in<br />
a Digital Age<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
With the arrival of Ursula v<strong>on</strong> der Leyen<br />
and the “geopolitical Commissi<strong>on</strong>”<br />
focused <strong>on</strong> the twin strategic<br />
priorities of digital and green, the Brussels<br />
bubble has been c<strong>on</strong>tinuously abuzz with the<br />
trendiest of catchphrases: artificial intelligence,<br />
and its impact <strong>on</strong> humankind.<br />
Any new technology since the steam<br />
engine has shaped society. In the 18th and<br />
19th century, the industrial revoluti<strong>on</strong> ripped<br />
through the established order and transformed<br />
fundamentally how we did things without much<br />
foresight as to c<strong>on</strong>sequences. In 2020, we have<br />
a duty to better guide societal evoluti<strong>on</strong> instead<br />
of building the digital society out of inertia and<br />
path-dependency. We have a m<strong>on</strong>umental<br />
opportunity to get artificial intelligence right from<br />
the start, and to leverage it to build the society of<br />
tomorrow in a more c<strong>on</strong>scientious fashi<strong>on</strong>. For<br />
this, we need to make the right choices now and<br />
to open up avenues for making the right choices<br />
in the future.<br />
What I feel is missing in the debates in<br />
Brussels and elsewhere, is a vanguard visi<strong>on</strong><br />
for artificial intelligence -- a visi<strong>on</strong> that projects<br />
us decades into the future and gives us a sense<br />
of directi<strong>on</strong>. Without dismissing the important<br />
c<strong>on</strong>versati<strong>on</strong>s around AI that are <strong>on</strong> the agenda<br />
today, we need to start thinking of the “why” of<br />
tomorrow to complement the “what” and the<br />
“how” of today. We need to start preparing<br />
for successfully dealing with the systemic<br />
transformati<strong>on</strong>s AI will bring to our societies, and<br />
nowhere is this more pressing than in the way<br />
the State and its instituti<strong>on</strong>s functi<strong>on</strong>. That, to<br />
my mind, is a global c<strong>on</strong>versati<strong>on</strong> we should be<br />
having, and the EU can lead the way in planning<br />
for the transformati<strong>on</strong> of the State in line with<br />
the AI revoluti<strong>on</strong>, and through this, become a<br />
truly geopolitical superpower as it aspires to be.<br />
My premise is simple: the modern state,<br />
as shaped after the Peace of Westphalia,<br />
risks becoming the most helpless and likely<br />
future victim of artificial intelligence. With a<br />
slow, ossified, often inefficient and ineffective<br />
bureaucracy and administrati<strong>on</strong>, the state has<br />
but a single advantage in the face of all other<br />
structures that organize human activity: its size<br />
and its m<strong>on</strong>opolies <strong>on</strong> providing security and --<br />
to varying extents -- social services, as well as<br />
<strong>on</strong> writing and enforcing rules for how society<br />
and ec<strong>on</strong>omy work. But its m<strong>on</strong>opoly <strong>on</strong> power,<br />
which allowed it to withstand massive shocks<br />
throughout history (from wars to large-scale rise<br />
of disruptive technologies) can make things turn<br />
terribly wr<strong>on</strong>g when it comes to AI, and not <strong>on</strong>ly<br />
in the EU but anywhere else.<br />
<strong>Artificial</strong> intelligence, al<strong>on</strong>gside the great<br />
opportunities it brings for humans, comes<br />
loaded with the means for the demise or the<br />
perversi<strong>on</strong> of the state. Suprastatal structures,<br />
such as the EU, are just <strong>on</strong>e step removed from<br />
the direct threat states face, but the danger they<br />
face is of the same nature.<br />
A challenge from market forces<br />
The first threat to the state’s existence comes<br />
from market forces -- a challenge to its ec<strong>on</strong>omic<br />
viability. As data roams (almost) freely in the<br />
hands of private companies, and as AI algorithms<br />
are being developed around the world and<br />
are improving efficiency in every sector of the<br />
ec<strong>on</strong>omy imaginable, challenges to the state’s<br />
already inefficient delivery of core services are<br />
increasing. When AI-powered private health<br />
becomes cheaper and more effective than state<br />
provided healthcare, how will the state justify its<br />
expenditure <strong>on</strong> health? When private AI-powered<br />
educati<strong>on</strong> -- pers<strong>on</strong>alized, cutting edge, datadriven,<br />
digital-age-ready -- becomes the norm,<br />
how will the state compete? When, ultimately,<br />
private AI-assisted security makes citizens feel<br />
safe and protected, how will the state justify its<br />
existence? When all services the state provides<br />
will be better provided by an AI-powered private<br />
sector, what will give the state the authority to<br />
make and enforce laws and rules? And when<br />
private companies begin reinventing rules<br />
and providing alternative systems of behavior<br />
“governance” that become more important to<br />
citizens than the rules of the state, how will the<br />
state maintain its m<strong>on</strong>opoly <strong>on</strong> enforcement,<br />
11
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
arbitrati<strong>on</strong>, and the dispensati<strong>on</strong> of justice?<br />
Before the rise of artificial intelligence and the data ec<strong>on</strong>omy,<br />
citizens entrusted the state with exclusive competences because<br />
the state was the <strong>on</strong>ly <strong>on</strong>e large enough to provide certain<br />
services. The state was the <strong>on</strong>ly instituti<strong>on</strong> large enough to collect<br />
taxes, it was the <strong>on</strong>ly instituti<strong>on</strong> large enough to administrate<br />
natural m<strong>on</strong>opolies, and it was the <strong>on</strong>ly instituti<strong>on</strong> large enough to<br />
maintain a standing army to counter other states’ standing armies.<br />
Enter data, the new fuel -- or even future currency -- of the<br />
digital ec<strong>on</strong>omy. And <strong>on</strong> this data, specifically because of the<br />
state’s inherent vulnerability to perversi<strong>on</strong> from collecting data<br />
and building AI, the private sector has a cvasi-m<strong>on</strong>opoly. The<br />
tables are turned: private companies are now increasingly more<br />
effective at providing what the state was historically entrusted<br />
to provide. And while there is yet no company that can match<br />
a modern state (even if because the state still holds a citizensentrusted<br />
m<strong>on</strong>opoly <strong>on</strong> making and enforcing rules), the trend<br />
is here, at our doorstep. As more and more of human activity<br />
becomes data-driven, data-fueled, and data-dependent, how can<br />
the state adapt, so that it can compete and survive?<br />
The threat of state perversi<strong>on</strong><br />
The sec<strong>on</strong>d threat AI brings to the very existence of the state<br />
comes from the opposite directi<strong>on</strong>: it offers the right set of<br />
tools for the state to acquire incredible power over its citizens.<br />
The more the state collects, stores, and processes increasingly<br />
sophisticated and detailed data <strong>on</strong> their citizens with the help of<br />
increasingly sophisticated AI, the more the instituti<strong>on</strong>al setup that<br />
ensures its democratic -- or, arguably, any type of -- functi<strong>on</strong>ing<br />
becomes pr<strong>on</strong>e to unraveling. Massive advances in <strong>on</strong>e functi<strong>on</strong>al<br />
area -- such as, for example, the ability to predict every single<br />
citizens’ potential criminal behavior or, oppositely, to identify and<br />
reward every citizen’s law-abiding behavior -- shift the instituti<strong>on</strong>al<br />
balance of power and alter the power competiti<strong>on</strong> between<br />
instituti<strong>on</strong>s. Resource allocati<strong>on</strong> between instituti<strong>on</strong>s gradually<br />
shifts to instituti<strong>on</strong>s that are more effective in providing for the<br />
“wellbeing” of the citizens and, in time, the state transforms into<br />
a giant populati<strong>on</strong> c<strong>on</strong>trol machine, c<strong>on</strong>tinuously and more and<br />
more intrusively optimizing the lives of its citizens. This is but <strong>on</strong>e<br />
scenario, which does not even need a totalitarian state to come<br />
to pass. Nor does the state -- or its leaders -- need to c<strong>on</strong>sciously<br />
choose to “c<strong>on</strong>trol” citizens. Once the state becomes capable<br />
to profile every single <strong>on</strong>e of its “customers” -- behaviorally,<br />
psychologically, politically, and in a myriad other ways -- there is no<br />
turning back. The technology allows it, for the first time in human<br />
history. And, incidentally, the state is since times immemorial<br />
the default repository of the pers<strong>on</strong>al data of its citizens. So the<br />
temptati<strong>on</strong> is there.<br />
By 2030, the EU must define a new rulebook, fit for the digital<br />
age, by which the state needs to play in dealing with citizens’<br />
data and in deploying AI. Brussels is rightfully worried by facial<br />
recogniti<strong>on</strong> technology, predictive policing, social scoring, and<br />
other blatantly dangerous technologies that could be misused<br />
by public authorities, especially when these technologies combine<br />
with potential biases and discriminati<strong>on</strong>. But the road to Big<br />
Brother needs not be a straight line. We need to think carefully<br />
and to anticipate any possible negative outcomes from using<br />
citizens’ pers<strong>on</strong>al data and placing powerful AI in the hands of<br />
state authorities, even in the most benign use cases. For this,<br />
the state itself needs to be transformed, so that it is resilient to<br />
perversi<strong>on</strong> even if endowed with dangerous technological tools.<br />
We need to reinvent the checks-and-balances and rewrite the<br />
social c<strong>on</strong>tract for the digital age so that the state survives and<br />
c<strong>on</strong>tinues to serve its purpose.<br />
Fake news and disinformati<strong>on</strong><br />
The third challenge to the state brought about by artificial<br />
intelligence comes from without -- the direct attack <strong>on</strong> truth brought<br />
about by AI-powered disinformati<strong>on</strong> and fake-news. Once truth<br />
unravels social order unravels as well, and the state is threatened.<br />
While propaganda, disinformati<strong>on</strong>, mass manipulati<strong>on</strong>, and fake<br />
news are all as ancient as human civilizati<strong>on</strong>, AI acts as a catalyst<br />
that amplifies these threats bey<strong>on</strong>d the c<strong>on</strong>tainment threshold.<br />
This is an existential challenge, first and foremost, to democratic<br />
systems, because democracy is fueled by truth: citizens need to<br />
know the truth in order to vote, and the vote is the fundamental<br />
building block of democratic systems. The democratic state is<br />
bound to unravel <strong>on</strong>ce third parties (whether states or n<strong>on</strong>-state<br />
actors, internati<strong>on</strong>al or domestic) can sway the results of electi<strong>on</strong>s<br />
through powerful AI-enabled tools. We are not yet there, but the<br />
progressi<strong>on</strong> is also not linear: <strong>on</strong>ce such mass manipulati<strong>on</strong> tools<br />
become powerful enough that they are bey<strong>on</strong>d c<strong>on</strong>tainment, the<br />
failure of the democratic state will be abrupt, not incremental.<br />
Social order is relevant for the stability of all societies, not just<br />
democracies. AI-powered fake news and disinformati<strong>on</strong> can also<br />
bring to their knees n<strong>on</strong>-democratic states, including those very<br />
states who are now building and using such tools to advance<br />
their internati<strong>on</strong>al agenda. And, in the future, authoritarian states<br />
might be even more vulnerable to this threat than democracies. As<br />
democratic societies develop antibodies to deal with such direct<br />
attacks <strong>on</strong> their democratic functi<strong>on</strong>ing, authoritarian states are<br />
more likely to deploy mass-manipulati<strong>on</strong> tools internally in order<br />
to c<strong>on</strong>trol their citizens and to project a preferred versi<strong>on</strong> of reality.<br />
But when the proliferati<strong>on</strong> of powerful disinformati<strong>on</strong> AI will make<br />
it accessible to smaller actors, including domestic challengers,<br />
maintaining a fabricated versi<strong>on</strong> of the truth and c<strong>on</strong>tinuously<br />
fine-tuning it will become increasingly difficult.<br />
The stability of the state, democratic or not, depends <strong>on</strong><br />
truth or, at the very least, <strong>on</strong> the stability of the local “truth”. The<br />
progress of society, including the digital transformati<strong>on</strong>, must be<br />
fueled by truth, because scientific progress can <strong>on</strong>ly happen when<br />
facts remain unchanged. And the rules-based internati<strong>on</strong>al order<br />
is fueled by truth, because rules and decisi<strong>on</strong>s in internati<strong>on</strong>al fora<br />
rely <strong>on</strong> facts and the truth. AI-powered tools that can manipulate<br />
populati<strong>on</strong>s <strong>on</strong> a large scale and change the truth to influence<br />
mass-scale outcomes are a threat to all states and a threat to the<br />
future global stability.<br />
12 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Transforming the state to<br />
withstand the challenges of AI<br />
The answer to all three challenges -- the<br />
increasing competiti<strong>on</strong> from private<br />
companies in providing fundamental<br />
services, the state’s vulnerability to<br />
perversi<strong>on</strong> from within if misusing<br />
technology, and the challenge to social<br />
order posed by AI-powered mass<br />
manipulati<strong>on</strong> tools -- is in the reinventi<strong>on</strong><br />
of the state as the safest and most trusted<br />
digital platform for its citizens. But how?<br />
First, the state needs to be the first<br />
to provide citizens a secure, trusted,<br />
and inalienable mechanism to own and<br />
use their pers<strong>on</strong>al data, including in<br />
the interacti<strong>on</strong> with the state itself. The<br />
Commissi<strong>on</strong>’s “secure European e-identity”,<br />
foreshadowed by Ursula v<strong>on</strong> der Leyen in<br />
her State of the Uni<strong>on</strong> address, is a step<br />
in the right directi<strong>on</strong>, but it needs a boost<br />
in magnitude and ambiti<strong>on</strong>. Looking at<br />
2030, the EU must provide its states and<br />
therefore its citizens with a soluti<strong>on</strong> to not<br />
just the ownership and the use of their<br />
pers<strong>on</strong>al data, but to its “m<strong>on</strong>etizati<strong>on</strong>”.<br />
If people are given a mechanism through<br />
which they can stand to gain quantifiable<br />
state-backed benefits from their data,<br />
perhaps to the same extent to which the<br />
state guarantees currency as a carrier of<br />
ec<strong>on</strong>omic value, market forces will prop-up<br />
the state in the uneven competiti<strong>on</strong> with<br />
the private sector. The Googles and the<br />
Amaz<strong>on</strong>s of the world will have to adapt<br />
to increasing demands for value from the<br />
data they use, and the state will have a<br />
newfound role in protecting its citizens’<br />
rights.<br />
Sec<strong>on</strong>d, the state needs to transform<br />
itself so as to become a de facto (and<br />
powerful) standard-setter in the way<br />
pers<strong>on</strong>al data is used. The state needs<br />
to become itself a data-agnostic digital<br />
platform, in which citizens plug their data<br />
to access services or to derive ec<strong>on</strong>omic<br />
benefits. The state needs to become<br />
the selfless and trusted repository of<br />
its citizens data, just like a bank or a<br />
safehouse. Rules must be rewritten and<br />
the functi<strong>on</strong>ing of the states’ instituti<strong>on</strong>s<br />
must be rebalanced so as to guarantee<br />
citizens the inviolability of their data. And,<br />
just like the current state pays workers for<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
their time and skill through direct hires and<br />
public tenders, so should it set the t<strong>on</strong>e for<br />
the digital future by offering citizens direct<br />
ec<strong>on</strong>omic value in those instances when<br />
it does need to use, temporarily and for<br />
limited and clearly-defined purposes, their<br />
pers<strong>on</strong>al data.<br />
Third, the state needs to open itself<br />
up for everybody to benefit -- citizens<br />
and industry alike -- thus setting both<br />
transparency and informati<strong>on</strong> standards.<br />
By opening the increasingly massive<br />
amounts of industrial data it generates,<br />
the state can be a standard-setter and<br />
a catalyst for the data-driven industrial<br />
transformati<strong>on</strong>, fueling innovati<strong>on</strong> and<br />
progress. Allowing industry to benefit<br />
from public data and from co-creating<br />
data with public authorities, encouraging<br />
and facilitating the sharing and comm<strong>on</strong><br />
use of industrial data, and relying <strong>on</strong> the<br />
natural m<strong>on</strong>opoly <strong>on</strong> certain types of data<br />
to set standards, the state can fuel growth<br />
and also become a trusted platform for its<br />
ec<strong>on</strong>omic actors, not just for its citizens.<br />
And by opening up its data and acting as<br />
a neutral actor in an ecosystem fueled by<br />
SHUTTERSTOCK<br />
data and informati<strong>on</strong>, the state can also<br />
guarantee a c<strong>on</strong>sistent and empiricsdriven<br />
versi<strong>on</strong> of “the truth” that is harder<br />
to undermine, boosting its own immunity<br />
to disinformati<strong>on</strong> and manipulati<strong>on</strong><br />
attacks from without.<br />
Given the ec<strong>on</strong>omies of scale and<br />
interoperability requirements for such<br />
transformative changes to the very<br />
foundati<strong>on</strong>s of the state, the European<br />
Uni<strong>on</strong> has a crucial role to play here. The<br />
traditi<strong>on</strong>al “state” will become increasingly<br />
insignificant in a digital ec<strong>on</strong>omy where<br />
data flows at incredible speeds all across<br />
the globe. No <strong>on</strong>e state can manage such<br />
transformati<strong>on</strong> <strong>on</strong> its own, and no <strong>on</strong>e<br />
state would benefit from it, should it go at<br />
it al<strong>on</strong>e. While according to the principle of<br />
subsidiarity member states will ultimately<br />
be resp<strong>on</strong>sible for many of the genetic<br />
changes the digital age requires of them<br />
to ensure their survival, the Uni<strong>on</strong> is in the<br />
perfect positi<strong>on</strong> to provide the ultimate<br />
transformati<strong>on</strong> handbook. By 2030, we<br />
need to have the handbook ready -- and<br />
with it, a new set of rules for the role of the<br />
Uni<strong>on</strong> in the digital transformati<strong>on</strong>.<br />
13
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
ISTOCK<br />
Politicizing the 5G debate<br />
is not good for Europe<br />
By Abraham Liu<br />
It is often said “may you live in<br />
interesting times”.<br />
These last few m<strong>on</strong>ths and<br />
particularly these last few weeks and<br />
days have certainly been interesting<br />
times.<br />
It is now clear that, after shall we say,<br />
a slightly unusual American electi<strong>on</strong>,<br />
there will be a new US Administrati<strong>on</strong> in<br />
January.<br />
It’s still too early to tell what the<br />
policies of President Elect Biden and<br />
Vice President Elect Harris will be, but<br />
by the team they are assembling around<br />
them, it looks like we may be returning<br />
to a world of multilateral cooperati<strong>on</strong><br />
and engagement, which from a business<br />
leader’s point of view, is encouraging.<br />
We have <strong>on</strong>ly to look at this year’s<br />
experiences to see the fruits of such<br />
multilateralism. With the last decade<br />
of global collaborati<strong>on</strong> in the ICT<br />
sector, the quality of the networks has<br />
enabled important aspects of our lives<br />
to c<strong>on</strong>tinue; like teleworking, remote<br />
medicine, home educati<strong>on</strong> and business<br />
c<strong>on</strong>tinuity. Specifically, the ability to<br />
make high-quality video c<strong>on</strong>ference<br />
calls is a direct result of global efforts<br />
and cooperati<strong>on</strong> to develop a single 4G<br />
14 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
It looks like we may be returning to a world<br />
of multilateral cooperati<strong>on</strong> and engagement,<br />
which from a business leader’s point of view,<br />
is encouraging.<br />
Abraham Liu<br />
Huawei Chief<br />
Representative to the<br />
EU Instituti<strong>on</strong>s<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
network standard – before the outgoing US<br />
administrati<strong>on</strong> came to power.<br />
During the last administrati<strong>on</strong>’s term<br />
though, we have seen attempts by the US<br />
to threaten others through its technological<br />
dominance. This has had the effect of<br />
encouraging Europeans to talk about their<br />
digital sovereignty, and the resilience of<br />
their networks. Did you know for example<br />
that the overwhelming amount of your data<br />
storage, cloud services, semic<strong>on</strong>ductors<br />
and technological infrastructure is from<br />
the United States? Of the top ten cloud<br />
providers in the EU, <strong>on</strong>ly 3 can truly be<br />
classed European in origin. Huawei <strong>on</strong> the<br />
other hand, does not deal in your data, or<br />
have access to it, nor do we m<strong>on</strong>etize it. We<br />
are not even compelled to collect data if<br />
ordered to - I am assured that the relevant<br />
Chinese law does not have extraterritorial<br />
applicati<strong>on</strong> – unlike, ir<strong>on</strong>ically, the US<br />
“Cloud Act”. We can actually help Europe<br />
with its digital sovereignty and strategic<br />
aut<strong>on</strong>omy ambiti<strong>on</strong>s.<br />
But I believe this talk of where a tech<br />
company comes from, promoted by<br />
the current US administrati<strong>on</strong>, is at best<br />
missing the point, and at worst a form<br />
of “corporate racism”. This doesn’t do<br />
any<strong>on</strong>e any good – least of all European<br />
c<strong>on</strong>sumers. Huawei was born in China, but<br />
its technology is not fully Chinese – just like<br />
any European or US tech company the<br />
technology is multinati<strong>on</strong>al.<br />
Apple products have European and<br />
Chinese comp<strong>on</strong>ents. Nokia products<br />
have Chinese and US parts. Huawei kit<br />
has European and US elements. The<br />
noti<strong>on</strong> that you have to choose between<br />
Chinese and US or European is false – you<br />
are already choosing Chinese in everything<br />
you buy. Indeed, while trying to exclude<br />
Huawei from 5G networks <strong>on</strong> the basis that<br />
we are Chinese, both the US and Europe<br />
are increasing their trade with China<br />
and Chinese companies, and European<br />
companies are expanding their markets<br />
inside China as well.<br />
It is as a direct result of this c<strong>on</strong>nected<br />
and interdependent technology supply<br />
chain that many European and American<br />
companies are also being damaged by the<br />
policies of the current US administrati<strong>on</strong>.<br />
I am hopeful that this message will<br />
res<strong>on</strong>ate with new lawmakers in the US,<br />
and their counterparts here in Europe,<br />
because there are so many benefits that<br />
the new technological developments in<br />
AI, quantum and cloud computing can<br />
bring - not least to Europe’s envir<strong>on</strong>mental<br />
protecti<strong>on</strong> goals as spelled out in its Green<br />
Deal, and its recovery plans. A 5G network<br />
without Huawei is a more expensive, less<br />
technologically advanced network.<br />
It’s probably too optimistic to c<strong>on</strong>clude<br />
that a new US administrati<strong>on</strong> will overturn<br />
or change the most restrictive policies<br />
towards Huawei in the short term – or<br />
even in the medium term. But while we<br />
are planning for the worst, we are hoping<br />
for the best.<br />
For me, it would be enough that in<br />
the coming weeks and m<strong>on</strong>ths, we here<br />
in Europe, and in Huawei, live through a<br />
period of greater communicati<strong>on</strong>, more<br />
collaborati<strong>on</strong>, and increased cooperati<strong>on</strong><br />
across countries and c<strong>on</strong>tinents. That we<br />
live in slightly less interesting times!<br />
15
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
We need to invest in our<br />
own digital sovereignty<br />
Axel Voss<br />
Member of the<br />
European Parliament<br />
from Germany (EPP),<br />
Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong><br />
<strong>Intelligence</strong> in a Digital<br />
Age<br />
By Axel Voss<br />
digital transformati<strong>on</strong><br />
c<strong>on</strong>tinues to have a str<strong>on</strong>g<br />
“The<br />
impact <strong>on</strong> our daily life and<br />
has led to radical changes in almost<br />
every part of Europe’s ec<strong>on</strong>omy and<br />
society. Although the EU Digital Single<br />
Market strategy of 2014 was a first<br />
important step to make the European<br />
Uni<strong>on</strong> future-proof for the digital era,<br />
I believe that many relevant policy<br />
questi<strong>on</strong>s remain unanswered.”<br />
This is the beginning of my own<br />
digital manifesto published in January,<br />
calling <strong>on</strong> the European Commissi<strong>on</strong><br />
to quickly draw up its plans to further<br />
develop the Digital Single Market.<br />
Almost a year later, my statement still<br />
stands. The strategies that were built<br />
up were very important first steps - but<br />
if we do not follow up <strong>on</strong> it by heavily<br />
investing in our own digital sovereignty,<br />
we will end up as a digital col<strong>on</strong>y of the<br />
US or China, uncompetitive and without<br />
a voice of our own.<br />
Currently, n<strong>on</strong>e of the top 15 digital<br />
companies is European. There is no<br />
significant European operating system,<br />
browser, social media network or search<br />
engine. The investment gap compared<br />
to the USA and China is estimated at<br />
€190 billi<strong>on</strong> per year. Our resp<strong>on</strong>se to<br />
this challenge has been a wide range<br />
of incoherent, fragmented interim<br />
soluti<strong>on</strong>s following lengthy decisi<strong>on</strong>making<br />
processes. It seems that we<br />
need to jump <strong>on</strong> the moving train, yet<br />
we are still standing at the platform<br />
discussing which seat to take. Not <strong>on</strong>ly<br />
will this approach prevent us from ever<br />
catching up with a rapidly changing<br />
technological envir<strong>on</strong>ment, it may also<br />
give our citizens the impressi<strong>on</strong> that the<br />
European political class has lost c<strong>on</strong>trol<br />
– a percepti<strong>on</strong> that could ultimately<br />
result in a significant loss of trust in our<br />
democratic system.<br />
Hence, we need to go bey<strong>on</strong>d first<br />
step measures. We need to invest<br />
heavily in our digital infrastructure and<br />
technology “made in the EU”. Although<br />
the Commissi<strong>on</strong> is already promising<br />
16 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
investments, what we need is a mobilisati<strong>on</strong> of private and<br />
public investments at the scale of the European Green Deal.<br />
In fact, a transformati<strong>on</strong> of our digital infrastructure will also<br />
strengthen our sustainability and thus both major initiatives<br />
should go hand in hand.<br />
We also need to invest in digital competence and<br />
educati<strong>on</strong>. Right now, a graduate in data science who wants<br />
We need to invest heavily in<br />
our digital infrastructure and<br />
technology “made in the EU”.<br />
Although the Commissi<strong>on</strong> is<br />
already promising investments,<br />
what we need is a mobilisati<strong>on</strong><br />
of private and public<br />
investments at the scale of<br />
the European Green Deal. In<br />
fact, a transformati<strong>on</strong> of our<br />
digital infrastructure will also<br />
strengthen our sustainability<br />
and thus both major initiatives<br />
should go hand in hand.<br />
SHUTTERSTOCK<br />
to develop new technologies will move to the Silic<strong>on</strong> Valley.<br />
The EU has to become an attractive locati<strong>on</strong> for innovati<strong>on</strong><br />
if it wants to remain a large competitive ec<strong>on</strong>omy. And<br />
thirdly, we need to invest in start-ups and digital companies<br />
to scale up. Only 12 big new European companies were able<br />
to scale up since 1951, compared to 51 in the US. Thus, we<br />
need to provide easy access to venture capital and ease the<br />
administrative burden <strong>on</strong> SMEs and start-ups.<br />
To achieve this, I would like to see clear prioritisati<strong>on</strong>. We<br />
need to have a clear picture about our areas of strength that<br />
we should reinforce, our critical deficiencies that we should<br />
overcome and the upcoming disruptive technologies in which<br />
str<strong>on</strong>g investments make sense. That way, we can prioritise<br />
in which technologies we want to be leading competitors,<br />
whether its cloud computing, automated cars or digital<br />
finance, and start heavily investing in these sectors as flagship<br />
areas. Our objective should be to become less dependent <strong>on</strong><br />
n<strong>on</strong>-European technologies and services, while establishing<br />
sound ethical, technological and security standards.<br />
To put such measures in place, we need to change the way<br />
political processes and governmental systems work by making<br />
legislative procedures more effective. To achieve a truly<br />
harm<strong>on</strong>ised Digital Single Market, we need EU regulati<strong>on</strong>s<br />
instead of directives with c<strong>on</strong>stant impact assessment. It took<br />
us ten years from the first c<strong>on</strong>sultati<strong>on</strong>s for GDPR to its actual<br />
implementati<strong>on</strong>. We cannot afford these delays in a race with<br />
competitors like China.<br />
While I commend the proposals made by the European<br />
Commissi<strong>on</strong>, the clock is ticking. The scenario of becoming a<br />
digital col<strong>on</strong>y will put our competitiveness at risk for the l<strong>on</strong>g<br />
term. If we do not have a say in shaping the next technological<br />
stage, actors who do not share our core values and standards<br />
will dominate the process, with troubling c<strong>on</strong>sequences for<br />
our privacy, security and our prosperity. Our goal should be<br />
to strive for a European way of digitalisati<strong>on</strong> in a Digital Single<br />
Market 2.0: human-centred, value-oriented and based <strong>on</strong> the<br />
social market ec<strong>on</strong>omy, without becoming protecti<strong>on</strong>ist.<br />
New digital innovati<strong>on</strong> is not possible without internati<strong>on</strong>al<br />
cooperati<strong>on</strong> and free flow of data. However, cooperati<strong>on</strong><br />
needs to be built <strong>on</strong> trust and respect of our values.<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
17
Fearless<br />
fear·less|ˈfir-ləs<br />
NEW EUROPE
Do not c<strong>on</strong>sider Hellenic<br />
American College for studies,<br />
if you do not want to:<br />
• Become a critical thinker and problem solver<br />
who seeks a career, not just a job<br />
• Acquire excepti<strong>on</strong>al communicati<strong>on</strong> skills<br />
in English<br />
Undergraduate Programs<br />
B.A. in English Language and Literature<br />
B.S. in Business Administrati<strong>on</strong><br />
B.S. in Engineering with a major<br />
in Electrical Engineering<br />
B.S. in Informatics<br />
B.S. in Psychology<br />
Bachelor of Music<br />
Graduate Programs<br />
M.A. in Applied Linguistics - TESOL<br />
M.A. in C<strong>on</strong>ference Interpreting<br />
M.A. in Translati<strong>on</strong><br />
M.S. in Informatics<br />
M.S. in Psychology<br />
M.S. in Sales Management<br />
Master of Business Administrati<strong>on</strong><br />
Ph.D. in Language & Communicati<strong>on</strong><br />
• Be part of a student-centered academic<br />
envir<strong>on</strong>ment<br />
• Develop research and analytical skills<br />
• Be an active learner participating in various<br />
hands-<strong>on</strong> projects related to your field of<br />
study<br />
• Expose yourself to a multinati<strong>on</strong>al<br />
envir<strong>on</strong>ment that cultivates global citizens<br />
But if you do…We are the <strong>on</strong>e<br />
you are looking for!<br />
Degree Programs in cooperati<strong>on</strong> with<br />
Massalias 22, 10680 Athens, Greece, Tel: +30 210 3680950, info@haec.gr, www.haec.gr<br />
Hellenic American University is accredited by the New England Commissi<strong>on</strong> of Higher Educati<strong>on</strong> (NECHE).<br />
Hellenic American University’s degree programs are regulated, approved, and regularly m<strong>on</strong>itored by the New Hampshire Department of Educati<strong>on</strong>, Divisi<strong>on</strong> of Educator Support and Higher Educati<strong>on</strong> – Higher<br />
Educati<strong>on</strong> Commissi<strong>on</strong>.<br />
Hellenic American College and Hellenic American University do not discriminate <strong>on</strong> the basis of race, color, nati<strong>on</strong>al and ethnic origin, gender, sexual orientati<strong>on</strong>, age, religi<strong>on</strong>, physical disability, or veteran status<br />
in the administrati<strong>on</strong> of their educati<strong>on</strong>al policies, admissi<strong>on</strong>s policies, scholarship and loan programs, and athletic or other school-administered programs.
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Digital Rights in the Age<br />
of Super <strong>Intelligence</strong><br />
By Eva Kaili<br />
Eva Kaili<br />
Member of the<br />
European Parliament<br />
from Greece (S&D),<br />
Chair of the Panel for<br />
the Future of Science<br />
and Technology<br />
(STOA), Member of the<br />
Committee <strong>on</strong> Industry,<br />
Research and Energy,<br />
Substitute Member of<br />
the Special Committee<br />
<strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong><br />
in a Digital Age<br />
In our transforming world, digital technology<br />
has the critical mass to push our fr<strong>on</strong>tiers and<br />
release unlimited potential. As the wave of digital<br />
transformati<strong>on</strong> soars high, improving our lives,<br />
industries and ec<strong>on</strong>omies, we must not overlook<br />
the risks that technologies and innovati<strong>on</strong>s pose<br />
<strong>on</strong> the fairness and cohesi<strong>on</strong> of our societies, and<br />
our rights as European citizens.<br />
Through the c<strong>on</strong>structi<strong>on</strong> of vast digital<br />
infrastructures, which track, m<strong>on</strong>itor and collect<br />
pers<strong>on</strong>al data at scale, major <strong>on</strong>line platforms<br />
are commodifying our data. An entire ecosystem<br />
of apps and companies c<strong>on</strong>trols huge data flows<br />
and ‘treats’ technology users as data suppliers via<br />
the impositi<strong>on</strong> of default choice architectures.<br />
The current form of ‘computati<strong>on</strong>al governance’<br />
in which we are ruled by algorithms and predictive<br />
analytics which make use and trade of individual<br />
human experiences to produce predicti<strong>on</strong>s<br />
of human behaviour has largely remained<br />
unregulated. We can still be in c<strong>on</strong>trol of safe AI,<br />
trustworthy and complimentary to humans.<br />
AI is at the core of this wave of digitisati<strong>on</strong>;<br />
the power it carries must be tamed before it is<br />
unleashed in its next form of super intelligence.<br />
Manifested in both software and hardware<br />
with intelligence expressed in lines of code and<br />
powerful automati<strong>on</strong>, it is already being used to<br />
solve complex problems, discover patterns or<br />
predict traffic, and make suggesti<strong>on</strong>s to users. AI<br />
is also tested for completely aut<strong>on</strong>omous vehicles,<br />
as deep machine learning could make decisi<strong>on</strong>s<br />
faster, safer and better than people. Progress<br />
remains incremental, however the more we<br />
advance our understanding and the further we<br />
develop this technology, the closer we reach the<br />
point where AI leaps from narrow to general,<br />
and then super intelligence. Narrow AI is already<br />
making simple suggesti<strong>on</strong>s and decisi<strong>on</strong>s at the<br />
service of people; however, the next level of AI<br />
sophisticati<strong>on</strong> can completely omit humans from<br />
the process and reach levels that humans will not<br />
be capable to understand or c<strong>on</strong>trol anymore.<br />
As most of human activity and industry has<br />
moved <strong>on</strong>line during the pandemic, and we<br />
become more dependent <strong>on</strong> an increasing amount<br />
of digital services, a series of critical questi<strong>on</strong>s<br />
arise: How can we build data infrastructures that<br />
recognise the role of data as a public good? Can<br />
we foresee an impartial system of data governance<br />
that could rec<strong>on</strong>cile the ast<strong>on</strong>ishing potential of<br />
these technologies with their significant human<br />
downsides? Do we as citizens know that when we<br />
use a search engine to find something <strong>on</strong>line, the<br />
engine is learning by searching us as well? Does the<br />
collecti<strong>on</strong> and processing of behavioural data by<br />
these technology platforms threaten to dismantle<br />
democracy and undermine the rule of law?<br />
To tackle these questi<strong>on</strong>s we need to come up<br />
with novel, innovative resp<strong>on</strong>ses, so that we make<br />
our future better than the grim outlook predicted in<br />
Black Mirror, the popular Netflix series. We have to<br />
be aware and cautious in designing the appropriate<br />
framework for AI; intelligent systems are used <strong>on</strong><br />
the pandemic fr<strong>on</strong>t for rapid COVID diagnosis<br />
and prognosis based <strong>on</strong> AI-enabled CT scanning<br />
analysis. AI systems are used now to upgrade our<br />
mapping systems, the management of resources,<br />
to make our mobility smarter, our agriculture more<br />
precise, with predictive maintenance and smart<br />
sensors that can c<strong>on</strong>tribute to a greener future,<br />
saving time, energy, and resources.<br />
As the impact of exp<strong>on</strong>ential technologies<br />
grows, a new framework is needed in Europe to<br />
harness its benefits and mitigate its risks. This<br />
framework must be built <strong>on</strong> ethical principles and<br />
binding standards, which elevate people’s trust in<br />
AI and ensure that in the digital age, people co-exist<br />
with intelligent systems without fearing exclusi<strong>on</strong>,<br />
manipulati<strong>on</strong>, oppressi<strong>on</strong> or discriminati<strong>on</strong>.<br />
20 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
SHUTTERSTOCK<br />
Retaining freedom of choice in a humancentric<br />
AI that would prevent brain<br />
computer interfaces challenging the nature<br />
and future of humanity. In c<strong>on</strong>trast to the<br />
trends of the Fourth Industrial Revoluti<strong>on</strong><br />
towards inequalities and dehumanizati<strong>on</strong>,<br />
technology and innovati<strong>on</strong> best practices<br />
need now to be bent back towards the<br />
service of humanity, and Europe could<br />
lead as a global rules and standards setter<br />
for the Fifth Industrial Revoluti<strong>on</strong>.<br />
The European principle-based<br />
framework for AI systems must translate<br />
and establish by law, respect to our rights<br />
in the digital age. Progress has been made<br />
with landmark initiatives such as the GDPR,<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
however more work is needed in this space<br />
to establish legally binding rules which<br />
provide developers and innovators with<br />
more legal clarity, and award individuals<br />
str<strong>on</strong>ger protecti<strong>on</strong>s of their rights and<br />
freedoms. At the foundati<strong>on</strong>al level,<br />
this framework must guarantee higher<br />
transparency and accountability, define<br />
the liability of AI systems, establishing<br />
standards to trace, audit, explain, appeal<br />
and reverse decisi<strong>on</strong>s made by AI during<br />
its entire lifecycle.<br />
The rapid development of automati<strong>on</strong><br />
in Europe must not reflect mistakes of<br />
the past; AI algorithms and systems must<br />
be trained <strong>on</strong> diversified sets, and their<br />
objectives have to be clearly defined<br />
and c<strong>on</strong>trolled to avoid risks of bias and<br />
discriminati<strong>on</strong> or data pois<strong>on</strong>ing. Decisi<strong>on</strong>s<br />
made by AI must be aligned with the<br />
collective ethical fabric that defines Europe<br />
throughout the lifecycle of intelligent<br />
systems and follow clear red lines, such<br />
as the risk-based approach that would,<br />
for example, completely ban research in<br />
aut<strong>on</strong>omous lethal weap<strong>on</strong>s or c<strong>on</strong>scious<br />
AI. The high-risk applicati<strong>on</strong>s should alert<br />
us of the fragility of our comm<strong>on</strong> values,<br />
as well as our rights as citizens, our<br />
resp<strong>on</strong>sibilities as policy-makers and our<br />
obligati<strong>on</strong>s to future generati<strong>on</strong>s.<br />
21
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Str<strong>on</strong>g protecti<strong>on</strong>s can enable the development of<br />
ethical AI in Europe, and force gatekeepers to respect with<br />
reciprocity the new architecture of data governance. The<br />
data ecosystem is ground-zero for AI innovati<strong>on</strong> in Europe.<br />
Data is the source of learning for AI systems and therefore<br />
very important for AI to become smarter. Privacy by design,<br />
c<strong>on</strong>sumer protecti<strong>on</strong>, and product safety will be at the top of<br />
our agenda since we will not compromise privacy for safety.<br />
More opti<strong>on</strong>s available to users, clear c<strong>on</strong>sent, and rewards<br />
for the use of our data, with str<strong>on</strong>ger security standards, as in<br />
cases of extreme pers<strong>on</strong>alizati<strong>on</strong> and micro targeting, would<br />
enable fairer AI in this evolving ecosystem, and inspire trust<br />
for European citizens. The EU has expertise in leading efforts<br />
to establish global norms and standards for data processing,<br />
which it can leverage in collaborati<strong>on</strong> with internati<strong>on</strong>al<br />
partners to achieve global c<strong>on</strong>sensus. In the European<br />
Parliament, we are currently working <strong>on</strong> an ambitious<br />
agenda to enable better, rules-based data collecti<strong>on</strong> and<br />
safer mechanisms to process, store, exchange, and transfer<br />
data. Standardisati<strong>on</strong> is necessary to ensure the high quality<br />
and safety of data, al<strong>on</strong>gside the establishment of data<br />
spaces governed by rules addressing the divergent levels of<br />
inherent risk of different sectors and technologies. Coupling<br />
these efforts with a secure digital ID can be instrumental in<br />
rethinking the current data ownership models.<br />
In the EU, a significant porti<strong>on</strong> of the budget and other<br />
funding instruments is earmarked for the development of<br />
digital skills, and digital educati<strong>on</strong> systems fit for purpose,<br />
as we must prepare the grounds for a seamless transiti<strong>on</strong><br />
of our societies and ec<strong>on</strong>omies into an era powered by<br />
intelligent systems. At both the nati<strong>on</strong>al and the EU levels,<br />
progress is necessary to calibrate educati<strong>on</strong> and professi<strong>on</strong>al<br />
training with the needs for digital skills, to enable greater<br />
understanding of AI across generati<strong>on</strong>s, and stimulate career<br />
paths into future jobs of AI innovati<strong>on</strong> and research.<br />
Our visi<strong>on</strong> is to foster technology that creates value for<br />
all, and enables our society to progress without impeding<br />
<strong>on</strong> the rights of Europeans. Striking this sensitive balance is<br />
a complicated endeavour, however the accelerati<strong>on</strong> of AI<br />
development, and the reach of its impact, implore us to get it<br />
right from the outset. The EU is a global champi<strong>on</strong> of privacy<br />
and data protecti<strong>on</strong> in the digital age. This achievement<br />
must be leveraged in establishing a safe ecosystem, where<br />
trustworthy AI enables us to achieve our objectives with<br />
regulatory c<strong>on</strong>trol, and establish a fair and equal society fit<br />
for the future.<br />
The European principle-based<br />
framework for AI systems must<br />
translate and establish by law,<br />
respect to our rights in the digital<br />
age. Progress has been made<br />
with landmark initiatives such as<br />
the GDPR, however more work is<br />
needed in this space to establish<br />
legally binding rules which provide<br />
developers and innovators with more<br />
legal clarity, and award individuals<br />
str<strong>on</strong>ger protecti<strong>on</strong>s of their rights<br />
and freedoms.<br />
22 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
<strong>Artificial</strong> <strong>Intelligence</strong>: a geopolitical<br />
and ec<strong>on</strong>omic game changer<br />
Adam Bielan<br />
Member of the<br />
European Parliament<br />
from Poland (ECR),<br />
Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong><br />
<strong>Intelligence</strong> in a Digital<br />
Age<br />
By Adam Bielan<br />
<strong>Artificial</strong> intelligence (AI) has been<br />
developing and broadening<br />
rapidly. It has increasingly<br />
become over the last years <strong>on</strong>e of the<br />
pillars of digital transformati<strong>on</strong>. AI’s<br />
influence now entails numerous sectors<br />
such as health, educati<strong>on</strong>, transport,<br />
cybersecurity, or research. AI is not <strong>on</strong>ly<br />
about our ec<strong>on</strong>omies, scaling-up effects<br />
or efficiency gains. As we are currently<br />
witnessing an escalating global race for<br />
digital leadership, it has turned into a<br />
geopolitical subject with increasingly<br />
high stakes.<br />
While lagging behind the tech<br />
giants, notably the American and<br />
Chinese digital empires, the EU now<br />
seeks to take part in the formulati<strong>on</strong><br />
of an internati<strong>on</strong>al AI approach and<br />
of its principles. In this c<strong>on</strong>text, the<br />
EU is highlighting the importance of<br />
strategic partnerships with key actors<br />
such as China, Japan and the USA, and<br />
stresses the significance of cooperating<br />
with regi<strong>on</strong>al organizati<strong>on</strong>s like the<br />
OECD and the Council of Europe. The<br />
EU Member States can indeed act as a<br />
balancing power between two visi<strong>on</strong>s<br />
for future AI developments. Critical<br />
ethical questi<strong>on</strong>s are <strong>on</strong> the table and<br />
an EU definiti<strong>on</strong> of AI directi<strong>on</strong>s thus can<br />
become an alternative to data models<br />
driven by state security and c<strong>on</strong>trol<br />
interests like in China, and str<strong>on</strong>gly<br />
influenced by the GAFAM interests<br />
and other large Internet platforms and<br />
industries like in the US. The safeguard<br />
of privacy and human rights is indeed<br />
<strong>on</strong>e of the core elements of Member<br />
States’ and the EU’s digital policy to the<br />
benefit of citizens.<br />
The European Uni<strong>on</strong> (EU), through<br />
new programs, increased funds and<br />
a new agenda <strong>on</strong> the matter, wishes<br />
to better address AI challenges and<br />
implement a new strategy vis à vis the<br />
24 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
opportunities <strong>Artificial</strong> <strong>Intelligence</strong> represents. Indeed, the<br />
recent years have seen the end of inacti<strong>on</strong> and the EU is<br />
now catching up with the ‘big giants’. Positive developments<br />
include the White Paper <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> published by<br />
the Commissi<strong>on</strong> and the set-up of a new Special Committee<br />
<strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in a Digital Age (AIDA) by the European<br />
Parliament. The latter recently adopted reports <strong>on</strong> ethical<br />
aspects, civil liability regime and intellectual property rights.<br />
SHUTTERSTOCK<br />
I believe this is the right directi<strong>on</strong> for citizens of the European<br />
Uni<strong>on</strong>.<br />
On top of this, the Covid-19 pandemic has underlined the<br />
importance for EU Member States and the European Uni<strong>on</strong><br />
in general to boost its investments in several areas in order<br />
to remain an epicenter of the geopolitical chessboard and a<br />
place for innovati<strong>on</strong> and ec<strong>on</strong>omic growth.<br />
C<strong>on</strong>cerning ec<strong>on</strong>omic perspectives, some progress has<br />
undeniably been made. Nevertheless, we still need to focus<br />
<strong>on</strong> skills, digital literacy, job creati<strong>on</strong> and to address job<br />
losses and people left behind. Digital inequalities are a more<br />
important than ever.<br />
Small and Medium Enterprises (SMEs) are the backb<strong>on</strong>e of<br />
the European ec<strong>on</strong>omy. They are the most challenged when<br />
new policies come into force and are a spring for innovati<strong>on</strong><br />
and growth. The EU Member States and instituti<strong>on</strong>s have kept<br />
this in mind, with plans to accompany SMEs to adapt their<br />
procedures, their business models and their ways of working<br />
using AI.However, much remains to be d<strong>on</strong>e in order to make<br />
this shared visi<strong>on</strong> a reality. Most notably, the necessity to<br />
have a clear and comprehensive framework is evident. A<br />
potentially complex regulatory framework will be much<br />
harder to digest for European SMEs, which lack extensive law<br />
departments compared to global players. Therefore, further<br />
acti<strong>on</strong>s are needed to bridge that gap. Digital Innovati<strong>on</strong> Hubs<br />
and Regulatory Sandboxes can represent in that regard a<br />
positive development, and administrative burden reducti<strong>on</strong><br />
policies should be a priority <strong>on</strong> our agenda.<br />
Also, I believe it is necessary to provide start-ups and SMEs<br />
access with the widest possible range of data and algorithm<br />
libraries as the lack of such a possibility is slowing down their<br />
development, which may eventually not be fast enough to<br />
challenge their internati<strong>on</strong>al competitors. Available data,<br />
together with its usability, accessibility and interoperability,<br />
is essential for a « culture of data » to grow in the European<br />
Uni<strong>on</strong>. The fragmentati<strong>on</strong> of European data has l<strong>on</strong>g<br />
prevented access to available data, with each Member State<br />
tending to work <strong>on</strong> its own nati<strong>on</strong>al data plan. This c<strong>on</strong>trasts<br />
with China, whose populati<strong>on</strong> and country size have early <strong>on</strong><br />
encouraged cloud and data storage practices.<br />
In additi<strong>on</strong> to SMEs, <strong>Artificial</strong> <strong>Intelligence</strong> could also better<br />
c<strong>on</strong>nect and mutually benefit the public sector. Strengthening<br />
collaborati<strong>on</strong> between the public sector and the small and<br />
medium enterprises can take many forms. I have for instance<br />
proposed a few m<strong>on</strong>ths ago a pilot project with fellow<br />
MEPs establishing the « EU GovTech platform », designed<br />
to bring the best out of well-functi<strong>on</strong>ing nati<strong>on</strong>al GovTech<br />
frameworks to the benefit of the entire single market. These<br />
frameworks enable the aggregati<strong>on</strong> of EU and nati<strong>on</strong>al public<br />
instituti<strong>on</strong>s’ demand for modern technologies. This means<br />
increasing access of EU’s innovative companies to the public<br />
procurement market.<br />
We, as European lawmakers, will require strenuous efforts<br />
to deliver innovati<strong>on</strong>-friendly and future-proof Single Market<br />
legislati<strong>on</strong> that remains both understandable for citizens<br />
and flexible for businesses. This is also our priority in AIDA<br />
committee. The envir<strong>on</strong>ment in which we find ourselves<br />
represents an opportunity for lawmakers and nati<strong>on</strong>al<br />
leaders to prove how serious we are in supporting SMEs.<br />
This is not <strong>on</strong>ly an ec<strong>on</strong>omic matter, defining the directi<strong>on</strong>s<br />
we want to lead our ec<strong>on</strong>omies. These questi<strong>on</strong>s are also<br />
intrinsically linked to a global scene in which the EU is trying to<br />
make a stand. That is why we need to step up our efforts, as<br />
AI, its framework and its development in the coming years will<br />
undeniably play a part in defining tomorrow’s global order.<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
25
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Is “Open Strategic Aut<strong>on</strong>omy”<br />
… “artificial” or “intelligent”?<br />
By Anna-Michelle Asimakopoulou<br />
Anna-Michelle<br />
Asimakopoulou<br />
Member of the<br />
European Parliament<br />
from Greece (EPP), Vice-<br />
Chair of the Committee<br />
<strong>on</strong> Internati<strong>on</strong>al Trade,<br />
Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong><br />
<strong>Intelligence</strong> in a Digital<br />
Age<br />
Paving the digital path<br />
for Europe should<br />
involve a human centric,<br />
transparent, trustworthy<br />
and inclusive artificial<br />
intelligence strategy. One<br />
that protects citizens and<br />
businesses in Europe and<br />
fosters innovati<strong>on</strong><br />
and competitiveness.<br />
At the 2020 State of the<br />
Uni<strong>on</strong> address, European<br />
Commissi<strong>on</strong> President Ursula<br />
v<strong>on</strong> der Leyen has unveiled her<br />
ambitious goal to make this decade<br />
“Europe’s Digital Decade”.<br />
The debate over digital<br />
developments is becoming more<br />
and more lively in Europe, and the<br />
accelerati<strong>on</strong> of the pace of digitizati<strong>on</strong><br />
as a result of the COVID-19 pandemic<br />
is shaping the “new normal”. The<br />
urgent need to analyze vast amounts<br />
of health data to understand and<br />
c<strong>on</strong>trol the virus has prompted us<br />
to invest more heavily in artificial<br />
intelligence and data management.<br />
The pandemic has also brought<br />
to light Europe’s overdependence in<br />
several critical and strategic sectors.<br />
Thus, the EU instituti<strong>on</strong>s were “forced”<br />
to revisit Europe’s industrial strategy<br />
in order to offer greater resilience<br />
and diversificati<strong>on</strong> of supply chains in<br />
strategic areas such as health supplies<br />
and medicine. As a result, Europe is<br />
today, more committed to becoming<br />
“strategically aut<strong>on</strong>omous”.<br />
<strong>Artificial</strong> <strong>Intelligence</strong> is at the<br />
forefr<strong>on</strong>t of the digital power race.<br />
The debate over Αrtificial Ιntelligence<br />
encompasses technological, ethical,<br />
legal, socio-ec<strong>on</strong>omic as well as<br />
geopolitical aspects. The OECD, the<br />
Council of Europe, the EU, the US,<br />
China and many other countries have<br />
developed their own strategies and<br />
guidelines <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong>.<br />
In additi<strong>on</strong> to the many standing<br />
committees dealing with AI in the<br />
European Parliament, the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong> in<br />
the Digital Age (AIDA) was created to<br />
take a horiz<strong>on</strong>tal approach in setting<br />
a l<strong>on</strong>g-term roadmap for AI in Europe.<br />
Europe is working hard to be<br />
at the forefr<strong>on</strong>t of the technology<br />
revoluti<strong>on</strong> and to ensure the<br />
competitiveness of its technology<br />
businesses. It strives to find the right<br />
balance, which promotes innovati<strong>on</strong><br />
and competitiveness, builds trust<br />
and protects our rights, values and<br />
principles. Europe aims to provide<br />
a trustworthy digital envir<strong>on</strong>ment to<br />
our citizens and businesses to shape<br />
the c<strong>on</strong>diti<strong>on</strong>s for the development<br />
and use of AI. This is <strong>on</strong>e of the main<br />
objectives of the European Strategy<br />
for Data and the Regulati<strong>on</strong> <strong>on</strong> Data<br />
Governance recently proposed by the<br />
European Commissi<strong>on</strong>, building up<strong>on</strong><br />
existing regulati<strong>on</strong>s such as pers<strong>on</strong>al<br />
data protecti<strong>on</strong> (GDPR), c<strong>on</strong>sumer<br />
protecti<strong>on</strong> and competiti<strong>on</strong> rules.<br />
Meanwhile, China is investing<br />
billi<strong>on</strong>s in AI, 5G, quantum computing<br />
and chip fabricati<strong>on</strong>, as part of a<br />
strategy, launched in July 2017, to<br />
establish itself as the world’s AI<br />
leader. By the year 2030, it expects<br />
to have a domestic AI industry worth<br />
at least $150 billi<strong>on</strong>. It steals IPR, tries<br />
to dominate internati<strong>on</strong>al standard<br />
setting bodies and lures countries<br />
into its techno orbit through the<br />
“digital silk road” initiative. China has<br />
1,4 bn people to produce data, and<br />
its’ own robust technology giants<br />
which it subsidizes with state aid<br />
and uses their products in the Party’s<br />
surveillance state.<br />
This potential of new<br />
“dependencies” namely <strong>on</strong> holders<br />
26 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
ISTOCK<br />
of certain critical technologies who use<br />
them in an “autocratic” manner, as well<br />
as <strong>on</strong> those c<strong>on</strong>trolling large volumes<br />
of data is why, Europe is striving to<br />
create its’ own “open yet sovereign<br />
Single Market for Data” which is crucial<br />
for AI development. Therefore, AI can<br />
be c<strong>on</strong>sidered as a “key” to achieving<br />
European “digital sovereignty” as part of<br />
the overall strategy of “Open Strategic<br />
Aut<strong>on</strong>omy.” As Commissi<strong>on</strong> Vice-<br />
President Maros Sefcovic emphasized<br />
“we can make Europe more resilient by<br />
boosting our open strategic aut<strong>on</strong>omy<br />
and building a fairer, climate-neutral and<br />
digitally sovereign future.”<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
There is much discussi<strong>on</strong> about<br />
what the meaning of “Open Strategic<br />
Aut<strong>on</strong>omy” actually is. Is it legitimate<br />
protecti<strong>on</strong> or undue and covert<br />
“protecti<strong>on</strong>ism?” “Aut<strong>on</strong>omy”, or<br />
“αυτονομία” in Greek, comes from the<br />
words αὐτός (autós, “self”) and νόμος<br />
(nómos, “law”). It literally means being<br />
free to make your own rules. Actually, It<br />
is your right to be able to make your own<br />
rules. Making these rules in Europe has<br />
produced our comm<strong>on</strong>ly shared values:<br />
democracy, the rule of law, and respect<br />
for human rights.<br />
Paving the digital path for Europe<br />
should involve a human centric,<br />
transparent, trustworthy and inclusive<br />
artificial intelligence strategy. One<br />
that protects citizens and businesses<br />
in Europe and fosters innovati<strong>on</strong><br />
and competitiveness. Becoming<br />
“protecti<strong>on</strong>ist” would not <strong>on</strong>ly be<br />
“artificial” it would also be untenable.<br />
A digitally sovereign or strategically<br />
aut<strong>on</strong>omous Europe, must be open<br />
to the world. It must create practical<br />
alliances with like-minded partners to<br />
set internati<strong>on</strong>al standards, and create<br />
rules based <strong>on</strong> comm<strong>on</strong>ly shared<br />
values. This is not <strong>on</strong>ly the “intelligent”<br />
choice, it is also the <strong>on</strong>ly democratic<br />
<strong>on</strong>e.<br />
27
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
An <strong>Artificial</strong> <strong>Intelligence</strong><br />
For Human Well-Being<br />
By Iban Garcia del Blanco<br />
SHUTTERSTOCK<br />
The massive rollout of artificial<br />
intelligence (AI) entails a<br />
technological leap that will cause<br />
(that is already causing) very substantial<br />
changes in the labor market, in the<br />
relati<strong>on</strong>ship with public authorities, in<br />
pers<strong>on</strong>al relati<strong>on</strong>ships and even in our<br />
own domestic life.<br />
Any technological progress carries<br />
benefits and risks. When Dworkin spoke<br />
of science, he highlighted its ambivalence<br />
both as a promise and as a threat. When<br />
we talk about artificial intelligence, we talk<br />
about benefits and/or risks <strong>on</strong> a scale not<br />
previously known, taking into account its<br />
intrinsic power. Due to its characteristics<br />
and its potential, this technology will take<br />
us through dizzying changes that would<br />
otherwise take generati<strong>on</strong>s. Fortunately,<br />
AI provides us with powerful tools to<br />
better address the major challenges of<br />
our time: the fight against climate change,<br />
the fight against depopulati<strong>on</strong>; and even<br />
to anticipate future pandemics such<br />
as COVID19, which has put up against<br />
the ropes our resilience as societies,<br />
and that we can now better address by<br />
accelerating research <strong>on</strong> drugs, vaccines,<br />
and developing state-of-the-art tracking<br />
applicati<strong>on</strong>s.<br />
As the regulatory superpower it claims<br />
to be, Europe must lead the way globally.<br />
I champi<strong>on</strong> a broad pact between the<br />
public and private sectors, each within<br />
the scope of their competences, that<br />
enables the unimpeded promoti<strong>on</strong> of<br />
technological development <strong>on</strong> the <strong>on</strong>e<br />
hand, and an ecosystem of trust for the<br />
European citizens’ peace of mind <strong>on</strong> the<br />
other. The ultimate goal of AI can <strong>on</strong>ly be<br />
the improvement our societies and our<br />
citizens’ lives.<br />
Early this year, the European<br />
Commissi<strong>on</strong> published a White Paper <strong>on</strong><br />
artificial intelligence together with a digital<br />
strategy –and by the beginning of 2021,<br />
it intends to release an all-encompassing<br />
regulati<strong>on</strong> <strong>on</strong> the phenomen<strong>on</strong>. In this<br />
regard, the European Parliament is <strong>on</strong>e<br />
step ahead: I had myself the h<strong>on</strong>or of<br />
being the rapporteur for the first European<br />
legislative initiative <strong>on</strong> the “Ethical aspects<br />
of artificial <strong>Intelligence</strong>, robotics and related<br />
technologies”, approved this past October<br />
with an extraordinary backing by the<br />
plenary of the Parliament, with the intent<br />
to indicate the Commissi<strong>on</strong> our ideas and<br />
positi<strong>on</strong>s related to the future regulati<strong>on</strong>.<br />
As noted above, Europe wants and can<br />
pi<strong>on</strong>eer the legal establishment of an ethical<br />
threshold that both protects European<br />
citizens from the possible adversities that<br />
this technological evoluti<strong>on</strong> entails, and<br />
28 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
provides an added value of trust to European AI in<br />
the world. An ethical threshold c<strong>on</strong>sistent with our<br />
European principles and values, reflected in the<br />
Charter of Fundamental Rights of the European<br />
Uni<strong>on</strong>, and c<strong>on</strong>gruent with our civilizing project.<br />
A legislati<strong>on</strong> inspired by a humanistic approach<br />
and focused <strong>on</strong> technological development. A<br />
regulati<strong>on</strong> applicable to both, AI developed in<br />
Europe or willing to operate in the Uni<strong>on</strong>, and with<br />
the aim of becoming a basic framework shared<br />
throughout the world <strong>on</strong> minimum requirements<br />
in the development and use of this technology.<br />
The planning of the sector should not be<br />
left off solely in the hands of the market. From<br />
the Parliament, we assert the need for public<br />
participati<strong>on</strong> that safeguards objectives bey<strong>on</strong>d<br />
the possible ec<strong>on</strong>omic profitability, so that the<br />
measurable aspects in terms of social profitability<br />
are also the subject of research and development:<br />
better public services, social resp<strong>on</strong>sibility,<br />
envir<strong>on</strong>mental sustainability, gender equality...<br />
I also perceive as essential a public-private<br />
collaborati<strong>on</strong> that allows mobilizing the sufficient<br />
amount of resources needed to catch up with<br />
the investment efforts d<strong>on</strong>e in other parts of the<br />
world -such is the strategy that the Commissi<strong>on</strong><br />
itself proposes in its White Paper.<br />
From a regulatory viewpoint, it is clear that<br />
the Uni<strong>on</strong> understands the need to establish<br />
a demanding framework for safeguarding<br />
citizens’ rights, as well as the essential principles<br />
c<strong>on</strong>tained in the Lisb<strong>on</strong> Treaty. N<strong>on</strong>etheless, we<br />
also believe that certain margin of maneuver is<br />
essential for the operators, and that they are the<br />
<strong>on</strong>es who assume the need to bind themselves<br />
with an ethical model in the development and<br />
deployment. Legislators are aware of the need not<br />
to establish an oppressive regulatory framework<br />
that stifles the initiative, and also that it will be<br />
impossible to understand such a complex and<br />
changing reality without a flexible framework<br />
that does not become obsolete with the arrival<br />
of the following innovati<strong>on</strong>. Likewise, <strong>on</strong>e of the<br />
main objectives is to avoid the fragmentati<strong>on</strong> of<br />
the European digital internal market, while also<br />
complying with the sacred principle of subsidiarity.<br />
That is why the report approved by the Parliament<br />
designs a top-down governance model: a<br />
European coordinating body able to harm<strong>on</strong>ize<br />
the regulatory development throughout the<br />
Uni<strong>on</strong> and to adapt quickly to technological<br />
developments; and entities resp<strong>on</strong>sible for<br />
administering and enforcing standards at the<br />
nati<strong>on</strong>al level. There has been an intense debate<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
Iban Garcia del<br />
Blanco<br />
Member of the<br />
European Parliament<br />
form Spain (S&D), Vice-<br />
Chair of the Committee<br />
<strong>on</strong> Legal Affairs,<br />
Member of the Special<br />
Committee <strong>on</strong> <strong>Artificial</strong><br />
<strong>Intelligence</strong> in a Digital<br />
Age<br />
around who should exercise that role without<br />
the report ending up unraveling it. In my opini<strong>on</strong>,<br />
the existence of a European agency in charge is<br />
necessary, with sufficient rule to avoid the need<br />
to regulate every small change that this c<strong>on</strong>stantly<br />
growing technological envir<strong>on</strong>ment is generating.<br />
The report approved by the European<br />
Parliament puts emphasis <strong>on</strong> the c<strong>on</strong>trol of the<br />
so-called “high-risk applicati<strong>on</strong>s”, for which it<br />
provides c<strong>on</strong>trols prior to their use and throughout<br />
their operati<strong>on</strong>, as it establishes demanding<br />
obligati<strong>on</strong>s related to transparency, auditability<br />
and reversibility. To define this high risk, the text<br />
proposes a combinati<strong>on</strong> of categories referring to<br />
the degree of predictability of putting a community<br />
or an individual at risk, the use and the sector<br />
where it will be used. Through this mixed system<br />
we intend to give legal certainty to the system and,<br />
at the same time, generate what has to be known<br />
as “trustworthy ecosystem”.<br />
Our objective is also that the design, the<br />
development, the c<strong>on</strong>trol and the supervisi<strong>on</strong><br />
of this regulatory framework be participated by<br />
all citizens, and especially by the most involved<br />
or affected individuals and groups. The text<br />
establishes a mandate for all European and nati<strong>on</strong>al<br />
supervisory bodies so that, in an indispensable<br />
way and periodically, they have the assistance<br />
and participati<strong>on</strong> of civil society in the design of<br />
the main characteristics of the model, with special<br />
attenti<strong>on</strong> to integrate the perspective of the small<br />
and medium companies, the uni<strong>on</strong>s, or the<br />
c<strong>on</strong>sumers. We intend not <strong>on</strong>ly a co-resp<strong>on</strong>sibility<br />
<strong>on</strong> the compliance and enforcement, but also their<br />
inclusi<strong>on</strong> somehow or other in the design of the<br />
governance mechanisms for the entire system.<br />
Finally, I would like to highlight the need to<br />
incorporate the general public into the debate and<br />
knowledge of the implicati<strong>on</strong>s of AI. This debate<br />
cannot be <strong>on</strong>ly reduced to a group of experts,<br />
operators, legislators and even associati<strong>on</strong>s and<br />
groups of users. Such a disruptive technology has<br />
to be subject to democratic c<strong>on</strong>trol; societies as a<br />
whole have to be aware of this reality and be able to<br />
make big decisi<strong>on</strong>s about the models of society they<br />
want. For this reas<strong>on</strong>, I prefer to talk about “digital<br />
literacy” rather than “acquisiti<strong>on</strong> of digital skills”<br />
-and I believe that it is the duty of public authorities<br />
to make the main c<strong>on</strong>cepts understandable to their<br />
citizens and to guarantee a mature public debate<br />
around AI. In this, the very survival of democracy<br />
is at stake, taking into account the disturbing social<br />
effects that an unc<strong>on</strong>trolled applicati<strong>on</strong> of AI can<br />
have in some areas.<br />
29
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
When the President of the<br />
European Commissi<strong>on</strong> made<br />
her first speech to the European<br />
Parliament in December 2019, she officially<br />
recognized “<strong>Artificial</strong> <strong>Intelligence</strong>” as an area<br />
of strategic importance for the European<br />
Uni<strong>on</strong>. Nine m<strong>on</strong>ths later, addressing <strong>on</strong>ce<br />
again the European Parliament in her maiden<br />
“State of the Uni<strong>on</strong>’ speech”, she had moved<br />
from spelling out “<strong>Artificial</strong> <strong>Intelligence</strong>” to<br />
talking in terms of ‘AI’ – so well-known is the<br />
technology within the EU bubble now. This is<br />
not so surprising when AI is being deployed<br />
across most (if not all) sectors of the ec<strong>on</strong>omy,<br />
from disease diagnosis to minimizing the<br />
envir<strong>on</strong>mental impact of farming.<br />
It is true that much work has been d<strong>on</strong>e<br />
by the European Commissi<strong>on</strong> since President<br />
Ursula V<strong>on</strong> der Leyen and her team took<br />
office. Already promised in December 2019<br />
was a “legislative proposal” <strong>on</strong> AI – what was<br />
delivered was an AI White Paper in February.<br />
While this, admittedly, is not a legislative<br />
proposal, it is a document that has kick-started<br />
the debate <strong>on</strong> human and ethical AI, the use of<br />
Big Data, and how these technologies can be<br />
used to create wealth for society and business.<br />
The Commissi<strong>on</strong>’s White Paper<br />
emphasizes the importance of establishing<br />
a uniform approach to AI across the EU’s 27<br />
member states, where different countries<br />
have started to take their own approach to<br />
regulati<strong>on</strong>, and thus potentially, are erecting<br />
barriers to the EU’s single market. It also,<br />
importantly for Huawei, talks about plans to<br />
take a risk-based approach to regulating AI.<br />
At Huawei we studied the White Paper<br />
with interest, and al<strong>on</strong>g with (more than<br />
1,250!) other stakeholders, c<strong>on</strong>tributed to<br />
the Commissi<strong>on</strong>’s public c<strong>on</strong>sultati<strong>on</strong>, which<br />
closed <strong>on</strong> 14 June, giving our input and ideas<br />
as experts working in this field.<br />
Finding the Balance<br />
The main point that we emphasized to the<br />
Commissi<strong>on</strong> is the need to find the right<br />
balance between allowing innovati<strong>on</strong> and<br />
ensuring adequate protecti<strong>on</strong> for citizens.<br />
In particular, we focused <strong>on</strong> the need for<br />
high-risk applicati<strong>on</strong>s to be regulated under<br />
a clear legal framework, and proposed ideas<br />
for what the definiti<strong>on</strong> of AI should be. In<br />
this regard, we believe the definiti<strong>on</strong> of AI<br />
should come down to its applicati<strong>on</strong>, with<br />
risk assessments focusing <strong>on</strong> the intended<br />
use of the applicati<strong>on</strong> and the type of impact<br />
Angeliki<br />
Dedopoulou<br />
Senior Manager for<br />
EU Public Affairs with<br />
Huawei Technologies<br />
30 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
SHUTTERSTOCK<br />
AI in<br />
the EU:<br />
Balancing<br />
Benefit &<br />
C<strong>on</strong>trol<br />
By Angeliki Dedopoulou<br />
resulting from the AI functi<strong>on</strong>. If there<br />
are detailed assessment lists and<br />
procedures in place for companies<br />
to make their own self-assessments,<br />
then this will reduce the cost of initial<br />
risk assessment – which must match<br />
sector-specific requirements.<br />
We have recommended that<br />
the Commissi<strong>on</strong> looks into bringing<br />
together c<strong>on</strong>sumer organizati<strong>on</strong>s,<br />
academia, member states, and<br />
businesses to assess whether an AI<br />
system may qualify as high-risk. There<br />
is already an established body set up<br />
to deal with these kinds of things – the<br />
standing Technical Committee High<br />
Risk Systems (TCRAI). We believe this<br />
body could assess and evaluate AI<br />
systems against high-risk criteria both<br />
legally and technically. If this body<br />
took some c<strong>on</strong>trol, combined with a<br />
voluntary labelling system, <strong>on</strong> offer<br />
would be a governance model that<br />
• c<strong>on</strong>siders the entire supply chain<br />
• sets the right criteria and targets<br />
the intended goal of transparency for<br />
c<strong>on</strong>sumers/businesses<br />
• incentivizes the resp<strong>on</strong>sible<br />
development and deployment of AI,<br />
and<br />
• Creates an ecosystem of trust.<br />
Outside of the high-risk applicati<strong>on</strong>s<br />
of AI, we have stated to the Commissi<strong>on</strong><br />
that the existing legal framework based<br />
<strong>on</strong> fault-based and c<strong>on</strong>tractual liability<br />
is sufficient – even for state-of-theart<br />
technologies like AI, where there<br />
could be a fear that new technology<br />
requires new rules. Extra regulati<strong>on</strong><br />
is however, unnecessary; it would be<br />
over-burdensome and discourage the<br />
adopti<strong>on</strong> of AI.<br />
From what we know of the current<br />
thinking within the Commissi<strong>on</strong>, it<br />
appears that it also plans to take a<br />
risk-based approach to regulating AI.<br />
Specifically, the Commissi<strong>on</strong> proposes<br />
focusing in the short-term <strong>on</strong> “highrisk”<br />
AI applicati<strong>on</strong>s – meaning either<br />
high-risk sectors (like healthcare) or in<br />
high-risk use (for example whether it<br />
produces legal or similarly significant<br />
effects <strong>on</strong> the rights of an individual).<br />
So, What Happens Next?<br />
The Commissi<strong>on</strong> has a lot of work to do<br />
in getting through all the c<strong>on</strong>sultati<strong>on</strong><br />
resp<strong>on</strong>ses, taking into account the<br />
needs of business, civil society, trade<br />
associati<strong>on</strong>s, NGOs and others. The<br />
additi<strong>on</strong>al burden of working through<br />
the cor<strong>on</strong>avirus crisis has not helped<br />
matters, with the formal resp<strong>on</strong>se from<br />
the Commissi<strong>on</strong> now not expected until<br />
Q1 2021.<br />
Cor<strong>on</strong>avirus has been a game-changer<br />
for technology use in healthcare of<br />
course, and will no doubt have an<br />
impact <strong>on</strong> the Commissi<strong>on</strong>’s thinking in<br />
this area. Terms such as “telemedicine”<br />
have been talked about for years, but the<br />
crisis has turned virtual c<strong>on</strong>sultati<strong>on</strong>s<br />
into reality – almost overnight.<br />
Bey<strong>on</strong>d healthcare we see AI<br />
deployment being c<strong>on</strong>tinuously rolled<br />
out in areas such as farming and in the<br />
EU’s efforts to combat climate change.<br />
We are proud at Huawei to be part of<br />
this c<strong>on</strong>tinuous digital development in<br />
Europe – a regi<strong>on</strong> in which and for which<br />
we have been working for 20 years.<br />
The development of digital skills is at<br />
the heart of this, which not <strong>on</strong>ly equips<br />
future generati<strong>on</strong>s with the tools to seize<br />
the potential of AI, but will also enable<br />
the current workforce to be active<br />
and agile in an ever-changing world:<br />
there is a need for an inclusive, lifel<strong>on</strong>g<br />
learning-based and innovati<strong>on</strong>-driven<br />
approach to AI educati<strong>on</strong> and training,<br />
to help people transiti<strong>on</strong> between jobs<br />
seamlessly. The job market has been<br />
heavily impacted by the crisis, and quick<br />
soluti<strong>on</strong>s are needed.<br />
As we wait for the Commissi<strong>on</strong>’s<br />
formal resp<strong>on</strong>se to the White Paper,<br />
what more is there to say about AI<br />
in Europe? Better healthcare, safer<br />
and cleaner transport, more efficient<br />
manufacturing, smart farming and<br />
cheaper and more sustainable energy<br />
sources: these are just a few of the<br />
benefits AI can bring to our societies,<br />
and to the EU as a whole. Huawei will<br />
work with EU policymakers and will<br />
strive to ensure the regi<strong>on</strong> gets the<br />
balance right: innovati<strong>on</strong> combined with<br />
c<strong>on</strong>sumer protecti<strong>on</strong>.<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
31
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
AI and misinformati<strong>on</strong><br />
Str<strong>on</strong>g rules for digital<br />
opini<strong>on</strong>-forming marketplaces<br />
By Alexandra Geese<br />
Alexandra<br />
Geese<br />
Member of the<br />
European Parliament<br />
from Germany<br />
(Greens-European Free<br />
Alliance), Member of<br />
the Special Committee<br />
<strong>on</strong> <strong>Artificial</strong> <strong>Intelligence</strong><br />
in a Digital Age<br />
The Covid crisis in<br />
particular has led to<br />
a false informati<strong>on</strong><br />
epidemic, with<br />
increased yearning<br />
for simple answer<br />
and quick fixes.<br />
Simultaneously, the<br />
corrective effect of real<br />
life c<strong>on</strong>tact is lacking.<br />
Who is protecting us?<br />
There is no privacy <strong>on</strong> the<br />
internet and nothing is as free<br />
as it looks. And yet we c<strong>on</strong>fide<br />
everything to it, even giving it authority<br />
over democratic discourse. That is<br />
dangerous. Because communicati<strong>on</strong><br />
<strong>on</strong> <strong>on</strong>line platforms is c<strong>on</strong>trolled<br />
by algorithms programmed with<br />
ruthless business interests, seeking<br />
to keep people <strong>on</strong> sites as l<strong>on</strong>g as<br />
possible to show them adverts. That<br />
works best with extreme c<strong>on</strong>tent, for<br />
our str<strong>on</strong>gest emoti<strong>on</strong>s are fear and<br />
anger.<br />
Hate speech, misinformati<strong>on</strong>,<br />
polarisati<strong>on</strong> and prejudices fall <strong>on</strong><br />
fertile ground, currently exacerbated<br />
by the Covid crisis. We are all<br />
spending much more time <strong>on</strong>line,<br />
thereby automatically coming across<br />
more false informati<strong>on</strong>, including<br />
disinformati<strong>on</strong> deliberately spread<br />
by domestic or foreign parties and a<br />
whole host of other factually incorrect,<br />
dec<strong>on</strong>textualised or obsolete c<strong>on</strong>tent<br />
which mislead people, reinforce<br />
dubious narratives and devalue true<br />
expert knowledge.<br />
That can have fatal c<strong>on</strong>sequences,<br />
such as D<strong>on</strong>ald Trump alleging<br />
in spring 2020 that it might be<br />
“interesting” to inject people with<br />
disinfectant to fight cor<strong>on</strong>avirus. A<br />
US government expert had previously<br />
correctly explained that bleach and<br />
disinfectant could quickly kill the<br />
cor<strong>on</strong>avirus <strong>on</strong> metal surfaces. The<br />
absolutely n<strong>on</strong>sensical idea which<br />
Trump then released into the world<br />
cost human lives, seeing significant<br />
increases in calls to the pois<strong>on</strong> centre.<br />
People were gargling with bleach. A<br />
family in Florida even sold bleach<br />
as a “miracle mineral soluti<strong>on</strong>” and<br />
founded a church to sell it.<br />
The Covid crisis in particular has led<br />
to a false informati<strong>on</strong> epidemic, with<br />
increased yearning for simple answer<br />
and quick fixes. Simultaneously, the<br />
corrective effect of real life c<strong>on</strong>tact is<br />
lacking. Who is protecting us?<br />
Not the platform operators,<br />
because misinformati<strong>on</strong> helps their<br />
business interests. Dubious but<br />
highly emoti<strong>on</strong>al c<strong>on</strong>tent boosts their<br />
profits. Fact-checking is a good way of<br />
preventing the worst excesses. But it<br />
is no panacea. Twitter and Facebook<br />
dem<strong>on</strong>strated how misleading c<strong>on</strong>tent<br />
can be handled differently during<br />
the US presidential electi<strong>on</strong>. Whilst<br />
Twitter placed fact-checking labels<br />
<strong>on</strong> numerous tweets by President<br />
Trump, Facebook and its algorithms<br />
allowed the “Stop the Steal” group to<br />
amass 200,000 members overnight,<br />
despite it disseminating false electi<strong>on</strong><br />
informati<strong>on</strong> and inciting violence, thus<br />
breaching the network’s terms and<br />
c<strong>on</strong>diti<strong>on</strong>s. However, democratic<br />
governments must not make arbitrary<br />
decisi<strong>on</strong>s about <strong>on</strong>line c<strong>on</strong>tent, either.<br />
Freedom of expressi<strong>on</strong> is precious,<br />
and no government should be able<br />
to intervene in c<strong>on</strong>tent moderati<strong>on</strong><br />
<strong>on</strong> a political whim.<br />
Rather, policy-makers ought to<br />
draft and implement clear rules.<br />
First of all, uniform rules for dealing<br />
with illegal c<strong>on</strong>tent. But the majority<br />
32 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
of misinformati<strong>on</strong> is not illegal. Should<br />
we prohibit people from sharing private<br />
c<strong>on</strong>tent, the validity of which they cannot<br />
judge? Of course not.<br />
What we can demand and desperately<br />
need, however, is transparency, first of<br />
all for recommendati<strong>on</strong> systems – the<br />
artificial intelligence employed to govern<br />
c<strong>on</strong>tent. A Wall Street Journal report<br />
found extremist c<strong>on</strong>tent in a third of<br />
all Facebook groups. The groups were<br />
recommended to 66% of members by a<br />
Facebook algorithm. We have a societal<br />
right to know about and publicly debate<br />
these decisi<strong>on</strong> mechanisms. Meaningful<br />
transparency will be increasingly<br />
important in the future - just think about<br />
deep fakes – but it is already urgent.<br />
AlgorithmWatch, a Berlin-based NGO,<br />
has made c<strong>on</strong>crete proposals regarding<br />
researchers’ and investigative journalists’<br />
access to raw data.<br />
The mechanisms which currently<br />
c<strong>on</strong>trol our communicati<strong>on</strong> and<br />
significant porti<strong>on</strong>s of the basic<br />
informati<strong>on</strong> in public debate are a<br />
veritable black box. Independent<br />
researchers <strong>on</strong>ly have restricted access<br />
to data, with access to public applicati<strong>on</strong><br />
programming interfaces (APIs) being<br />
increasingly curtailed in recent years,<br />
leaving Facebook, YouTube and the like<br />
to use every psychological trick to exploit<br />
their knowledge m<strong>on</strong>opoly. As a society,<br />
we must put an end to this. We ought to<br />
be able to co-determine rules for digital<br />
opini<strong>on</strong>-forming marketplaces, based <strong>on</strong><br />
hard facts currently withheld from us by<br />
c<strong>on</strong>glomerates.<br />
I propose “Social Media Councils”<br />
as a model for public debate, similar<br />
to Citizens’ Assemblies in Ireland,<br />
comprising civil society, experts for<br />
freedom of expressi<strong>on</strong>, democracy<br />
and technology, and representatives of<br />
groups particularly affected by hatred<br />
and hate speech. They can trigger public<br />
debates based <strong>on</strong> evidence gained<br />
through transparency obligati<strong>on</strong>s,<br />
identify good and bad practice, and<br />
issue recommendati<strong>on</strong>s for acti<strong>on</strong> to<br />
politicians. Facebook and its internal<br />
ethics committee seek to privatise<br />
precisely this public debate. Facebook’s<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
own internal appraisal of its practices<br />
is commendable, but in the l<strong>on</strong>g run a<br />
committee answerable to the CEO will<br />
never decide in society’s favour over its<br />
own boss’s business interests. That is<br />
why we need space for public reflecti<strong>on</strong>.<br />
Prominent Facebook critics such as<br />
Carole Cadwalladr, who exposed the<br />
Cambridge Analytica scandal, and Roger<br />
McNamee, early investor in Facebook<br />
and venture capitalist, have joined forces<br />
in the “Real Facebook Oversight Board” to<br />
publicise particularly dubious practices.<br />
Micro-targeting and targeted<br />
advertising, i.e. the disseminati<strong>on</strong> of<br />
advertising to very small target groups<br />
based <strong>on</strong> previously collected and<br />
collated data, must also urgently be<br />
banned. The resulting gigantic database<br />
with milli<strong>on</strong>s of highly detailed user<br />
profiles also enables misleading<br />
messages to be spread to especially<br />
ISTOCK<br />
susceptible users and is thus highly<br />
problematic. Moreover, Google and<br />
Facebook c<strong>on</strong>trol large secti<strong>on</strong>s of the<br />
ad-tech market, thereby c<strong>on</strong>stantly<br />
expanding their market shares to<br />
the detriment of European media<br />
outlets. Nowadays, press publishers’<br />
proceeds are also highly dependent<br />
<strong>on</strong> behaviour-based advertisement.<br />
However, the Dutch public broadcaster<br />
NPO has dem<strong>on</strong>strated that equally<br />
successful methods for c<strong>on</strong>text-based<br />
advertisement can exist without pers<strong>on</strong>al<br />
data or spying <strong>on</strong> and pursuing people<br />
via various websites and into their offline<br />
lives. A viable press financing model<br />
is needed to reinforce substantiated<br />
reporting as a foundati<strong>on</strong> of liberty and<br />
democracy, whilst not inadvertently<br />
weakening that very democracy<br />
through the unwanted disseminati<strong>on</strong> of<br />
misinformati<strong>on</strong>.<br />
33
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
ISTOCK<br />
Teach AI in schools.<br />
Win the tech race in 2030.<br />
By K<strong>on</strong>stantinos Kyranakis<br />
Ten years ago every<strong>on</strong>e was talking about the future<br />
referring to 2020 as a really special year. Every nati<strong>on</strong>,<br />
every business, every sports team was making big<br />
plans. Well, COVID ruined most of them, except <strong>on</strong>e. The<br />
COVID crisis accelerated technological advancements almost<br />
everywhere in the world. Zoom calls, Amaz<strong>on</strong> orders, <strong>on</strong>line<br />
project planners, gaming, Netflix and so many quarantine<br />
apps skyrocketed this year.<br />
However the tech king that has really expanded<br />
horiz<strong>on</strong>tally this year is <strong>Artificial</strong> <strong>Intelligence</strong> (AI). Billi<strong>on</strong>s<br />
of people have experimented with face filters <strong>on</strong> a mobile<br />
app. Trilli<strong>on</strong>s of <strong>on</strong>line purchases have been made because<br />
of pers<strong>on</strong>alized suggesti<strong>on</strong>s. And surely while 2020 has not<br />
been the best year for travelling, all those who went through<br />
internati<strong>on</strong>al airports for some reas<strong>on</strong> might have noticed<br />
automatic passport verificati<strong>on</strong>.<br />
34 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
K<strong>on</strong>stantinos<br />
Kyranakis<br />
Member of the Hellenic<br />
Parliament, former<br />
President of the Youth<br />
of the European<br />
People’s Party<br />
Chances are that you fall at least in<br />
<strong>on</strong>e of those categories and if so you have<br />
been exposed to the power of AI which<br />
undoubtedly became a big part of our lives<br />
in 2020 without making too much noise.<br />
From face recogniti<strong>on</strong> to self-driving cars,<br />
AI is growing exp<strong>on</strong>entially and we need<br />
to get ready for what is coming.<br />
Kids entering first grade at elementary<br />
schools this year, will be joining the<br />
workforce in the decade of 2030. If<br />
today we c<strong>on</strong>sider AI to be part of the<br />
successful operati<strong>on</strong>s of a company or<br />
a government, we can <strong>on</strong>ly imagine that<br />
ten years from today it will be absolutely<br />
necessary. It will be omnipresent and<br />
its impact will be taken for granted. All<br />
industries, including transport, logistics,<br />
medicine, c<strong>on</strong>structi<strong>on</strong>, defense,<br />
architecture, customer service and even<br />
tax authorities will be utilizing AI. We have<br />
a duty to prepare the next generati<strong>on</strong> for<br />
what is to come.<br />
That is the reas<strong>on</strong> I submitted a<br />
proposal to the Greek Parliament for AI<br />
to be taught in schools. Obviously, no <strong>on</strong>e<br />
is expecting for 1st graders to become<br />
Pyth<strong>on</strong> developers, but if they learn from<br />
a young age how to think algorithmically,<br />
understand the purposes of AI and realise<br />
what this technology can achieve, then<br />
they will be able to go <strong>on</strong> and accomplish<br />
goals that are hard to even c<strong>on</strong>ceive<br />
today.<br />
C<strong>on</strong>cretely, basic methods and<br />
patterns can be taught in the early<br />
stages of elementary school, followed<br />
by establishing a good base in intuitive<br />
programming languages towards the<br />
end of elementary. Stepping <strong>on</strong> these<br />
building blocks, Machine Learning can<br />
be progressively taught as a less<strong>on</strong> that<br />
combines mathematics and computer<br />
science, two subjects that already exist<br />
in schools. Gamifying this subject with<br />
competiti<strong>on</strong>s such as Robotics will<br />
also prove beneficial. Note that these<br />
steps create a hazy path that could be<br />
followed and mainly aim to initiate the<br />
c<strong>on</strong>versati<strong>on</strong> of AI training in schools. I<br />
expect that more bulletproof plans will be<br />
created by expert committees when that<br />
time comes.<br />
A great experiment is going <strong>on</strong> in<br />
Scandinavia right now. Finland has<br />
foreseen the potential impact of AI and<br />
created a course for efficient and effective<br />
AI training for all, with an ultimate<br />
objective of at least 1% of the populati<strong>on</strong><br />
getting trained. Similarly, Sweden spends<br />
over half a billi<strong>on</strong> € <strong>on</strong> AI research<br />
programs such as W.A.S.P and heavily<br />
invests in its universities. Already in 2018<br />
French President Macr<strong>on</strong> announced the<br />
nati<strong>on</strong>’s l<strong>on</strong>g term AI strategy and while<br />
also menti<strong>on</strong>ing that the country will be<br />
investing north of 1.5billi<strong>on</strong> € to boost<br />
France’s AI capabilities. These are just<br />
some of the examples that illustrate that<br />
Europe is taking this issue seriously.<br />
All that said, no investment will matter<br />
if students across Europe today will not<br />
become part of the AI revoluti<strong>on</strong>. For kids<br />
with an inclinati<strong>on</strong> towards STEM subjects,<br />
highlighting to them the power of AI from<br />
an early age will <strong>on</strong>ly benefit them as they<br />
will have a l<strong>on</strong>ger time to cultivate a good<br />
understanding in the field. For all other<br />
students, that will eventually focus <strong>on</strong><br />
other professi<strong>on</strong>s they will also likely end<br />
up utilizing the power of AI, so learning<br />
about it from a young age will ultimately<br />
give them skills and knowledge so as<br />
to become competitive and rise to the<br />
challenges of the 2030 decade.<br />
Europe can become a leader in AI,<br />
10 years from now - so l<strong>on</strong>g that it starts<br />
creating the future leaders of the subject<br />
today, in schools. It follows that investing<br />
in this proposal will lead to further<br />
advancement of universities, state funded<br />
research programs and easier talent<br />
discovery by european companies. If we<br />
want more “unicorns” in Europe, we need<br />
bright young people that will have the<br />
visi<strong>on</strong> and the power to make european<br />
companies, global tech winners.<br />
Some people say access to capital is<br />
the key to win the tech race. I think the<br />
most valuable asset is human capital.<br />
Technology is built by humans. And if<br />
humans are intelligent enough to build<br />
artificially intelligent machines that could<br />
outsmart them, we owe it to our future to<br />
educate our kids and make them smarter<br />
than we are.<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
35
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
AI and Machine Learning in<br />
Digital Campaigning – will<br />
bots bring us closer than ever<br />
or will they ruin our system?<br />
By Hannes Kirchbaumer<br />
Hannes<br />
Kirchbaumer<br />
Founder and CEO of<br />
KIRCHBAUMER.COM,<br />
a digital marketing and<br />
political campaigning<br />
c<strong>on</strong>sulting agency<br />
based in Vienna<br />
A<br />
rtificial intelligence (AI) impacts numerous aspects of life, from smart<br />
devices helping out in homes, self-driving cars, to the way we c<strong>on</strong>sume<br />
informati<strong>on</strong>. With the ability to analyze data and learn from it, AI is<br />
revoluti<strong>on</strong>izing our world. Many industries have a high demand for AI capabilities<br />
including the medical field, banking and of course: marketing. With the help of<br />
very effective data analysis, combined with an ability to adapt to input, AI can<br />
identify marketing trends. Organizati<strong>on</strong>s, brands and political parties can now<br />
leverage AI to save time and resources through automated digital marketing<br />
services.<br />
Using statistical techniques, machine learning algorithms can easily identify<br />
patterns in data. By analyzing <strong>on</strong>line behavior of users which includes their<br />
news and retail c<strong>on</strong>sumpti<strong>on</strong> patterns, relati<strong>on</strong>ships and social media patterns,<br />
it is possible to create unique psychographic and behavioral user profiles. This<br />
makes targeted advertising campaigns possible that can be sent to each user<br />
based <strong>on</strong> their individual psychology. The more interacti<strong>on</strong> data is available, the<br />
better. It is all about sending the right informati<strong>on</strong> to the right pers<strong>on</strong> at the<br />
right time. This goal has been here for a l<strong>on</strong>g time and with the right data to be<br />
processed, it became reality.<br />
FIVE MAIN BENEFITS OF AI IN DIGITAL MARKETING<br />
1) MORE EFFECTIVENESS<br />
AI helps businesses remove the guesswork to create a more effective<br />
marketing strategy with data-driven analysis. Live dashboards and journey<br />
analysis indicate what works best and how users react to certain events.<br />
2) BETTER USER EXPERIENCE<br />
AI makes it possible to provide any audience with the right c<strong>on</strong>tent. This<br />
applies not <strong>on</strong>ly to websites, but also to newsletters and social media.<br />
3) HIGHER PRODUCTIVITY<br />
Automati<strong>on</strong> of repetitive tasks becomes possible, saving time and m<strong>on</strong>ey.<br />
Simultaneously, new resources are freeing up. Employees, who used to be<br />
36 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
occupied with those tasks, can now focus<br />
<strong>on</strong> even more important <strong>on</strong>es.<br />
4) UNDERSTANDING THE AUDIENCE<br />
With AI, data can be analyzed to<br />
predict behaviors and decisi<strong>on</strong> of any<br />
target user. The outcome can be used for<br />
more tests that will deliver new insights<br />
and live results.<br />
5) BETTER DECISIONS<br />
AI empowers decisi<strong>on</strong>-making and<br />
helps to create better-performing<br />
c<strong>on</strong>tent. It is now also possible to let AI do<br />
the creati<strong>on</strong> work as l<strong>on</strong>g as some basic<br />
informati<strong>on</strong> and graphics are provided.<br />
Often, the website c<strong>on</strong>tent might already<br />
be enough for AI to make some well and<br />
often high-performing ad creatives.<br />
AI IN POLITICAL CAMPAIGNS<br />
When working in the political<br />
campaigning field, the first topic in a<br />
discussi<strong>on</strong> about AI will often be the<br />
possible misuse of it. By knowing the<br />
profiles of the users/voters, AI can of<br />
course also be used to send pers<strong>on</strong>alized,<br />
political messages that are insincere and<br />
fake. C<strong>on</strong>tent that highlights a different<br />
side of an argument can be sent to<br />
potential voters and every voter gets a<br />
different versi<strong>on</strong> that is in line with their<br />
expectati<strong>on</strong>s. That helps the candidate<br />
in developing a general opini<strong>on</strong> that is<br />
in his or her favor.<br />
In the 2016 US presidential electi<strong>on</strong>s,<br />
it was found that automated social media<br />
bots were used to increase the Twitter<br />
traffic for pro-Trump hashtags.<br />
They had roughly double the activity<br />
of his rival, Hillary Clint<strong>on</strong>. On Facebook,<br />
Cambridge Analytica provided audience<br />
data that made super micro targeted<br />
campaigns possible and it was found<br />
that the way they had gathered data was<br />
against the policy.<br />
That did not benefit Facebook’s<br />
general image – naturally, they had to do<br />
something about it. Suddenly, targeting<br />
for political campaigns and other social<br />
issues was reduced significantly by the<br />
big players.<br />
But when checking whether<br />
audiences and targeting settings are<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
acceptable and in line with policy, it is<br />
again AI that is coming in. In the 2020 US<br />
presidential electi<strong>on</strong>s, it was also used to<br />
mark posts with a fact-checking warning.<br />
Another misuse of AI and machine<br />
learning is the development of so called<br />
“deepfakes”. It c<strong>on</strong>cerns audio or video<br />
c<strong>on</strong>tent generated by AI that shows<br />
some<strong>on</strong>e saying or doing something<br />
that was never said or d<strong>on</strong>e in reality.<br />
These kind of videos appear extremely<br />
authentic and can be used as a way of<br />
influencing people. Ir<strong>on</strong>ically, it might<br />
need AI again to detect this kind of<br />
c<strong>on</strong>tent.<br />
THE FUTURE OF AI IN DIGITAL<br />
MARKETING & CAMPAIGNING<br />
<strong>Artificial</strong> intelligence is an incredible<br />
technology that can work w<strong>on</strong>ders for any<br />
organizati<strong>on</strong>. Using AI digital marketing<br />
strategies will help deliver improved<br />
customer experience and market more<br />
effectively. AI digital marketing also<br />
changes the way we build websites and<br />
interact with customers.<br />
SHUTTERSTOCK<br />
But what must public and political<br />
organizati<strong>on</strong>s do to be able to use both<br />
the audience and creati<strong>on</strong> part in a good<br />
way?<br />
Everything comes down to collecting<br />
your own data, own AI and machine<br />
learning soluti<strong>on</strong>s and get in more direct<br />
c<strong>on</strong>tent with your users. It is still possible<br />
to gain first-party data with campaigns<br />
<strong>on</strong> the big networks. Exactly that data<br />
should be used to create a perfect<br />
database to work with.<br />
The demand of GDPR-compliant<br />
community-building services with an<br />
AI-based, smart marketing automati<strong>on</strong><br />
functi<strong>on</strong>ality are very high at the moment<br />
and I would recommend all kinds of<br />
organizati<strong>on</strong>s to find a way to build a<br />
so-called “client data platform”. With<br />
increasing restricti<strong>on</strong>s <strong>on</strong> third party<br />
data, it is an absolute obligati<strong>on</strong> to have<br />
your first party data in perfect c<strong>on</strong>diti<strong>on</strong>.<br />
It will help organizati<strong>on</strong>s not <strong>on</strong>ly reach<br />
their users more efficiently, but also give<br />
them valuable informati<strong>on</strong> they might<br />
not have expected.<br />
37
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
<strong>Artificial</strong> intelligence’s great impact<br />
<strong>on</strong> low and middle-skilled jobs<br />
By Sybrand Brekelmans & Georgios Petropoulos<br />
Sybrand<br />
Brekelmans<br />
& Georgios<br />
Petropoulos<br />
Sybrand Brekelmans:<br />
Research assistant at<br />
Bruegel<br />
Georgios Petropoulos:<br />
Marie Curie<br />
Skłodowska Research<br />
Fellow at MIT and<br />
Bruegel and postdoctoral<br />
fellow at the<br />
MIT Initiative <strong>on</strong> the<br />
Digital Ec<strong>on</strong>omy<br />
T<br />
he academic literature suggests<br />
that, in the past decades,<br />
technological progress has<br />
led to job polarisati<strong>on</strong> in European<br />
Uni<strong>on</strong> countries. While computer<br />
technologies and robots have<br />
replaced, to some extent, routine<br />
middle-skilled jobs such as machine<br />
operati<strong>on</strong>, c<strong>on</strong>structi<strong>on</strong> work or<br />
administrative work, they have also<br />
led to an increase in complementary,<br />
n<strong>on</strong>-routine high-skilled jobs (eg<br />
managers, professi<strong>on</strong>als) and in<br />
low-skilled jobs (eg agriculture,<br />
cleaning and pers<strong>on</strong>al care services).<br />
However, our new research suggests<br />
that the new technologies that have<br />
emerged since 2010 – artificial<br />
intelligence and machine learning –<br />
are set to change drastically the job<br />
landscape over the next few decades.<br />
These technologies are likely to have<br />
a deeper impact across a wider range<br />
of jobs and tasks, including possible<br />
destructi<strong>on</strong> of low-skilled jobs.<br />
<strong>Artificial</strong> intelligence (AI) systems<br />
are able to perform tasks that involve<br />
decisi<strong>on</strong>-making, therefore changing<br />
the impact of automati<strong>on</strong> <strong>on</strong> the<br />
workforce.<br />
AI-powered technologies can<br />
now retrieve informati<strong>on</strong>, coordinate<br />
logistics, handle inventories, prepare<br />
taxes, provide financial services,<br />
translate complex documents, write<br />
business reports, prepare legal briefs<br />
and diagnose diseases. Moreover,<br />
they are set to become much better<br />
at these tasks in the next few years<br />
thanks to machine learning (ML):<br />
computers fed by big data can learn,<br />
practice skills and ultimately improve<br />
their own performances and perform<br />
their assigned tasks more efficiently.<br />
Our new working paper evaluates<br />
the ‘probability of automati<strong>on</strong>’ for<br />
different jobs, using data from 24<br />
European countries.<br />
This probability is initially<br />
computed at the job task level and<br />
then aggregated at the occupati<strong>on</strong>al<br />
level (Table 1). Since each job<br />
c<strong>on</strong>sists of a variety of tasks, with<br />
different potential for automati<strong>on</strong>,<br />
the probability of automati<strong>on</strong> at the<br />
job level does not necessarily mean<br />
the destructi<strong>on</strong> of jobs, but rather<br />
whether automati<strong>on</strong> can significantly<br />
transform the nature of those jobs.<br />
We use this measure of<br />
automati<strong>on</strong> in an aggregate<br />
framework where jobs are grouped<br />
into three different categories of skill:<br />
low, middle and high-skilled jobs.<br />
Figure 1 shows the results.<br />
These results suggest that<br />
artificial intelligence and machine<br />
learning will have different impacts<br />
compared to computer and robotic<br />
technologies, which caused job<br />
polarisati<strong>on</strong> (drop in routine middleskilled<br />
jobs and increase in lowskilled<br />
jobs). In c<strong>on</strong>trast, AI is highly<br />
likely to significantly alter not <strong>on</strong>ly<br />
38 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Table 1: European jobs with the highest and lowest probabilities of automati<strong>on</strong><br />
Figure 1: Exposure to automati<strong>on</strong> of different skill groups<br />
middle-skilled jobs, but also low-skill employment. Moreover,<br />
while the high skilled are relatively less at risk from AI and<br />
ML-induced transformati<strong>on</strong>, its impact is still n<strong>on</strong>-negligible<br />
for these jobs.<br />
The results also suggest a future transformati<strong>on</strong> of work.<br />
In middle and low-skilled jobs, AI systems will complete the<br />
easily automated tasks while humans c<strong>on</strong>tinue to perform<br />
those that cannot be automated.<br />
A high probability of automati<strong>on</strong> may also be associated<br />
with the creati<strong>on</strong> of new tasks and jobs though the productivity<br />
gains from adopting AI technologies, but these jobs and tasks<br />
will most likely be high-skilled.<br />
The transformative nature of AI and ML requires proactive<br />
measures to re-design labour markets. The workforce needs<br />
to be prepared for the upcoming changes, while the efficiency<br />
gains from these technologies should be harnessed. Countries<br />
with high degrees of labour flexibility, high quality science<br />
educati<strong>on</strong> and less pervasive product market regulati<strong>on</strong>s tend<br />
to have higher skill-oriented job structures and are therefore<br />
less exposed to labour transformati<strong>on</strong> due to automati<strong>on</strong>.<br />
Brekelmans S., G. Petropoulos (2020), ‘<strong>Artificial</strong> intelligence’s great impact <strong>on</strong> low and middle-skilled jobs’, Bruegel Blog, 29 June, available<br />
at https://www.bruegel.org/2020/06/artificial-intelligences-great-impact-<strong>on</strong>-low-and-middle-skilled-jobs/<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
39
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
ISTOCK<br />
Global cooperati<strong>on</strong><br />
is the <strong>on</strong>ly way forward<br />
The ICT industry and its critical role<br />
The crackdowns felt<br />
across the industry<br />
Through my l<strong>on</strong>g career in the Informati<strong>on</strong><br />
and Communicati<strong>on</strong> Technology (ICT)<br />
industry, I have been lucky enough to be at<br />
the fr<strong>on</strong>t seat, watching as the world enters<br />
this ever-changing era of digitalisati<strong>on</strong>.<br />
Over the years, we have developed<br />
a unique ecosystem that enables<br />
interacti<strong>on</strong>s between services, systems,<br />
data, software, and people. In its very<br />
essence, the digital world is the pathway<br />
through which we c<strong>on</strong>nect, share, and<br />
transform together as a global community.<br />
We are in the midst of the Fourth Industrial<br />
Revoluti<strong>on</strong>, seeing our society transform,<br />
from innovative business models to the<br />
Internet of Things (IoT), e-Learning to the<br />
rise of remote working. We are all part of<br />
an interc<strong>on</strong>nected digital ecosystem and<br />
it must remain so if we are to succeed in<br />
tackling the world’s biggest challenges.<br />
Digital technologies are increasingly<br />
becoming the heart of the way we live<br />
and work. What’s less understood, but no<br />
less important, is that digital technologies<br />
are becoming essential to our future <strong>on</strong><br />
the Planet, its ecosystems, and inclusive<br />
ec<strong>on</strong>omic development. The way that we<br />
40 December 2020 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Luis Neves<br />
CΕΟ of the Global<br />
Enabling Sustainability<br />
Initiative (GeSI)<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | December 2020<br />
use digital technologies will determine our<br />
future.<br />
Recent research by GeSI, c<strong>on</strong>ducted<br />
with the support of Deloitte, finds that the<br />
applicati<strong>on</strong> and use of digital technology will<br />
be essential to achieving the bold objectives<br />
set by the UN Sustainable Development<br />
Goals (SDGs). What’s particularly urgent is<br />
to mobilise the use of digital technologies to<br />
transiti<strong>on</strong> to a low carb<strong>on</strong> ec<strong>on</strong>omy that will<br />
mitigate the worst c<strong>on</strong>sequences of climate<br />
change. Digital technologies, GeSI’s research<br />
finds, have the unique capability to enable<br />
nati<strong>on</strong>s to decouple carb<strong>on</strong> emissi<strong>on</strong>s from<br />
ec<strong>on</strong>omic growth – enabling the world to<br />
raise the standard of living while slowing<br />
temperature increases.<br />
Yet just when the world’s governments<br />
need to be collaborating to promote global<br />
technology strategies for sustainability, we<br />
see the increase of techno-nati<strong>on</strong>alism.<br />
Around the world, countries are instituting<br />
destructive, zero-sum policies that are<br />
hampering access to and the applicati<strong>on</strong><br />
of technology that would help the world<br />
advance to achieve the SDGs.<br />
• The US and Japan are restricting<br />
Chinese made comp<strong>on</strong>ents that<br />
affect the progress of building 5G<br />
networks<br />
• China has retaliated by ordering its<br />
public instituti<strong>on</strong>s and government<br />
agencies to stop using foreign made<br />
computers and software<br />
• In Brussels, the EU has called for<br />
the creati<strong>on</strong> of a U.S.-EU Trans-<br />
Atlantic ec<strong>on</strong>omic model that block<br />
China’s attempts to influence global<br />
standards in 5G and other nextgenerati<strong>on</strong><br />
technologies.<br />
• Middle East countries face<br />
technological sancti<strong>on</strong>s<br />
• India has explored ways to<br />
nati<strong>on</strong>alize data.<br />
The list of examples goes <strong>on</strong>. Arguments<br />
abound regarding the importance of<br />
protecting nati<strong>on</strong>al security and ec<strong>on</strong>omic<br />
sovereignty. Yet the arguments overlook<br />
a larger, more urgent questi<strong>on</strong> regarding<br />
collective security.<br />
As it stands the situati<strong>on</strong> will inevitably<br />
lead to the fragmentati<strong>on</strong> of the global<br />
supply chain, decrease competiti<strong>on</strong> and<br />
increase prices for European c<strong>on</strong>sumers and<br />
manufacturers.<br />
We need a global effort to adopt cutting<br />
edge networks, super-fast processors, smart<br />
data analytic capabilities, etc. and apply<br />
them to the most pressing challenges that<br />
the world faces regarding climate change,<br />
biodiversity loss, water scarcity, poverty,<br />
health, and educati<strong>on</strong>, am<strong>on</strong>g others.<br />
The more techno-nati<strong>on</strong>alism and its<br />
bans, restricti<strong>on</strong>s, and crackdowns we see,<br />
the more the world will notice the effects<br />
of these acti<strong>on</strong>s which will be felt not <strong>on</strong>ly<br />
across industry but also society.<br />
Everything today depends <strong>on</strong> digital<br />
technologies. Digital technologies have<br />
been the engine of growth and prosperity.<br />
They keep us c<strong>on</strong>nected. They are at the<br />
centre of our global society as the COVID-19<br />
pandemic has shown. They are meant to be<br />
used for good, for peace, and development.<br />
Finally, they need to be accessible to all<br />
and universal. We need to do everything to<br />
ensure this Purpose is maintained.<br />
The path forward<br />
What are we as an industry to do? What<br />
are we as citizens, policy makers, to do? We<br />
must choose a smarter future. With our<br />
unparalleled capabilities and interc<strong>on</strong>nected<br />
systems, what we must do now is c<strong>on</strong>tinue<br />
down the path we have already set for<br />
ourselves: towards the achievement of the<br />
Sustainable Development Goals (SDGs).<br />
With the European Uni<strong>on</strong> ramping up its<br />
digital strategy, no company should be left<br />
behind so we can ensure competiti<strong>on</strong> and<br />
a fair level playing field aiming at a str<strong>on</strong>ger,<br />
smarter, more sustainable world. Global<br />
cooperati<strong>on</strong> is the <strong>on</strong>ly way forward.<br />
The challenges our society are currently<br />
facing and will c<strong>on</strong>tinue to face requires<br />
smart, innovative minds working together<br />
to find impactful soluti<strong>on</strong>s and fast. The ICT<br />
sector has a unique opportunity to rebuild a<br />
society that has been hampered by disease,<br />
politics, and war. Technology can accelerate<br />
acti<strong>on</strong>, but it requires a robust, shared<br />
ambiti<strong>on</strong>. Critical to a shared ambiti<strong>on</strong> is a<br />
str<strong>on</strong>g leadership to secure it. We must take<br />
the lead <strong>on</strong> global transformati<strong>on</strong>s to create a<br />
more prosperous, inclusive, and sustainable<br />
world with digital at its core. Failure is not an<br />
opti<strong>on</strong>. We can <strong>on</strong>ly succeed if we move<br />
forward together.<br />
41
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
The dynamics of data accumulati<strong>on</strong><br />
By Julia Anders<strong>on</strong><br />
Each new sunrise increases the certainty that the sun will rise<br />
tomorrow. The more we know about the past, the better we<br />
can predict the future. The same holds true in important AI and<br />
related machine learning (ML) applicati<strong>on</strong>s, where great volumes of<br />
data are needed to reach marketable levels of predictive accuracy.<br />
To maintain dynamism in ML markets, competiti<strong>on</strong> authorities<br />
must thus ensure that all participants access the data they need to<br />
compete viably. But how much is enough? Is more always better,<br />
or is there a point where additi<strong>on</strong>al observati<strong>on</strong>s do little to change<br />
predicti<strong>on</strong>s? Do we need to observe the sun rise a billi<strong>on</strong> times to<br />
know what will happen tomorrow?<br />
Training data is a scarce resource<br />
As detailed in a previous blog, machine learning (ML), a subfield<br />
of AI, is a predicti<strong>on</strong> technology that generates new informati<strong>on</strong><br />
(‘predicti<strong>on</strong>s’) based <strong>on</strong> existing informati<strong>on</strong> (‘data’). Notable ML<br />
applicati<strong>on</strong>s include automated driving, image recogniti<strong>on</strong>, language<br />
processing and search. [1]<br />
Like any product, ML models are <strong>on</strong>ly as good as the raw<br />
material. Access to adequate training data is critical for ML applicati<strong>on</strong><br />
vendors—which The Ec<strong>on</strong>omist went as far as to qualify as “the<br />
world’s most valuable resource”.[2]<br />
Where do ML providers and adopters obtain this valuable<br />
resource? Varian (2018) lists eight potential sources, including: as a byproduct<br />
of operati<strong>on</strong>s (e.g. generated from machines and sensors),<br />
web scraping, data vendors (e.g. Nielsen), cloud providers (e.g.<br />
Amaz<strong>on</strong>), public sector data, and offering a service (e.g. ReCAPTCHA).<br />
Yet, despite a variety of sources, training data can be an<br />
important bottleneck for businesses trying to develop or implement<br />
ML—and an <strong>on</strong>-going c<strong>on</strong>cern for ML models that require regular<br />
retraining. According to a 2019 European Commissi<strong>on</strong> report, access<br />
to adequate training data is a key limiting factor for the development<br />
of ML applicati<strong>on</strong>s.[3]<br />
One particular issue is that data must be ‘big’—a term that is<br />
often used to refer to datasets for which individual observati<strong>on</strong>s<br />
carry little informati<strong>on</strong>al c<strong>on</strong>tent, so that value is derived from having<br />
a large number of observati<strong>on</strong>s (Chakraborty, 2017). In many ML<br />
applicati<strong>on</strong>s, data volume greatly affects model performance.<br />
Image recogniti<strong>on</strong> algorithms, for instance, must be trained <strong>on</strong><br />
large volumes of data (Carrière-Swallow et al., 2019). The questi<strong>on</strong><br />
here is: How much (relevant) data is enough to make a valuable ML<br />
model?[4]<br />
Ec<strong>on</strong>omies of scale and the volume of data needed<br />
The relati<strong>on</strong>ship between data volume and returns from ML<br />
is a c<strong>on</strong>tentious topic. Hal Varian and Pat Bajari, chief ec<strong>on</strong>omists<br />
at Google and Amaz<strong>on</strong>, respectively, argue that model accuracy<br />
increases with data sample size, but at a decreasing rate.[5]<br />
Intuitively, in teaching me (or an algorithm) what Labradors looks<br />
like, the first ten Labradors are more informative than the following<br />
ten. In technical terms, data exhibits ‘decreasing returns to scale’.<br />
Figure 1. Decreasing returns to scale of Stanford Dogs data for<br />
ML accuracy Source: Varian (2018)<br />
If data exhibits decreasing returns to scale, then we might<br />
c<strong>on</strong>clude that firms <strong>on</strong>ly need to secure a large-enough lumpsum<br />
of data to compete viably. Ec<strong>on</strong>omists Ajay Agrawal,<br />
Joshua Gans, and Avi Goldfarb disagree.[7] Their argument<br />
runs as follows: even if data exhibits decreasing returns (as<br />
in Figure 1), a slight lead in data quantity may induce a slight<br />
lead in quality that attracts users. More users generate more<br />
data, which drives higher quality. This is the so-called ‘data<br />
feedback loop’ illustrated in Figure 2. Over time, a small initial<br />
data advantage can translate into a significant share of the user<br />
base and of the market. In the l<strong>on</strong>g run, this self-reinforcing<br />
dynamic can lead to market dominance.<br />
In essence, Agrawal et al. distinguish between technical returns<br />
to data (i.e. accuracy) and ec<strong>on</strong>omic returns to data (i.e. market<br />
shares), and argue that even if data exhibits decreasing returns in<br />
a technical sense (as in Figure 1), returns can be increasing in an<br />
ec<strong>on</strong>omic sense.<br />
42 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
Figure 2. The data feedback loop<br />
Source: based <strong>on</strong> OECD (2016)<br />
One of the mechanisms underlying the data feedback loop<br />
relates to the increased granularity that comes with volume. Google<br />
and Microsoft’s differing views in the area of internet search ML<br />
illustrates this point.[8]<br />
Microsoft, in justifying a 10-year deal for Yahoo search in 2009,[9]<br />
has advanced that higher performance is a direct c<strong>on</strong>sequence of<br />
scale.[10] Following this argument, Google search algorithm is better<br />
than Yahoo’s because more people use Google. More people using<br />
Google means that Google can collects more data and provide better<br />
search results, attracting more users, and so <strong>on</strong>. This is the feedback<br />
loop in acti<strong>on</strong>.<br />
Google has expressed scepticism about Microsoft’s scale<br />
argument.[11] Yahoo’s billi<strong>on</strong>s of searches should be sufficient to<br />
compete effectively.[12] Yahoo’s inferior performance would reflect<br />
the lower quality of its algorithm,[13] rather than the amount of data<br />
it holds. To quote Google’s Varian: “it’s not quantity or quality of the<br />
ingredients that make a difference, it’s the recipes”.<br />
Microsoft has countered that its billi<strong>on</strong>s of searches are<br />
insufficient. Many queries are extremely rare and, for these queries,<br />
Google has more observati<strong>on</strong> points, simply because it holds more<br />
data. Google thus produces better results for rare queries.[14]<br />
Empirical work by Schaefler and al. (2018) vindicate Microsoft’s<br />
claims. Schaefler and al. (2018) examine the impact of additi<strong>on</strong>al<br />
user data <strong>on</strong> the quality of internet search results. They disentangle<br />
the effects of more data from the effect of better algorithms and<br />
find that both matter. In their oven, Varian’s analogy falls flat: both<br />
ingredients and recipe are critical to good cooking.<br />
The Google and Microsoft disagreement also exposes the<br />
str<strong>on</strong>g behavioural and structural assumpti<strong>on</strong>s that underlie the<br />
feedback loop argument. In the internet search case, for instance,<br />
the argument requires that (i) c<strong>on</strong>sumers can detect small quality<br />
differences, (ii) quality differences matters to users (or at least a few),<br />
and (iii) switching costs are sufficiently low. The same reas<strong>on</strong>ing<br />
should not be expected to hold, however, in all ML markets.<br />
In speech to text applicati<strong>on</strong>s, for instance, it is not clear that small<br />
quality differences are obvious to the user, or that they would matter<br />
enough to justify the potentially large costs of switching providers<br />
(e.g. learning to use a new applicati<strong>on</strong>). On the other extreme, some<br />
applicati<strong>on</strong> areas are very sensitive to quality differences. These are<br />
applicati<strong>on</strong>s where accuracy is a matter of life and death.[15] Who<br />
would accept a sec<strong>on</strong>d-best medical diagnostic?[16]<br />
No <strong>on</strong>e wants Dr Bing when Dr Google is down the<br />
corridor. For these applicati<strong>on</strong>s, returns to predicti<strong>on</strong><br />
quality may amount to the whole market, i.e. “the<br />
winner takes all”.<br />
A large share of the empirical work around<br />
returns to scale for data focuses <strong>on</strong> <strong>on</strong>e ML<br />
applicati<strong>on</strong>: internet search.[17] More research<br />
is needed to determine where the feedback loop<br />
argument holds, because the answer is likely to vary<br />
across different applicati<strong>on</strong> area.<br />
Ec<strong>on</strong>omies of scope and the volume of<br />
data needed<br />
Glen Weyl and Eric Posner present a different<br />
argument in the debate. In their 2018 book Radical<br />
Markets, the Microsoft ec<strong>on</strong>omist and University of<br />
Chicago law professor argue that the returns to data<br />
are increasing in scale, but with a twist.<br />
Decreasing returns to scale, they claim, <strong>on</strong>ly hold<br />
in pre-ML ‘standard’ statistical c<strong>on</strong>texts. Standard<br />
statistics addresses relatively simple problems (e.g.<br />
estimate a populati<strong>on</strong> average) for which rough<br />
Julia Anders<strong>on</strong><br />
Research Analyst,<br />
Bruegel<br />
estimates are sufficient:<br />
An entrepreneur who wants to open a wealth<br />
management firm in a neighbourhood wants to<br />
know whether the average income is $100,000 or<br />
$200,000 but doesn’t need to know that it is $201,000<br />
rather than $200,000.[18]<br />
ML tackles problems that are much more<br />
complex, according to Posner and Weyl. These<br />
new, harder tasks are more valuable than earlier,<br />
easier <strong>on</strong>es. They also call for more data, and more<br />
complex data. For these problems, the more the<br />
data, the better it can address these harder tasks.<br />
Therefore, and still following Posner and Weyl, the<br />
more the data, the more valuable it is.<br />
To illustrate the authors’ point, c<strong>on</strong>sider training<br />
an ML to recognise dogs in pictures. This model<br />
requires a certain quantity of data to produce<br />
accurate results. But, as illustrated in Figure 1,<br />
the value of additi<strong>on</strong>al data drops past a certain<br />
threshold. However, as the volume of data c<strong>on</strong>tinues<br />
to grow (and complexity al<strong>on</strong>g with it), the algorithm<br />
c<strong>on</strong>tinues to learn. At some point, it can perform<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
43
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
an additi<strong>on</strong>al, more complex task, such as labelling objects in the<br />
photographs. Again, the value of additi<strong>on</strong>al data for the task of<br />
labelling objects flattens out past a certain threshold. Once again,<br />
as the volume of data c<strong>on</strong>tinues to grow, the algorithm c<strong>on</strong>tinues<br />
to learn. At some point, it can perform an additi<strong>on</strong>al, even more<br />
difficult task, such as understanding the nature of the acti<strong>on</strong>s in<br />
the photographs.<br />
Therefore, the value of data grows to the extent that harder<br />
problems need more data. Weyl and Posner’s argument is, implicitly,<br />
that data exhibits ec<strong>on</strong>omies of scope. As Figure 3 illustrates, we<br />
end up with a picture that challenges Varian’s decreasing returns to<br />
scale hypothesis. Weyl and Posner note that the value of data may<br />
not increase forever: we could see a future where ML has “learned<br />
everything”. But until then, they claim, returns to the volume of data<br />
are increasing.<br />
Figure 3. Value of data as a functi<strong>on</strong> of the number of<br />
observati<strong>on</strong>s in a typical ML<br />
Source: Weyl and Posner (2018)<br />
Data as a barrier to entry<br />
Ec<strong>on</strong>omic theory teaches us that, where there are increasing<br />
returns to scale, m<strong>on</strong>opolies naturally emerge—the largest firm<br />
being the most ec<strong>on</strong>omically efficient.[19] Likewise, where they<br />
materialise, increasing returns to scale to data could be expected to<br />
work to c<strong>on</strong>centrate ec<strong>on</strong>omic gains in the hands of data-rich firms.<br />
M<strong>on</strong>opolisati<strong>on</strong>, however, is not necessarily syn<strong>on</strong>ymous with<br />
m<strong>on</strong>opolistic behaviour—i.e. supra-competitive pricing, reduced<br />
quality, and/or hampered innovati<strong>on</strong>. So l<strong>on</strong>g as the market leaders<br />
are challenged by the prospect of competitive entry by other firms,<br />
m<strong>on</strong>opolised markets can be competitive.<br />
Increasing returns to scale and scope for data, however, might<br />
work in some cases to entrench the incumbent’s positi<strong>on</strong> by creating<br />
very high barriers to entry for prospective entrants. This could be a<br />
problem in markets where there is no substitute for a dataset that<br />
is essential, i.e. where there is <strong>on</strong>ly <strong>on</strong>e dataset and it cannot be<br />
replicated, dispensed or purchased.[20] As US FTC Commissi<strong>on</strong>er<br />
Terrell McSweeny has noted, “it may be that an incumbent has<br />
significant advantages over new entrants when a firm has a database<br />
that would be difficult, costly, or time-c<strong>on</strong>suming for a new firm to<br />
match or replicate.”[21]<br />
Incumbency advantage could be particularly pr<strong>on</strong>ounced in the<br />
digital space. Leading ML users were <strong>on</strong>ce leading data collectors,<br />
but increasingly are leading ML providers. Amaz<strong>on</strong>, for instance,<br />
originally sold books as a way to gather pers<strong>on</strong>al data <strong>on</strong> affluent,<br />
educated shoppers (Ezrachi et al. 2016). It is now a leading provider<br />
of ML services <strong>on</strong> the cloud.<br />
‘Data scarcity’ can seem oxymor<strong>on</strong>ic. The world overflows with<br />
informati<strong>on</strong>—right up to toilet seats, the new data hotbed. The<br />
volume of data created and copied each year is expected to reach<br />
44 x 1021 bytes in 2020 — 40 times more bytes than there are stars in<br />
the observable universe.[22] Yet the vast amounts of data collected<br />
<strong>on</strong>line form unique sets of behavioural data that may be harder and<br />
harder to replicate in light of the self-reinforcing dynamics described<br />
above.<br />
Why not let prospective entrants buy the necessary data in<br />
<strong>on</strong>e of the many dedicated marketplaces (such as datapace.io)? A<br />
c<strong>on</strong>cern is that data incumbents have str<strong>on</strong>g reas<strong>on</strong>s not to sell their<br />
data: increasing ec<strong>on</strong>omic returns to data tend to create perverse<br />
incentives for firms to establish a data advantage and erect barriers<br />
to entry thereafter (Cockburn et al., 2018). Incumbents might prefer<br />
hoarding the data they collect in order to, first, gain an advantage<br />
over their competitors and, later, curtail market entry (J<strong>on</strong>es et al.,<br />
2018).<br />
Two factors that could tend to compound these<br />
challenges.<br />
First, as previously noted, data-rich firms could benefit not <strong>on</strong>ly<br />
from ec<strong>on</strong>omies of scale, but also from ec<strong>on</strong>omies of scope. Data<br />
acquired for a particular purpose may be valuable in other c<strong>on</strong>texts,<br />
granting incumbent firms an advantage over new entrants in<br />
adjacent markets (Goldfarb et al., 2018). For instance, data collected<br />
in the c<strong>on</strong>text of search queries can be used to inform a shoppingrecommendati<strong>on</strong><br />
algorithm. Ec<strong>on</strong>omies of scope could leave little<br />
room for potential entrants looking to grow outside the incumbents’<br />
market segments.<br />
Sec<strong>on</strong>d, even if a firm successfully enters a new ML applicati<strong>on</strong><br />
market, incumbents may be in a positi<strong>on</strong> to use their rich data to<br />
detect the competitive threat, and acquire the new entrant firm<br />
before the incumbent’s positi<strong>on</strong> is challenged (so-called ‘killer<br />
acquisiti<strong>on</strong>s’).<br />
Implicati<strong>on</strong>s for competiti<strong>on</strong> policy[23]<br />
Absent interventi<strong>on</strong>, a possible market outcome could be high<br />
c<strong>on</strong>centrati<strong>on</strong> and low c<strong>on</strong>testability in data-reliant markets. This<br />
implies the need for competiti<strong>on</strong> policy scrutiny around data access.<br />
Should unique and n<strong>on</strong>-substitutable datasets be c<strong>on</strong>sidered an<br />
‘essential facility’, <strong>on</strong> par with local loops for fixed teleph<strong>on</strong>y?<br />
Forced sharing can create inefficiencies, e.g. in the form investment<br />
44 January 2021 | <str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g>
PERSPECTIVES ON ARTIFICIAL INTELLIGENCE<br />
disincentives, but is well established EU policy in many of the network industries. Under what c<strong>on</strong>diti<strong>on</strong>s would the benefits of forced<br />
sharing outweigh the costs?<br />
The emergence of ML as a general-purpose technology raises difficult empirical and normative questi<strong>on</strong>s. Does the relati<strong>on</strong>ship between<br />
data accumulati<strong>on</strong> and ec<strong>on</strong>omic returns give data-rich incumbents a significant and self-reinforcing advantage? Are competiti<strong>on</strong> authorities<br />
equipped to discern and analyse data-driven m<strong>on</strong>opolistic returns? These questi<strong>on</strong>s are high <strong>on</strong> the new European Commissi<strong>on</strong>’s agenda,[24]<br />
and for good reas<strong>on</strong>s. M<strong>on</strong>opolistic behaviour by ML providers could slow the adopti<strong>on</strong> of technology critical for EU competitiveness,<br />
especially hitting those smaller firms that lack the knowledge and resources to build alternative capacity in-house. If technological revoluti<strong>on</strong>s<br />
are distributi<strong>on</strong>al earthquakes, competiti<strong>on</strong> authorities should work to ensure that every<strong>on</strong>e lands <strong>on</strong> their feet.<br />
[1] This post focuses <strong>on</strong> ML applicati<strong>on</strong>s.<br />
[2] See https://medium.ec<strong>on</strong>omist.com/will-big-data-create-a-new-untouchable-business-elite-8dc23bcaa7cb<br />
[3] DG COMP 2019, citing https://medium.com/machine-intelligence-report/data-not-algorithms-is-key-to-machine-learning-success-69c6c4b79f33, https://www.<br />
edge.org/resp<strong>on</strong>se-detail/26587, and http://www.spacemachine.net/views/2016/3/datasets-over-algorithms.<br />
[4] While this post focuses <strong>on</strong> issues pertaining to the volume of data, other characteristics of data are just as important for generating value. These include the<br />
other so-called ‘4Vs’ of data: volume, but also velocity (i.e. frequency), variety (e.g. administrative data, social media data, pictures, etc), and veracity (i.e. representative<br />
of the target populati<strong>on</strong>, free of bias, etc). For a firm to have a competitive advantage over these other characteristics can also generate important ec<strong>on</strong>omic benefits.<br />
For the purpose of this blog, I note that securing a sufficient volume of data appears to be necessary but not sufficient to having a competitive AI/ML business.<br />
[5] Varian (2018) and Bajari et al. (2018)<br />
[6] In particular, Bajari et al. (2018) find that the length of histories is robustly helpful in improving the demand forecast quality, but at a diminishing rate; whereas<br />
the number of products in the same category is not (with a few excepti<strong>on</strong>s where it exhibits diminishing returns to scale).<br />
[7] Agrawal, Ajay, Joshua Gans, and Avi Goldfarb. 2018a. Predicti<strong>on</strong> Machines: The Simple Ec<strong>on</strong>omics of <strong>Artificial</strong> <strong>Intelligence</strong>. Cambridge, MA: Harvard Business<br />
Review Press.<br />
[8] As related in Goldfarb et al. (2018)<br />
[9] See the deal’s press release: https://news.microsoft.com/2009/07/29/microsoft-yahoo-change-search-landscape/<br />
[10] See https://www.cnet.com/news/googles-varian-search-scale-is-bogus/<br />
[11] See Hal Varian in a CNET interview: “the scale arguments are pretty bogus in our view” (https://www.cnet.com/news/googles-varian-search-scale-is-bogus/)<br />
[12] “the amount of traffic that Yahoo, say, has now is about what Google had two years ago” and “when we do improvements at Google, everything we do<br />
essentially is tested <strong>on</strong> a 1 percent or 0.5 percent experiment to see whether it’s really offering an improvement. So, if you’re half the size, well, you run a 2 percent<br />
experiment.” Source: ibid<br />
[13] i.e. in performing tasks such as crawling, index, or ranking.<br />
[14] The European Commissi<strong>on</strong>’s DG COMP made similar claims in the c<strong>on</strong>text of the Google Shopping case. DG COMP claimed that general search service has<br />
to receive at least a certain minimum volume of queries in order to improve the relevance of its results for uncomm<strong>on</strong> queries because users evaluate the relevance<br />
of a general search service <strong>on</strong> the basis of both comm<strong>on</strong> and uncomm<strong>on</strong> queries. See para. 288 of the EC decisi<strong>on</strong> (https://ec.europa.eu/competiti<strong>on</strong>/antitrust/cases/<br />
dec_docs/39740/39740_14996_3.pdf)<br />
[15] Cockburn et al. (2019).<br />
[16] An podcast episode from the Ec<strong>on</strong>omist brings this point to life (https://www.ec<strong>on</strong>omist.com/podcasts/2019/10/09/the-promise-and-peril-of-ai)<br />
[17] See Schaefler and al. (2018): “In perhaps no other market has the questi<strong>on</strong> of the role of data stirred such a vivid discussi<strong>on</strong> am<strong>on</strong>g industry participants,<br />
academic experts, and policy advocates than in general internet search.”<br />
[18] Glen Weyl and Eric Posner, 2018. Radical Markets<br />
[19] https://cs.stanford.edu/people/eroberts/cs181/projects/1997-98/microsoft-vs-doj/ec<strong>on</strong>omics/returns.html<br />
[20] See Calvano et al. (2020) for a survey of the literature around these issues in digital markets.<br />
[21] Commissi<strong>on</strong>er Terrell McSweeny, Opening Remarks for a Panel Discussi<strong>on</strong>, “Why Regulate Online Platforms?: Transparency, Fairness, Competiti<strong>on</strong>, or<br />
Innovati<strong>on</strong>?” at the CRA C<strong>on</strong>ference in Brussels, Belgium, at 5 (Dec. 9, 2015), https://www.ftc.gov/system/files/documents/public_statements/903953/mcsweeny_-_<br />
cra_c<strong>on</strong>ference_remarks_9-12-15.pdf.<br />
[22] includes data generated <strong>on</strong>line and by IoT and c<strong>on</strong>nected devices. Source: Word Ec<strong>on</strong>omic Forum citing Rac<strong>on</strong>teur (https://www.weforum.org/agenda/2019/04/<br />
how-much-data-is-generated-each-day-cf4bddf29f/)<br />
[23] Note that a range of issues lie at the intersecti<strong>on</strong> of privacy and competiti<strong>on</strong>, including data ownership, reuse, transparency, sharing. These issues are bey<strong>on</strong>d<br />
the score of this post and will not be explored here.<br />
[24] See the missi<strong>on</strong> statement of European Commissi<strong>on</strong> President Ursula v<strong>on</strong> der Leyen, which instructs Margarethe Vestager: “In the first 100 days of our<br />
mandate, you will coordinate the work <strong>on</strong> a European approach <strong>on</strong> artificial intelligence, including its human and ethical implicati<strong>on</strong>s. This should also look at how<br />
we can use and share n<strong>on</strong>-pers<strong>on</strong>alised big data to develop new technologies and business models that create wealth for our societies and our businesses.” (https://<br />
ec.europa.eu/commissi<strong>on</strong>/sites/beta-political/files/missi<strong>on</strong>-letter-margrethe-vestager_2019_en.pdf)<br />
Anders<strong>on</strong>, J. (2020), ‘The dynamics of data accumulati<strong>on</strong>’, Bruegel Blog, 11 February.<br />
https://bruegel.org/2020/02/the-dynamics-of-data-accumulati<strong>on</strong>/<br />
<str<strong>on</strong>g>OUR</str<strong>on</strong>g> <str<strong>on</strong>g>WORLD</str<strong>on</strong>g> | January 2021<br />
45
Fearless<br />
fear·less|ˈfir-ləs<br />
NEW EUROPE
WE SHARE THE SAME VISION<br />
WE GROW TOGETHER<br />
BETTER TOGETHER