13-10-2018
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
SCIENCE & TECH<br />
SATUrDAY, OcTOBer <strong>13</strong>, <strong>2018</strong><br />
5<br />
Tech’s gender bias nothing new<br />
Marie Hicks<br />
A recent report revealed Amazon's AI recruiting<br />
technology developed a bias against women because it was<br />
trained predominantly on men's resumes. Although<br />
Amazon shut the project down, this kind of mechanized<br />
sexism is common and growing - and the problem isn't<br />
limited to AI mishaps.<br />
Facebook allows the targeting of job ads by gender,<br />
resulting in discrimination in online job advertisements<br />
for traditionally male-dominated jobs from construction<br />
to policing. The practice has long been illegal in traditional<br />
print media - but Facebook's targeting tools encourage it.<br />
Not only can this affect whether women and non-binary<br />
people can see ads; it also affects male job-seekers who are<br />
older and therefore viewed as less desirable by many<br />
employers. Facebook has come under fire for illegal<br />
advertising practices in the past: notably, it scrapped<br />
thousands of microtargeting categories after a 2016<br />
ProPublica report showed how it allowed racial<br />
discrimination in housing ads.<br />
The platform has repeatedly refused to take<br />
responsibility for what people do on it, echoing the<br />
behavior of other Silicon Valley companies. Gendered and<br />
racialized harassment online goes largely unchecked.<br />
Likewise, Google's YouTube has come under fire for<br />
algorithms that appear to push radicalizing far-right<br />
content onto casual viewers, while Google itself has faced<br />
accusations that its image search and autocomplete<br />
features rely on and strengthen racist and sexist<br />
stereotypes.<br />
As online platforms strip away civil rights protections<br />
intended to correct biases in earlier forms of<br />
communication, it serves as an example of the dangerous<br />
tendency of our current, and supposedly progressive,<br />
technologies to re-create discriminatory patterns of the<br />
past. Indeed, these problems fit a pattern in the long<br />
trajectory of the history of technology.<br />
Today, jobs in computing, if advertised on Facebook,<br />
would likely be targeted to men because these jobs are<br />
located in an already male-dominated field. In the early<br />
days of electronic computing, however, the work was<br />
strongly associated with women. It was feminized because<br />
it was seen as deskilled and unimportant. This quickly<br />
began to change as computers became indispensable in all<br />
areas of government and industry. Once it became clear<br />
that those who knew how to use them would have great<br />
power and influence, female programmers lost out despite<br />
having all the requisite skills. Britain's computerization is<br />
a cautionary tale: women were repeatedly and<br />
progressively denied promotions or boxed out of their<br />
jobs, particularly when they married or had children.<br />
Top executives of Facebook, Amazon and Alphabets Inc during a meeting at Trump Tower.<br />
Photo: Shannon Stapleton<br />
When they left, they were replaced by men. This created<br />
disastrous labor shortages that ultimately forced Britain's<br />
decline as a computing superpower.<br />
Women continued to program, but they had to do it<br />
without the support of major institutions. One example<br />
was the entrepreneur Stephanie "Steve" Shirley, who used<br />
a masculine nickname to sidestep sexism. Shirley started a<br />
freelance programming company with an explicitly<br />
feminist business model after finding herself unable to<br />
advance in government and industry. She employed<br />
hundreds of other women who had similarly had to leave<br />
the workforce. Shirley gave these women an opportunity<br />
to use their skills in the service of the nation's economy by<br />
giving them the option to work from home, filling some of<br />
the gaps left by this exodus of trained computer<br />
professionals from full-time computing work.<br />
Shirley's business, built on women's labor and<br />
expertise, went on to become a multimillion-dollar<br />
corporation that did mission-critical programming for<br />
government and private industry. As the government<br />
scrambled for male computing talent, for instance, a team<br />
of her female programmers, led by Ann Moffatt,<br />
programmed the black box for the Concorde jet. As<br />
Shirley's business flourished, many other companies and<br />
even the British government itself suffered for lack of<br />
programming talent.<br />
The irony is that this shortage had been intentionally<br />
engineered by the refusal to continue to employ female<br />
technologists in these newly prestigious jobs. Throughout<br />
history, when jobs are seen as more important, or are<br />
better paid, women are squeezed out - hence the need for<br />
protective legislation that ensures equality of opportunity<br />
in hiring and job advertisements.<br />
In computing today, a field that claims to value diversity,<br />
engineers at Facebook and other companies are building<br />
tools that rollback the advances of women in the<br />
workforce, as the industry undoes the civil rights<br />
protections enacted to ensure that what happened in early<br />
computing does not happen again.<br />
When industries ignore their pasts, they tend not only<br />
to repeat previous mistakes, but also to worsen current<br />
problems. Silicon Valley's gender problems are well<br />
known, and despite companies' claims that they are<br />
trying to address the problem, progress has been slow<br />
and uneven. This is not surprising when we consider the<br />
context. Although the industry is facing a reckoning<br />
today, for decades the stories that we told about<br />
computing technology focused on inexorable success,<br />
rather than taking seriously the possibility that our new<br />
technologies were failing us. High technology became<br />
virtually synonymous with progress and the greater<br />
application of computing to all manner of social problems<br />
was seen as a good in and of itself. As a result, we are<br />
largely blind to the errors of the past. We fail to see the<br />
problems in our present and the reasons behind them<br />
because we are too accustomed to seeing computing as a<br />
success story.<br />
The refusal to talk about computing's failures in the past<br />
has not served us, or present-day computing, well. Rather,<br />
it has hidden problems that have plagued the field since its<br />
inception. Facebook's discriminatory practices towards<br />
female users in everything from job advertisements to<br />
harassment can be traced back to its predecessor, the beta<br />
site set up by Mark Zuckerberg while at Harvard that stole<br />
female undergrads' pictures from internal Harvard<br />
servers. The site, known as Facemash, objectified the<br />
women for an audience invited to rate their relative<br />
attractiveness. When we consider Facebook's current<br />
problems in this light, they not only seem less surprising<br />
but also potentially more solvable.<br />
Lessons like this are critical today because high<br />
technology has an outsize effect on every aspect of our<br />
daily lives, and it is also, in many ways, steadily moving us<br />
back towards a past that we thought we had forgotten.<br />
Much of the anti-racist and anti-sexist legislation of the<br />
20th century has been invisibly rolled back by tech<br />
infrastructures that invite users to see their online actions<br />
as unmoored from real life - whether in the realm of hate<br />
speech or job advertisements.<br />
Strong representation of women in the labor market is<br />
key, historically and today, for women to be able to assert<br />
their rights in all aspects of their lives. Companies like<br />
Facebook cannot be allowed to divide and conquer by<br />
gender, race, sexuality, age, disability, or any other<br />
number of categories people have fought to protect by law<br />
as deserving of equal rights.<br />
Weaponized AI enabling<br />
perpetual wars<br />
A technician checks a server in a data centre.<br />
Do DWeb programs use as much<br />
energy as cloud-based services?<br />
Jack Schofield<br />
The main aim of the decentralised web<br />
(DWeb) is to remove the power of<br />
centralised "gatekeepers" such as<br />
Facebook and Google, who hoover up<br />
the world's data and monetise it by<br />
selling advertising. It reminds me of the<br />
original concept of the web, where every<br />
computer would be both a client and a<br />
server, sharing information on a more or<br />
less equal basis. Of course, that is not<br />
how real life works. What actually<br />
happens is that you get a power law<br />
distribution with a few large entities and<br />
a long tail of small ones.<br />
As Clay Shirky wrote in 2003: "In<br />
systems where many people are free to<br />
choose between many options, a small<br />
subset of the whole will get a<br />
disproportionate amount of traffic (or<br />
attention, or income), even if no<br />
members of the system actively work<br />
towards such an outcome. This has<br />
nothing to do with moral weakness,<br />
selling out, or any other psychological<br />
explanation. The very act of choosing,<br />
spread widely enough and freely<br />
enough, creates a power law<br />
distribution."<br />
The web still has plenty of variety, but<br />
almost everyone is familiar with one<br />
giant search engine, one giant retailer,<br />
one giant auction site, one giant social<br />
network, one giant encyclopaedia, and<br />
so on. Indeed, there is only one giant<br />
internet where there used to be dozens of<br />
competing networks using many<br />
different protocols.<br />
Obviously, it would be better if we all<br />
agreed these things in advance, based on<br />
open standards. However, people vote<br />
with their wallets, and competition<br />
results in de facto standards instead of<br />
de jure ones. Examples include<br />
Microsoft Windows, Google Search and<br />
Facebook. Each triumphed in a<br />
competitive marketplace. I am not<br />
saying this is the ideal solution, just that,<br />
in most cases, it's inevitable.<br />
One of the problems with returning to<br />
a decentralised web is that the internet is<br />
no longer decentralised. It has been<br />
redesigned around giant server farms,<br />
high-speed pipes and content delivery<br />
networks. It looks increasingly like a<br />
broadband television network because<br />
that is what it actually does most of the<br />
time.<br />
Today's web is being optimised for the<br />
delivery of Netflix movies, BBC<br />
programmes on iPlayer, Spotify music,<br />
live streams of every major sporting<br />
event, and so on. You can upload your<br />
own live streams but communications<br />
are asynchronous: your downloads are<br />
much faster, and much more reliable,<br />
than your uploads. It's really easy to<br />
watch 1TB of movies but an exercise in<br />
frustration trying to upload a 1TB harddrive<br />
backup.<br />
If you really want to save energy and<br />
internet resources, stop streaming stuff.<br />
Broadcast TV and radio can reach tens of<br />
millions of people, and adding another<br />
million adds relatively little in the way of<br />
extra power consumption. There is<br />
school of thought that it is better for the<br />
environment to use CDs or DVDs for<br />
albums or films you go back to again and<br />
again, or you could at least use digital<br />
files stored on your PC or smartphone.<br />
Photo: Juice Images<br />
And rather than using Graphite to<br />
replace Google Docs or Microsoft Office,<br />
just use a word processor offline. If you<br />
run Windows, you already have a text<br />
editor (Notepad) and a simple word<br />
processor (WordPad), and there are<br />
plenty of free alternatives. That will<br />
reduce global energy use and increase<br />
your privacy.<br />
It's really simple. If you don't want<br />
Google to read your documents, don't<br />
write your documents on Google's<br />
computers. And if you don't want cloud<br />
servers using energy, don't use the cloud.<br />
Companies such as Amazon AWS,<br />
Microsoft and Google are covering the<br />
world with server farms to make<br />
information more easily available. That's<br />
harder to do with real distributed<br />
systems because the thousands or<br />
millions of separate computers may be<br />
turned off or otherwise unavailable<br />
when you need the data they are storing.<br />
Worse, unless it's replicated, you could<br />
lose data.<br />
It's true that server farms consume<br />
an ever-growing amount of electricity,<br />
much of it used for cooling purposes.<br />
However, the cost is a powerful<br />
incentive for operators to use cheaper<br />
renewables, such as solar panels, and<br />
to reduce their power consumption in<br />
other ways. For example, Facebook<br />
has built a data centre in the north of<br />
Sweden where the air is freezing cold,<br />
while Microsoft is experimenting with<br />
underwater data centres that are easier<br />
to cool. Microsoft is also sponsoring<br />
tree planting in Ireland as part of its<br />
commitment to becoming carbon<br />
neutral.<br />
Ben Tarnoff<br />
Last month marked the 17th<br />
anniversary of 9/11. With it came a new<br />
milestone: we've been in Afghanistan<br />
for so long that someone born after the<br />
attacks is now old enough to go fight<br />
there. They can also serve in the six<br />
other places where we're officially at<br />
war, not to mention the <strong>13</strong>3 countries<br />
where special operations forces have<br />
conducted missions in just the first half<br />
of <strong>2018</strong>.<br />
The wars of 9/11 continue, with no<br />
end in sight. Now, the Pentagon is<br />
investing heavily in technologies that<br />
will intensify them. By embracing the<br />
latest tools that the tech industry has to<br />
offer, the US military is creating a more<br />
automated form of warfare - one that<br />
will greatly increase its capacity to wage<br />
war everywhere forever.<br />
On Friday, the defense department<br />
closes the bidding period for one of the<br />
biggest technology contracts in its<br />
history: the Joint Enterprise Defense<br />
Infrastructure (Jedi). Jedi is an<br />
ambitious project to build a cloud<br />
computing system that serves US forces<br />
all over the world, from analysts behind<br />
a desk in Virginia to soldiers on patrol in<br />
Niger. The contract is worth as much as<br />
$<strong>10</strong>bn over <strong>10</strong> years, which is why big<br />
tech companies are fighting hard to win<br />
it. (Not Google, however, where a<br />
pressure campaign by workers forced<br />
management to drop out of the<br />
running.)<br />
At first glance, Jedi might look like<br />
just another IT modernization project.<br />
Government IT tends to run a fair<br />
distance behind Silicon Valley, even in a<br />
place as lavishly funded as the<br />
Pentagon. With some 3.4 million users<br />
and 4 million devices, the defense<br />
department's digital footprint is<br />
immense. Moving even a portion of its<br />
workloads to a cloud provider such as<br />
Amazon will no doubt improve<br />
efficiency.<br />
But the real force driving Jedi is the<br />
desire to weaponize AI - what the<br />
defense department has begun calling<br />
"algorithmic warfare". By pooling the<br />
military's data into a modern cloud<br />
platform, and using the machinelearning<br />
services that such platforms<br />
provide to analyze that data, Jedi will<br />
help the Pentagon realize its AI<br />
ambitions.<br />
The scale of those ambitions has<br />
grown increasingly clear in recent<br />
months. In June, the Pentagon<br />
established the Joint Artificial<br />
Intelligence Center (JAIC), which will<br />
oversee the roughly 600 AI projects<br />
currently under way across the<br />
department at a planned cost of $1.7bn.<br />
And in September, the Defense<br />
Advanced Research Projects Agency<br />
(Darpa), the Pentagon's storied R&D<br />
wing, announced it would be investing<br />
up to $2bn over the next five years into<br />
AI weapons research.<br />
So far, the reporting on the Pentagon's<br />
AI spending spree has largely focused<br />
on the prospect of autonomous<br />
weapons - Terminator-style killer<br />
robots that mow people down without<br />
any input from a human operator. This<br />
is indeed a frightening near-future<br />
scenario, and a global ban on<br />
autonomous weaponry of the kind<br />
sought by the Campaign to Stop Killer<br />
Robots is absolutely essential.<br />
But AI has already begun rewiring<br />
warfare, even if it hasn't (yet) taken the<br />
form of literal Terminators. There are<br />
less cinematic but equally scary ways to<br />
weaponize AI. You don't need<br />
algorithms pulling the trigger for<br />
algorithms to play an extremely<br />
dangerous role.<br />
To understand that role, it helps to<br />
understand the particular difficulties<br />
posed by the forever war. The killing<br />
itself isn't particularly difficult. With a<br />
military budget larger than that of<br />
China, Russia, Saudi Arabia, India,<br />
France, Britain and Japan combined,<br />
and some 800 bases around the world,<br />
the US has an abundance of firepower<br />
and an unparalleled ability to deploy<br />
that firepower anywhere on the planet.<br />
The US military knows how to kill.<br />
The harder part is figuring out whom to<br />
kill. In a more traditional war, you<br />
simply kill the enemy. But who is the<br />
enemy in a conflict with no national<br />
boundaries, no fixed battlefields, and no<br />
conventional adversaries?<br />
This is the perennial question of the<br />
forever war. It is also a key feature of its<br />
design. The vagueness of the enemy is<br />
what has enabled the conflict to<br />
continue for nearly two decades and to<br />
expand to more than 70 countries - a<br />
boon to the contractors, bureaucrats<br />
and politicians who make their living<br />
from US militarism. If war is a racket, in<br />
the words of marine legend Smedley<br />
Butler, the forever war is one the longest<br />
cons yet.<br />
Automation has greatly increased US Military's capacity to wage war everywhere forever.<br />
Photo: Getty