Excavating AI

Fabulous essay by Kate Crawford and Trevor Paglen, uncovering the politics and biases embedded in the guge image databases that have been used for training machine learning software. Here’s how it begins:

You open up a database of pictures used to train artificial intelligence systems. At first, things seem straightforward. You’re met with thousands of images: apples and oranges, birds, dogs, horses, mountains, clouds, houses, and street signs. But as you probe further into the dataset, people begin to appear: cheerleaders, scuba divers, welders, Boy Scouts, fire walkers, and flower girls. Things get strange: A photograph of a woman smiling in a bikini is labeled a “slattern, slut, slovenly woman, trollop.” A young man drinking beer is categorized as an “alcoholic, alky, dipsomaniac, boozer, lush, soaker, souse.” A child wearing sunglasses is classified as a “failure, loser, non-starter, unsuccessful person.” You’re looking at the “person” category in a dataset called ImageNet, one of the most widely used training sets for machine learning.

Something is wrong with this picture.

Where did these images come from? Why were the people in the photos labeled this way? What sorts of politics are at work when pictures are paired with labels, and what are the implications when they are used to train technical systems?

In short, how did we get here?

The authors begin with a deceptively simple question: What work do images do in AI systems? What are computers meant to recognize in an image and what is misrecognised or even completely invisible? They examine the methods used for introducing images into computer systems and look at “how taxonomies order the foundational concepts that will become intelligible to a computer system”. Then they turn to the question of labeling: “how do humans tell computers which words will relate to a given image? And what is at stake in the way AI systems use these labels to classify humans, including by race, gender, emotions, ability, sexuality, and personality?” And finally, they turn to examine the purposes that computer vision is meant to serve in our society and interrogate the judgments, choices, and consequences of providing computers with these capacities.

This is a really insightful and sobering essay, based on extensive research.

Some time ago Crawford and Paglen created an experimental website — ImageNet Roulette — which enabled anyone to upload their photograph and then pulled up from the ImageNet database how the person would be classified based on their photograph. The site is now offline, but the Guardian journalist Julia Carrie Wong wrote an interesting article about it recently in the course of which she investigated how it would classify/describe her from her Guardian byline photo. Here’s what she found.

Interesting ne c’est pas? Remember, this is the technology underpinning facial recognition.

Do read the whole thing.

Quantum supremacy?

This morning’s Observer column:

Something intriguing happened last week. A paper about quantum computing by a Google researcher making a startling claim appeared on a Nasa website – and then disappeared shortly afterwards. Conspiracy theorists immediately suspected that something sinister involving the National Security Agency was afoot. Spiritualists thought that it confirmed what they’ve always suspected about quantum phenomena. (It was, as one wag put it to me, a clear case of “Schrödinger’s Paper”.) Adherents of the cock-up theory of history (this columnist included) concluded that someone had just pushed the “publish” button prematurely, a suspicion apparently confirmed later by stories that the paper was intended for a major scientific journal before being published on the web.

Why was the elusive paper’s claim startling? It was because – according to the Financial Times – it asserted that a quantum computer built by Google could perform a calculation “in three minutes and 20 seconds that would take today’s most advanced classical computer … approximately 10,000 years”. As someone once said of the book of Genesis, this would be “important if true”. A more mischievous thought was: how would the researchers check that the quantum machine’s calculation was correct?

A quantum computer is one that harnesses phenomena from quantum physics, the study of the behaviour of subatomic particles, which is one of the most arcane specialisms known to humankind…

Read on

Computational propaganda continues to increase — and evolve

A new report from the Computational Propaganda group at the Oxford Internet Institute shows that states are increasingly using weaponising social media for information supression, disinformation and political manipulation. The researchers found “evidence of organized social media manipulation campaigns which have taken place in 70 countries, up from 48 countries in 2018 and 28 countries in 2017. In each country, there is at least one political party or government agency using social media to shape public attitudes domestically.”

Other findings:

  • social media has been exploited by authoritarian regimes in 26 countries to suppress basic human rights, discredit political opponents and drown out dissenting opinions.

  • A handful of sophisticated state actors use computational propaganda for foreign influence operations. Facebook and Twitter attributed foreign influence operations to seven countries (China, India, Iran, Pakistan, Russia, Saudi Arabia, and Venezuela) who have used these platforms to influence global audiences.

  • China has become a major player in the global disinformation order. Until the 2019 protests in Hong Kong, most evidence of Chinese computational propaganda occurred on domestic platforms such as Weibo, WeChat, and QQ. But China’s new-found interest in aggressively using Facebook, Twitter, and YouTube should raise concerns for democracies.

  • Facebook remains the platform of choice for social media manipulation. In 56 countries, the researchers found evidence of formally organized computational propaganda campaigns on Facebook. Interestingly, the exploitation of Facebook’s targeted advertising machineseems to be on the decline. In the case studies the researchers studied, advertising was not central to the spread of disinformation. Instead the campaigns created memes, videos or other kinds of content tailored to exploit platforms’ algorithms and their amplifying effects — effectively getting virality for free.

There’s a good NYT report summarising the researchers’ findings.

Liberal delusions

“What it’s like to take the Aspen Institute Executive Seminar: It’s like the edited highlights of a humanities course at a liberal arts college, apparently. You fill in a Goopy-sounding questionnaire about your “leadership journey”; you discuss texts from Plato, Marx and Hobbes; on day three or four the people in your group start getting on one another’s nerves and breaking down in tears; you stage a potted version of Antigone to round off the week; you go home better-connected. Price: $11,350 all in.”

Linda Kinstler

Some institutions still work

Conor Gearty (an eminent human rights lawyer) wrote an interesting blog post about the Supreme Court’s decision that Johnson’s advice to the Queen on proroguing Parliament was unlawful. Excerpt:

Why did the Court do it? The constitutional reason – an entirely good one – is that the Court has deduced from the fundamental principles of representative democracy and accountable government a set of constraints on power that flow from these principles and which must, as a result, adhere to all exercises of public power, including those of the most senior political figures in the land (paras 41 and 46).

The deeper truth lying behind how these principles were deployed in this case leads us to something that was once a commonplace but these days is a glory rarely to be found in the shrill word of Brexit politics. In law, reason still matters. Facts are relevant. Nonsense doesn’t work. How can you justify the Prime Minister’s power by saying he is accountable to Parliament when you have just dispensed with Parliament? Why on earth do you need to cancel Parliament for weeks to do a Queen’s speech? Deceitful or deliberately obtuse replies to these basic questions might get you through a three-minute media interview or a noisy prime minister’s question time, but they can’t survive the forensic attentions of independently-minded lawyers with time to draw the non sequiturs, the contradictions and the lies to the surface.

This case is not about the judges seizing the policy agenda whatever the critics of the outcome might say. It is concerned with process not substance, with how things get done rather than what is done. Strongly hostile to democracy in days gone by, the judiciary have now embraced its fundamental tenets, taking to heart what we all say matters to us. In this decision, the judges are oiling the democratic machine, not telling it what to produce.

Great stuff. Worth reading in full.

Brexit, Trump, Johnson and civic virtue

Pondering the implications of the UK Supreme Court’s unanimous judgment this morning that Boris Johnson’s advice to the Queen to prorogue Parliament was unlawful, it was interesting to see the resurgence of the question of whether Johnson would obey the law — surely the first time that question has been asked seriously in recent British history. (The same question also reoccurs regularly in relation to Trump.) A conversation this morning with the philosopher Tom Simpson led me to read an article of his in which he critiques the political philosopher Philip Petit’s view on republican freedom. In the article, Simpson quotes a passage from Volume 1 of Quentin Skinner’s Foundations of Modern Political Thought in which he identifies two main approaches to virtue and corruption in political thought since the Renaissance:

”One stresses that government is effective whenever its institutions are strong, and corrupt whenever its machinery fails to function adequately. (The greatest exponent of this outlook is Hume.) The other approach suggests by contrast that if the men who control the institutions of government are corrupt, the best possible institutions cannot be expected to shape or constrain them, whereas if the men are virtuous, the health of the institutions will be a matter of secondary importance. This is the tradition (of which Machiavelli and Montesquieu are the greatest representatives) which stresses that it is not so much the machinery of government as the proper spirit of the rulers, the people and the laws which needs above all to be sustained.”

That passage — “if the men who control the institutions of government are corrupt, the best possible institutions cannot be expected to shape or constrain them, whereas if the men are virtuous, the health of the institutions will be a matter of secondary importance” — seems to me to be the key to our current crisis. We have seen that some of the institutions of American democracy (Congress, the Courts) are having trouble controlling Trump, who is clearly corrupt. The question for today is whether the UK Supreme Court and the conventions of the UK’s unwritten patchwork-quilt of a ‘constitution’ will be able to control the equally meretricious Johnson (and his equally unscrupulous Svengali, Dominic Cummings).

Watch this space, because at the moment nobody knows.