- Dealing With Bias in Artificial Intelligence Three women with extensive experience in A.I. speak on the topic and how to confront it.
- History as a giant data set: how analysing the past could help save the future Fascinating long read about the theory that human history goes in cycles.
- The Royal Mint has melted down a million Brexit commemorative coins created for October 31, 2019 One of the lesser consequences of Brexit madness.
- The collapse of the information ecosystem poses profound risks for humanity intriguing metaphor: disinformation is to our media ecosystem as fossil fuels are to the environment.
Bias in machine learning
Nice example from Daphne Keller of Google:
Another notion of bias, one that is highly relevant to my work, are cases in which an algorithm is latching onto something that is meaningless and could potentially give you very poor results. For example, imagine that you’re trying to predict fractures from X-ray images in data from multiple hospitals. If you’re not careful, the algorithm will learn to recognize which hospital generated the image. Some X-ray machines have different characteristics in the image they produce than other machines, and some hospitals have a much larger percentage of fractures than others. And so, you could actually learn to predict fractures pretty well on the data set that you were given simply by recognizing which hospital did the scan, without actually ever looking at the bone. The algorithm is doing something that appears to be good but is actually doing it for the wrong reasons. The causes are the same in the sense that these are all about how the algorithm latches onto things that it shouldn’t latch onto in making its prediction.
Addressing bias in algorithms is crucial, especially in domains like healthcare where accurate predictions are vital. One effective approach to recognizing and mitigating biases is to rigorously test the algorithm in scenarios similar to its real-world applications. Suppose a machine-learning algorithm is trained on data from specific hospitals to predict fractures from X-ray images. In this case, it may appropriately incorporate prior knowledge about patient populations in those hospitals, resulting in reliable predictions within that context. However, the challenge arises when the algorithm is intended to be used in different hospitals not present in the initial training data set. To avoid unintended biases, a robust evaluation process is essential, and the use of a mobile learning management system can prove beneficial. Such a system enables continuous monitoring and assessment of the algorithm’s performance across various hospital settings, ensuring it doesn’t latch onto irrelevant factors and provides accurate predictions based on genuine medical insights.
To recognize and address these situations, you have to make sure that you test the algorithm in a regime that is similar to how it will be used in the real world. So, if your machine-learning algorithm is one that is trained on the data from a given set of hospitals, and you will only use it in those same set of hospitals, then latching onto which hospital did the scan could well be a reasonable approach. It’s effectively letting the algorithm incorporate prior knowledge about the patient population in different hospitals. The problem really arises if you’re going to use that algorithm in the context of another hospital that wasn’t in your data set to begin with. Then, you’re asking the algorithm to use these biases that it learned on the hospitals that it trained on, on a hospital where the biases might be completely wrong.
Facebook’s strategic obfuscation
Facebook’s Carolyn Everson, vice president of global marketing solutions, was interviewed by Peter Kafka at the 2019 Code Media conference in Los Angeles yesterday. Vox had a nice report of the interview. This section is particularly interesting:
When pressed on Facebook’s refusal to fact-check political ads, Everson tried to defend the company’s stance by referencing the rules that govern how broadcasters must handle political advertisements. In the US, the Federal Communications Commission has extensive guidelines for television and radio broadcasters around political advertising that bar broadcasters from censoring ads or from taking down ones that make false claims. Those guidelines don’t apply to online platforms, including Facebook, but the company has consistently tried to hide behind them.
“We have no ability, legally, to tell a political candidate that they are not allowed to run their ad,” Everson said.
That’s complete baloney. Facebook is not bound by any regulations governing TV ads. It can shut down anyone or anything it likes or dislikes.
After the interview, a Facebook spokeswoman walked back the comments and said that Everson misspoke when she said Facebook was legally barred from refusing to run political ads.
An audience member also asked Everson why Facebook has decided to allow right-wing website Breitbart to be listed in its new News tab, which is ostensibly an indication that Breitbart offers trusted news, despite being a known source of propaganda. “We’re treating them as a news source; I wouldn’t use the term ‘trusted news,’” Everson said, pointing out that Facebook will also include “far-left” publications.
Which of course raises interesting questions about Facebook’s standards for determining the “integrity” of the news sources it includes in its tab, which the company extolled when it launched the feature in October.
Linkblog
- For the Eurostar Tribe, Brexit Could Make Brussels Feel Even Farther Away
- The secret to enjoying a long winter. Lovely blog post by Jason Kottke.
- Hook, Line and Sinker: Do Tinder Matches and Meet Ups Lead to One-Night Stands? Looks like clickbait, but is actually a full-on academic article, complete with statistics — “means, SDS, and median numbers of matches, meet ups, one-night stands, and meetings with an interest for long-term committed relationship”, zzz. Since I’m not a Tinder user, all of this is purely theoretical.
- The cognitive impact of air pollution Alarming survey by Patrick Collison, co-founder of Stripe and surely one of the most cerebral of tech founders around.
Networked totalitarianism, contd.
From today’s New York Times:
“Ying shou jin shou” — “Round up everyone who should be rounded up.”
The echo of “1984,” “Brave New World” or “Fahrenheit 451” is unmistakable. But this is not dystopian fiction. It’s a real bureaucratic directive prepared by the Chinese leadership, drawing on a series of secret speeches by Xi Jinping, China’s authoritarian leader, on dealing ruthlessly with Muslims who show “symptoms” of religious radicalism.
There’s nothing theoretical about it: Based on these diktats, hundreds of thousands of Uighurs, Kazakhs and other Muslims in the western Xinjiang region have been rounded up in internment camps to undergo months or years of indoctrination intended to mold them into secular and loyal followers of the Communist Party.
And further to the revelations discussed in yesterday’s post,
Should students ask whether their missing parents had committed a crime, they are to be told no, “it is just that their thinking has been infected by unhealthy thoughts. Freedom is only possible when this ‘virus’ in their thinking is eradicated and they are in good health.”
The Times comments:
That someone from within the unforgiving, secretive Chinese leadership would take the enormous risk of leaking 403 pages of internal documents to a Western newspaper is in itself amazing, especially since the documents include an 11-page report summarizing the party’s investigation into the activities of Wang Yongzhi, an official who was supposed to manage a district where Uighur militants had staged a violent attack but who eventually developed misgivings about the mass detention facilities he had built. “He refused,” said the report, “to round up everyone who should be rounded up.” After September 2017, Mr. Wang disappeared from public view.
Quote of the Day
“Economic growth, democracy, and CO2 have always been intertwined. Growth and democracy barely existed until coal fuelled the industrial revolution. Can democracy survive without carbon? We are not going to find out. No electorate will vote to decimate its own lifestyle. We can’t blame bad politicians or corporates. It us: we will always choose growth over climate.”
- Simon Kuper, “The Myth of Green Growth”, Financial Times, October 26/27, 2019.
Linkblog
- How to report a tech conference A masterclass in insightful summarising by Steven Sinofsky
- In 2029, AI Will Make Prejudice Much Worse Credibly dystopian?
- Why we must all reduce our moral certainties If we want to live in a democracy, we have to reach compromises that make most people feel secure. Interesting essay on openDemocracy.
- Against Economics David Graeber’s long review in the New York Review of Books of Robert Skidelsky’s new book, Money and Government: The Past and Future of Economics
Quote of the Day
“The forest was shrinking, but the trees kept voting for the axe, who had convinced them that, because his handle was made of wood, he was one of them”
- Turkish proverb
Networked totalitarianism: the script
The New York Times’s scoop on the treatment of China’s Uyghur population is an eye-opener, even to those of us who suspected the worst.
The students booked their tickets home at the end of the semester, hoping for a relaxing break after exams and a summer of happy reunions with family in China’s far west.
Instead, they would soon be told that their parents were gone, relatives had vanished and neighbors were missing — all of them locked up in an expanding network of detention camps built to hold Muslim ethnic minorities.
The authorities in the Xinjiang region worried the situation was a powder keg. And so they prepared.
The leadership distributed a classified directive advising local officials to corner returning students as soon as they arrived and keep them quiet. It included a chillingly bureaucratic guide for how to handle their anguished questions, beginning with the most obvious: Where is my family?
Samples:
”They’re in a training school set up by the government”. If pressed, officials were to explain that the parents were not criminals, but were nevertheless not free to leave the “schools”.
The question-and-answer script also included a barely-concealed threat: students were to be told that their behaviour could either shorten or extend the detention of their relatives. “I am sure that you will support them, because this is for their own good”, officials were advised to say, “and also for your own good”.
Lots more in that vein.
So why are Western countries still treating China as a proper member of the international community? (The answer is that one has to play nice with a superpower, I guess.) And what will it take to change things? Answer: probably only a brutal Tienanmen-style suppression of the protests in Hong Kong would have that effect.
On the other hand, it’s not that long ago that the UK had a system of interning people without trial in Northern Ireland. But at least it had the excuse of a genuine emergency. As far as one can see, the Uighurs don’t pose any threat to the Chinese state.
Our feudal present
In late August, a black-sailed ship appeared in the harbor carrying a 16-year-old visionary, a girl who had sailed from the far north across a great sea. A mass of city-dwellers and travelers, enthralled by her prophecies, gathered to welcome her. She had come to speak to the nations of Earth, to castigate us for our vanities and warn us of coming catastrophe. “There were four generations there cheering and chanting that they loved her,” the writer Dean Kissick observed. “When she came ashore, it felt messianic.”
Thus begins “In 2029, the Internet Will Make Us Act Like Medieval Peasants”, a lovely essay by Max Read on what tech is doing to us.
“Looking around lately”, he writes,
I am reminded less often of Gibson’s cyberpunk future than of J.R.R. Tolkien’s fantastical past, less of technology and cybernetics than of magic and apocalypse. The internet doesn’t seem to be turning us into sophisticated cyborgs so much as crude medieval peasants entranced by an ever-present realm of spirits and captive to distant autocratic landlords. What if we aren’t being accelerated into a cyberpunk future so much as thrown into some fantastical premodern past?
Wonderful stuff. Worth reading in full.