Ben Lorica, O'Reilly's chief data scientist, has posted slides and notes from his talk at last December's Strata Data Conference in Singapore, "We need to build machine learning tools to augment machine learning engineers." Read the rest
In 2015, a black software developer named Jacky Alciné revealed that the image classifier used by Google Photos was labeling black people as "gorillas."
In ADGN: An Algorithm for Record Linkage Using Address, Date of Birth, Gender, and Name, newly published in Statistics and Public Policy, a pair of researchers from Harvard and Tufts build a statistical model to analyze the impact of the voter ID laws passed in Republican-controlled states as part of a wider voter suppression project that was explicitly aimed at suppressing the votes of racialised people, historically likely to vote Democrat. Read the rest
More than a dozen major US airports are now covered in facial-recognition cameras, installed by the DHS to scan people departing on international flights without the legally mandated federal review process. Read the rest
Designers use metrics and a/b splitting experiments to maximize engagement with their products, seeking out the phrases that trigger emotional responses in users -- like a smart scale that congratulates you on losing weight -- but these systems are context-blind, so they are unable to distinguish between people who might be traumatized by their messages (say, a woman who's just miscarried late in her pregnancy, being congratulated on her "weight loss"). Read the rest
"The Trouble with Bias," Kate Crawford's (previously) keynote at the 2017 Neural Information Processing Systems is a brilliant tour through different ways of thinking about what bias is, and when we should worry about it, specifically in the context of machine learning systems and algorithmic decision making -- the best part is at the end, where she describes what we should do about this stuff, and where to get started. (via 4 Short Links) Read the rest
In a pair of open letters to Letter to The Honorable Elaine C. Duke, Acting Secretary of Homeland, a coalition of more than 100 tech liberties groups and leading technology experts urged the DHS to abandon its plan to develop a black-box algorithmic system for predicting whether foreigners coming to the USA to visit or live are likely to be positive contributors or risks to the nation. Read the rest
There are 50 hospitals on 5 continents that use Watson for Oncology, an IBM product that charges doctors to ingest their cancer patients' records and then make treatment recommendations and suggest journal articles for further reading. Read the rest
A year ago, the AI Now Institute released its inaugural report on the near-future social and economic consequences of AI, drawing on input from a diverse expert panel representing a spectrum of disciplines; now they've released a followup, with ten clear recommendations for AI implementations in the public and private sector. Read the rest
Rodney Brooks -- eminent computer scientist and roboticist who has served as head of MIT's Computer Science and Artificial Intelligence Laboratory and CTO of Irobot -- has written a scorching, provocative list of the seven most common errors made (or cards palmed) by pundits and other fortune-tellers when they predict the future of AI. Read the rest
When the 9/11 commission reported back on the intelligence failures that led to the attacks 16 years ago, they identified a key problem: America's spy agencies had collected so much useless, indiscriminate information (haystacks) that they couldn't find the useful, salient facts (needles) buried there. Read the rest
The Perspective API (previously) is a tool from Google spinoff Jigsaw (previously) that automatically rates comments for their "toxicity" -- a fraught business that catches a lot of dolphins in its tuna net. Read the rest
Once big data systems agglomerate enough data about you to predict whether you are likely to get sick or badly injured, insurers will be able to deny coverage (or charge so much for it that it amounts to the same thing) to anyone who is likely to get sick, forcing everyone who might ever need insurance into medical bankruptcy, and turning Medicaid into a giant "high-risk pool" that taxpayers foot the bill for. Read the rest
Nesta's Juan Mateos-Garcia proposes that "entropic forces" make algorithmic decision-making tools worse over time, requiring that they be continuously maintained and improved (this is also a key idea from Cathy O'Neil's Weapons of Math Destruction: a machine-learning system is only honest if someone is continuously matching its predictions to reality and refining its model based on the mistakes it makes). Read the rest
Eminent computer scientist Ed Felten has posted a short, extremely useful taxonomy of four ways that an algorithm can fail to be accountable to the people whose lives it affects: it can be protected by claims of confidentiality ("how it works is a trade secret"); by complexity ("you wouldn't understand how it works"); unreasonableness ("we consider factors supported by data, even when you there's no obvious correlation"); and injustice ("it seems impossible to explain how the algorithm is consistent with law or ethics"). Read the rest
The best way to fight gerrymandering is to prove to courts that electoral districts have been unfairly formed, a tactic that's been used successfully in places like North Carolina; but for this to work, you need good demographic data to show that the district is unfair, and for that, you need an accurate census. Read the rest
