Archives 22 May 2024

In October 2003, Mark Zuckerberg created his first viral site: not Facebook, but FaceMash. Then a college freshman, he hacked into Harvard’s online dorm directories, gathered a massive collection of students’ headshots, and used them to create a website on which Harvard students could rate classmates by their attractiveness, literally and figuratively head-to-head. The site, a mean-spirited prank recounted in the opening scene of The Social Network, got so much traction so quickly that Harvard shut down his internet access within hours. The math that powered FaceMash — and, by extension, set Zuckerberg on the path to building the world’s dominant social-media empire — was reportedly, of all things, a formula for ranking chess players: the Elo system.

Fundamentally, what an Elo rating does is predict the outcome of chess matches by assigning every player a number that fluctuates based purely on performance. If you beat a slightly higher-ranked player, your rating goes up a little, but if you beat a much higher-ranked player, your rating goes up a lot (and theirs, conversely, goes down a lot). The higher the rating, the more matches you should win. That is what Elo was designed for, at least. FaceMash and Zuckerberg aside, people have deployed Elo ratings for many sports — soccer, football, basketball — and for domains as varied as dating, finance, and primatology. If something can be turned into a competition, it has probably been Elo-ed. Somehow, a simple chess algorithm has become an all-purpose tool for rating everything. In other words, when it comes to the preferred way to rate things, Elo ratings have the highest Elo rating. […]

Elo ratings don’t inherently have anything to do with chess. They’re based on a simple mathematical formula that works just as well for any one-on-one, zero-sum competition — which is to say, pretty much all sports. In 1997, a statistician named Bob Runyan adapted the formula to rank national soccer teams — a project so successful that FIFA eventually adopted an Elo system for its official rankings. Not long after, the statistician Jeff Sagarin applied Elo to rank NFL teams outside their official league standings. Things really took off when the new ESPN-owned version of Nate Silver’s 538 launched in 2014 and began making Elo ratings for many different sports. Some sports proved trickier than others. NBA basketball in particular exposed some of the system’s shortcomings, Neil Paine, a stats-focused sportswriter who used to work at 538, told me. It consistently underrated heavyweight teams, for example, in large part because it struggled to account for the meaninglessness of much of the regular season and the fact that either team might not be trying all that hard to win a given game. The system assumed uniform motivation across every team and every game. Pretty much anything, it turns out, can be framed as a one-on-one, zero-sum game.
Arpad Emmerich Elo, creator of the Elo rating system, understood the limitations of his invention. “It is a measuring tool, not a device of reward or punishment,” he once remarked. “It is a means to compare performances, assess relative strength, not a carrot waved before a rabbit, or a piece of candy given to a child for good behavior.”

Axon on Tuesday announced a new tool called Draft One that uses artificial intelligence built on OpenAI’s GPT-4 Turbo model to transcribe audio from body cameras and automatically turn it into a police report. Axon CEO Rick Smith told Forbes that police officers will then be able to review the document to ensure accuracy. From the report:
Axon claims one early tester of the tool, Fort Collins Colorado Police Department, has seen an 82% decrease in time spent writing reports. “If an officer spends half their day reporting, and we can cut that in half, we have an opportunity to potentially free up 25% of an officer’s time to be back out policing,” Smith said. These reports, though, are often used as evidence in criminal trials, and critics are concerned that relying on AI could put people at risk by depending on language models that are known to “hallucinate,” or make things up, as well as display racial bias, either blatantly or unconsciously.

“It’s kind of a nightmare,” said Dave Maass, surveillance technologies investigations director at the Electronic Frontier Foundation. “Police, who aren’t specialists in AI, and aren’t going to be specialists in recognizing the problems with AI, are going to use these systems to generate language that could affect millions of people in their involvement with the criminal justice system. What could go wrong?” Smith acknowledged there are dangers. “When people talk about bias in AI, it really is: Is this going to exacerbate racism by taking training data that’s going to treat people differently?” he told Forbes. “That was the main risk.”

Smith said Axon is recommending police don’t use the AI to write reports for incidents as serious as a police shooting, where vital information could be missed. “An officer-involved shooting is likely a scenario where it would not be used, and I’d probably advise people against it, just because there’s so much complexity, the stakes are so high.” He said some early customers are only using Draft One for misdemeanors, though others are writing up “more significant incidents,” including use-of-force cases. Axon, however, won’t have control over how individual police departments use the tools.

Police in multiple major US cities have figured out a trick to circumvent their cities’ bans on facial recognition technology. Just ask a friend in a city without any such restrictions to do it for you.

It’s not immediately clear how widespread such side-stepping of facial recognition restrictions in the US may be. According to the Washington Post’s review of police documents in Austin, Texas, and San Francisco, however, it’s definitely something that’s gone on in at least two major US metropolitan areas.

WaPo reported over the weekend that SFPD and APD have both sought help from neighboring cop shops on occasion, though with different levels of success. San Francisco cops have reportedly sought facial recognition searches from other agencies on five occasions since 2019, but never got a match back; Austin police, on the other hand, have sent at least 13 searches since 2020.

“Some” of the searches from APD returned hits, and multiple suspects in Austin have been arrested and charged as a result of the searches, WaPo said.