More than 20 academics signed an open letter published Tuesday pushing for a “reorientation” of the field to focus on artificial intelligence’s potential to worsen problems like inequality and crime, as well as potential risks like humans losing control over autonomous systems leading to extinction. Signatories include the University of Toronto’s Geoffrey Hinton and Gillian Hadfield; Mila’s Yoshua Bengio; the University of British Columbia’s Jeff Clune; and celebrity scholars Yuval Noah Harari and Daniel Kahneman. (The Logic)

Talking point: AI could do plenty of good, the letter says, but the academics are concerned the risks are receiving too little attention. They’re calling for firms and other AI financers to spend at least a third of their budgets on safety and ethics; deep learning pioneers Bengio and Hinton have previously argued the ratio should be as high as one to one. The letter also calls on regulators to keep track of who’s using compute capacity to develop models, and require their makers to register them. Governments should license “exceptionally capable” systems, and make firms liable for any preventable harms their products cause. Similar proposals have been the subject of debate in the sector.