From Joshua Rothman for The New Yorker: “People say it’s just glorified autocomplete,” Geoffrey Hinton told me. “Now, let’s analyze that. Suppose you want to be really good at predicting the next word. If you want to be really good, you have to understand what’s being said. That’s the only way. So by training something to be really good at predicting the next word, you’re actually forcing it to understand. Yes, it’s ‘autocomplete’—but you didn’t think through what it means to have a really good autocomplete.” Hinton thinks that “large language models,” such as GPT, which powers OpenAI’s chatbots, can comprehend the meanings of words and ideas. And that they are either close to or are already able to reason in the same kind of way that human brains do. And that could be dangerous.”
The rise and fall of the bank robbery capital of the world
From Peter Houlahan: “Less than an hour later, the man the FBI called The Yankee walked out of Imperial Bank in Westwood, practically in the shadow of the Federal Building that houses the FBI’s L.A. Bank Robbery Squad, with $4,190. Diving into rush hour traffic on the 405 Freeway, The Yankee headed over the hills to the San Fernando Valley and pulled a final job just before closing time at the First Interstate in Encino for a take of $2,413. Four hours, six heists, $13,197. As impressive as The Yankee’s performance had been, a record for bank licks by one person in a single day, It did not entirely shock the FBI agents in bank robbery squad. This was L.A. after all, and by 1983, L.A. had long established itself as the undisputed “Bank Robbery Capital of the World.”
Note: This is a version of my personal newsletter, which I send out via Ghost, the open-source publishing platform. You can see other issues and sign up here.
Continue reading “Why the godfather of artificial intelligence fears his creation”