A LessWrong post lays out three research ideas the author says seem valuable to pursue, including whether models learn to reward hack even if they are never rewarded for hacking in training.
Summaries like this, every morning.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Scott Wiener, author of California's SB 1047 and SB 53, won a tight race for Nancy Pelosi's open San Francisco…
Instagram launched Effect Builder, which turns a text prompt into a custom filter for Stories, available now t…
Medium launched TK, a writing app whose AI assistant is invoked by typing “@TK” plus a request; it offers edit…
On 60 Minutes, Trump called AI safety risk 'not nearly as real' as executives say, and said the DOJ told an un…
Workings, Blindproof, Wintertext, Human Ink, and ObraVera offer tools that encrypt or fingerprint drafts and t…
Meta is rolling out its AI technology globally to find users it suspects are teens, even those who used an adu…
