In which I ask meta.ai a few questions about myself and the world. No real *new* news here: The answers are plausible but wrong: https://www.tbray.org/ongoing/When/202x/2024/04/18/Meta-AI-oh-my
Decent entertainment value; still waiting for progress on modeling meaning.
@timbray I got my 5 year old a “my first book about computers” from the library. It has a lot about how computers are hyper literal and need everything broken into small steps of if/then and loops. LLMs don’t work like a computer. It’s a real problem. A computer shouldn’t half remember who Tim Bray is. “Hmm yeah, definitely heard that name before. Isn’t he the guy that invented JSON?”
"The answers are plausible but wrong": the official slogan of LLMs.
@timbray I asked ChatGPT about myself when it came out and it said I work at Google which was funny as I never did (I did try to interview once but bailed out of the process as it was too hard!)