For the posts in my "LLMs don't actually know anything" thread, I do my best to verify them. These are four that I'm pretty sure used to work but no longer do. That suggests to me that there's some poor team at Google whose job it is to keep track of when people are mocking their "AI" and suppress the bad results.

I think that is hilarious. I would love to know how many individual fixes they've accumulated like this. It reminds me of a competent secretary endlessly covering for an incompetent boss.

This entry was edited (1 month ago)