AI Might Take Over the World... of Idiots

A discussion of a recent Grok transcription of a Jew speaking Hebrew:

Let me be very clear: That Grok post is a lie. The only true words in that “translation” are “the man speaks”, nothing beyond that. He speaks Hebrew not Yiddish; he says nothing about Zionists, shekels or any other money, or buying anything let alone the world; he says nothing bad about non-Jews, he doesn’t use the word “goy” in any way, and does not curse anyone. He doesn’t use words phonetically similar to any of that. There is simply no way that algorithmic transcription and translation yielded this result.

'And yet, Grok was firm about repeating this claim when challenged, even providing a (completely fake) Yiddish transcription, and elaborations on what the young man supposedly said. In one reply to me Grok insisted the video showed the man holding money. Usually it’s easy to bully an LLM into agreeing with you against its previous answer, but not this time.

Lots of people trust AI. I am sure Goebbels would have been pleased. There are things that AI can do, but trusting it for anything that you cannot independently verify is akin to doing a Google search for Trump is a felon and trusting the first item on on the list.