Mastodon Feed
Boosted by jonny@neuromatch.social ("jonny (nonvenomous)"):
david_chisnall@infosec.exchange ("David Chisnall (*Now with 50% more sarcasm!*)") wrote:
I am not good at spotting LLM-generated output. The last time I was on a grant review, there were a bunch of proposals that I dig into and they didn’t seem to make sense. When we came to discuss them, the other reviewers said ‘I think this is AI’, ‘I thought so too’, ‘okay, skip this one’. There were a lot of similarities between the ones that got flagged like this because they all included the grant call in the prompt and so tied things into it in the same weird nonsensical ways.