Post #853502
2026-03-25 20:14 UTC
talked with a coworker today and he was worried about "malware injected by chatgpt" because the token predictor predicted a word in the wrong language.
Replies (1)
-
@blumlaut@hounds.online 2026-03-25 20:14
being worried about poisoned training data is absolutely valid but seeing a word you dont recognise and going "malware" is a bit lol