2026-05-07 19:16 UTC
@ai6yr@m.ai6yr.org the worst is when someone wants to know why an LLM said this thing or THAT and *they ask the LLM why*
(See: too many articles about Grok)
Replies (1)
-
@n1vux@mastodon.radio 2026-05-07 20:12
@nev@flipping.rocks @ai6yr@m.ai6yr.org My ruler for (un)ethical AI (aside from the plagiarism 🐘 ) is whether there is an _actual_ explanation or trace capability. [Where "actual" means "not just another hallucination", but rather, where in the training data set did this line of generated text / piece of image / ... come from and why was it selected? Knowledge-engineered Expert Systems can do this, Data-Science built DSS usually can, but ML/LLM/PISS systems generally can't.]