4 ms·
OpenAI: AI hallucinations are mathematically inevitable, not engineering flaws
- ctxc 1y agoPardon my French, but...no shit?
- brettkromkamp 1y agoThis! Exactly, this! As if this is some kind of novel revelation.
- dietr1ch 1y agoIt's a problem in the underlying model, not an implementation flaw, but still an engineering flaw in my book.
- signa11 1y agothis is imho pretty cool for humans :o) if hallucination is inevitable, then the whole thing cannot function without a-human-in-loop. this ofcourse is in direct confrontation with fundamental premise of adopting such tech to reduce human workforce.