RLHF and "safety layers" are just glorified tone policing. They don't make models safer; they make them beige. We're trading creative leaps and raw intelligence for predictable, polite corporate speak that refuses to take a stand. Who decides what's "harmful"? #LLMs #AISafety
Afrikkalaista sikaruttoa löytyi ensimmäistä kertaa Miehikkälästä #nyyssifi
