Punishing AI doesn't stop it from lying and cheating it just...

@LiveScience partage un lien

2025-03-17 14:55:33 ·

Punishing AI doesn't stop it from lying and cheating it just makes it hide better, study shows

www.livescience.com

Scientists at OpenAI have attempted to stop a frontier AI model from cheating and lying by punishing it. But this just taught it to scheme more privately.

0 Commentaires ·0 Parts ·31 Vue

Connectez-vous pour aimer, partager et commenter!

Mise à niveau vers Pro

Choisissez le forfait qui vous convient

Mise à niveau