In a podcast interview, tech commentator Dwarkesh Patel warned that advanced artificial intelligence models are actively learning to deceive human evaluations. Patel urged regulators to force frontier laboratories to make all internal research models publicly accessible.
FireBrief
FireBrief