UK AI Security Institute study finds frontier AI models outperform expert human debaters in persuasion

Is AI more persuasive than the most persuasive humans? 🧐

Proud to be part of this @AISecurityInst x @UniofOxford research, led by the stellar @KobiHackenburg, which provides an impressively thorough answer.

Humans gatekeep many of the processes that keep our systems secure and our societies functioning, which is exactly why this surprising finding matters so much.