Are AIs getting dangerously good at persuasion? OpenAI says “not yet.”
At this point, anyone following artificial intelligence is familiar with the many (often flawed) benchmarks companies use to demonstrate a model's effectiveness at everything from math and logical reasoning to…