A secret meeting in 2025 brought together some of the world’s leading mathematicians to test OpenAI’s latest major language model, o4-mini.
The experts at the meeting were amazed at how much the model’s answers sounded like those of a real mathematician when it provided a complex proof.
“I have never seen this type of reasoning in models before,” Ken OnoProfessor of number theory at the University of Virginia said at the time. “That’s what a scientist does.”
But that was it artificial intelligence Is the (AI) model getting more credit than it deserves? And are we at risk of accepting AI-derived evidence without fully understanding it?
Ono acknowledged that the model may provide compelling – but possibly wrong – answers.
“If you say something with enough authority, people just get scared,” Ono said. “I think o4-mini has mastered proof by intimidation; it says everything with so much confidence.”
Confidence and the appearance of a good argument used to be good signs…
