Mathematician Joel David Hamkins Says AI Is Not Ready for Mathematical Research

Joel David Hamkins, a well-known mathematician and a logic professor at the University of Notre Dame, is convinced that existing AI is not a good tool for mathematical research. According to his statement, existing LLMs contain a lot of wrong answers, with poor logic.

During a recent conversation on the Lex Fridman podcast, Hamkins explained that he had tried several expensive and advanced AI models, but the experience was disappointing. He said, "I used them a lot, but they didn't seem to be helpful at all for mathematical work."

According to Hamkins, the biggest problem with AI isn't that it's wrong, but that it doesn't admit its mistakes. Even when they point out obvious errors in the AI's reasoning, the model simply says, "Everything is fine" and persists. Hamkins compared this to humans, saying that if a human behaves this way, one wouldn't want to talk to them again. According to him, this attitude makes it difficult to trust AI.

It is a fact that today, some people say that AI has solved tough mathematical problems. However, there are also mathematicians who do not agree with that statement. In fact, Terence Tao, the Nobel-winning mathematician, also has concerns that the proofs created by AI might look correct but might also have serious errors that no mathematician would ever validate.

Hamkins clearly states that performing well on tests or benchmarks does not mean that AI will be equally useful in real research. In his words, "AI is not yet trustworthy when it comes to mathematical reasoning." Although he believes AI may improve in the future, at the moment it is far from becoming a true 'research partner' for a professional mathematician.