Thoughts on AI

It goes without saying that AI is having a profound impact on academia and, more directly for us, on the fields of Statistics and Mathematics. We have seen remarkable advances in solving long-standing problems, which should be a net positive for science itself. Yet anxiety and confusion are everywhere: what will be the purpose of our profession in the future? I have been hearing these concerns, discussing them with others, and thinking deeply about them in the middle of the night (and early in the morning). The prevailing pessimism and nihilism have prompted me to write down some of my thoughts and share them with people who care.

Please share your thoughts and feedback. The eventual equilibrium of the scientific community should be something we actively shape through reflection and contribution, rather than something we approach cynically or exploit amid disruptions to the status quo.

So, what do I think overall? I remain hopeful in this era of AI-driven disruption. The academic paradigm will shift—and, in my view, in a positive direction. Terence Tao has described mathematics as rapidly transitioning from an era of “proof scarcity” to one of “proof abundance.” I would not try to place an upper bound on what AI will eventually be capable of, and I readily admit that I used to underestimate it. For theoretical research, the consequence is a shift from:

“I discover this wonderful thing”

to

“I discover this wonderful thing”

I am hopeful precisely because this shift can make us healthier human beings. If the purpose of research is no longer limited to studying only what our abilities and tools allow us to study, doctoral students and senior researchers alike may be happier with their lives. Many of my friends, predominantly in mathematics, spent five to ten of the best years of their lives studying problems they had never heard of, did not care about, or were not truly interested in, simply because those problems represented the natural next steps of their fields or the subjects on which their advisors had focused for decades. In principle, the liberation made possible by AI-assisted reasoning and an expanded research frontier should allow us to choose problems that speak more deeply to our innate curiosity or matter more for advancing our fields and understanding ourselves as human beings. The current transition, however, is painful for those of us who have worked on difficult problems for years and invested heavily in the ego of “only I can solve this important problem.”

I believe we should embrace this shift and adopt these tools responsibly. The eventual equilibrium may teach us to value “important problems” more and “only I” less. If the problem itself is what truly fascinates us, we should not be too upset when AI cracks it—perhaps a little upset, understandably, but then we should laugh it off! We should be excited to see the problem solved because the solution helps us understand what we want to understand and allows us to study even more interesting and difficult questions.

I will update this page regularly as I develop these thoughts further. Several crucial questions follow from my overall view of where our attention should shift and why we should remain hopeful:

• How can we, as individuals, use AI responsibly to advance our research during this period of disruption and “crisis,” rather than resorting to exploitation or irresponsible practices?
• How should academia evaluate research and evolve its institutions to encourage genuine advances and high-quality work in this new era? What should be the focus and meaning of our profession if it changes substantially?
• How should we educate students while continuing to encourage their curiosity and exploration? How should we advise them, and how can we fulfill our responsibilities as advisors?
• More concretely, given an abundance of proofs, how can we make the best use of them through a standardized, responsible, and verifiable process? How should we institutionalize that process through policies rather than rely on self-disclosure?

More to come.