Ittai Abraham:AI 或将让数学发现的速度超过人类理解
a16z 把四位研究者拉到一起聊数学家的处境,Ittai Abraham 的观察挺有意思:AI 能给 Lean 证明但给不出好解释,因为“验证证明”容易,“验证解释好坏”没法做成奖励信号。对关心形式化验证和 AI 数学研究的人都值得看。
a16z Frontier 圆桌中,研究者 Ittai Abraham 提出,AI 已经能解开千年数学问题并给出 Lean 形式化证明,却无法给出人类可理解的解释。他认为原因在于强化学习任务容易定义——判断一个证明正确与否是二元的,而评估一个解释是否优于另一个解释缺乏可验证的奖励信号,且写作可读内容需要对读者心智模型的把握。
他预测短期内会出现双轨状态:AI 高速推进“发现前沿”,人类数学家则继续承担“理解前沿”的工作,把 AI 的产出转化为直觉、教材和可传播的知识。圆桌还讨论了 Lean 如何自动校验证明、十亿页的证明是否算证明、年轻数学家如何看待这一转变,以及数学教授现在该教什么等议题。
Ittai Abraham says AI could push the discovery frontier of mathematics ahead of human understanding: "It's kind of surprising that AI can solve a Millennium problem and provide a Lean proof, but cannot provide a human understandable explanation." "Writing something that is human understandable requires you to have a mental model... what the reader understands, what are their intuitions, how it learns." "It is much easier to give a reinforcement task where you have a proof and you have to say, 'Is that a correct proof or an incorrect proof?' It is much harder to show an explanation of, let's say, a proof or a lemma and ask, 'Is this a good explanation or is this better than the previous explanation?'" "My short-term prediction is that this might take a while where we'll have AI be very, very good at pushing the discovery frontier, but we'll still have humans do a lot of work on kind of pushing the human understanding frontier." "Now, obviously, AI is going to help that task. But a lot of the work will have to actually be done by humans." @ittaia
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Lean, formal verification, AI math, reinforcement learning.
Source attribution: curated from a16z (@a16z).