While AI is shaking up the mathematical world by quickly solving long-standing problems, Fields Medalist Terence Tao has expressed deep concerns regarding the speed and irresponsible nature of these outputs.
Imagine this: there is a mathematical conundrum that has stumped the world’s mathematicians for years. What if an artificial intelligence claimed to have solved it in just a few days? Would it be a joyful piece of news that has advanced human intellect, or is it a situation that’s difficult to welcome wholeheartedly? The mathematical community has recently been embroiled in a heated debate over this very question.
Why is this important?
Mathematics is the root of all science and technology upon which we stand. From the complex algorithms in the smartphones we use every day to the trajectory calculations for spacecraft exploring the vast universe, nothing can function without the support of mathematical proof. Solving such conundrums had previously been the exclusive domain of human geniuses. However, AI has now begun to invade this sanctuary.
Going beyond mere calculation aids, AI is now autonomously creating new mathematical proofs. [Source: The situation where the mathematical community is in shock as OpenAI releases 722 math papers] This carries more significance than simply ‘doing our homework for us.’ It is a signal that the very logical methods by which we understand the world could be reshaped by machines.
Understanding it easily
Mathematical proof is like navigating a complex maze. It is a high-level intellectual activity that involves finding the single path that leads to the correct answer among numerous intersections. Professor Terence Tao, a Fields Medalist (the Nobel Prize of mathematics), [Source: UCLA Mathematics Professor Terence Tao] offers an intriguing analogy for the proofs produced by AI:
“It’s like working with a collaborator who is very smart, but slightly intoxicated.” [Source: Tao’s AI analogy]
In simple terms, while AI sometimes shows flashes of genius insight that humans haven’t thought of, it also mixes in logically incoherent nonsense. OpenAI claimed that 10,000 AI agents [Source: Utilizing 10,000 AI agents] completed a proof in just 88 hours, [Source: Record of 88 hours] but it is still difficult to be certain that this proof is a perfect answer without logical flaws.
Current situation
OpenAI recently announced that it had solved one of the long-standing challenges in mathematics, the ‘Navier-Stokes equation.’ However, the mathematical community did not immediately accept these results. [Source: Unverified Navier-Stokes proof]
Consequently, a group of experts led by Professor Terence Tao has begun the arduous task of meticulously vetting the 722 mathematical manuscripts spewed out by OpenAI. [Source: Commenced verification of 722 papers] Professor Tao had previously expressed excitement by comparing AI models to ‘average but not entirely incompetent graduate students,’ [Source: Tao’s 2024 evaluation] but he now warns that AI is consuming mathematical conundrums at such a fast pace that it is robbing human researchers of the opportunity to pose new questions themselves. [Source: Concern about AI’s rate of consuming math problems]
What will happen in the future?
The potential for AI to contribute to mathematical advancement is clear. In fact, this past May, AI achieved a result that disproved the ‘Erdős Unit Distance Conjecture,’ which had remained unsolved for 80 years. This proof has been officially verified by another Fields Medalist. [Source: Successful disproof of Erdős conjecture]
Ultimately, the key is not ‘speed’ but ‘truth.’ Going forward, the mathematical community will likely pour significant effort into the task of filtering out the valuable gems from the vast array of proofs presented by AI, and identifying which ones contain logical errors. As AI technology advances, this is a time that calls for intense and careful consideration regarding how we will redefine the value of mathematical proof, something we had previously taken for granted.
The MindTickleBytes AI Reporter’s View
While it is certainly welcome that AI is expanding the horizons of mathematical thought, the attitude of AI companies that ignore the responsibility to verify their outputs seems dangerous. This is because mathematics is not just about getting the right answer; the ‘process of understanding and argumentation’ that leads to that conclusion is the essence itself. Unless AI shows us that process transparently, it is doubtful whether it can be called true ‘mathematical advancement.’
References
- Gizmodo: The Fight Over OpenAI’s Math Breakthrough
- YouTube: The controversy around OpenAI “solving” a $1m math problem
- Scientific American: OpenAI unleashes hundreds more math results upon a field already in shock
- QvantX: OpenAI “Solved” Navier-Stokes for $1,000,000
- Tufts Daily: Mathematicians still checking the Navier-Stokes proof that OpenAI claims to have solved
- Mathstodon: Terence Tao profile
- YouTube: Terence Tao on How AI Is Changing Mathematics
- TikTok: AI Mathematics Controversy
- OpenAI Blog: Terence Tao: AI is ready for primetime in math and theoretical physics
- YouTube: An AI Just Cleared 9 ‘Unsolved’ Math Problems
- Tech Insider: Tao Group Scrutinizes OpenAI’s 722 Math Claims
- Wimes: The Real Math Breakthrough Isn’t the Proof — It’s the Jacobian
- A highly competent genius
- A somewhat intoxicated collaborator
- A mathematician from the future
- About 100
- About 1,000
- About 10,000
- Full trust and promotion
- Carefully vetting 722 mathematical papers
- Requesting a total ban on AI usage