AI is already transforming fields like graphic design, movie making, and computer programming, but it might soon upturn the hard sciences as well.
David Silver, a prominent researcher at Google Deepmind and the lead researcher at AlphaGo, has offered some compelling insights into the future of mathematics and the role of artificial intelligence. He hints at a future where AI mathematicians not only solve some of the world’s most challenging mathematical puzzles but fundamentally reshape the field itself.
“The Clay Mathematics Institute, in the year 2000, offered a million-dollar prize for seven different mathematical problems,” Silver was asked in an interview. “Human mathematicians have had a quarter of a century to try and solve them, and only one has fallen. Do you think potentially the next one could go to AI?”
“Yes, I do,” Silver replied. “I think that it might take time. I don’t think we’re there yet. I think there’s a long way before AI systems are capable of doing this. AI is on the right track, and systems like AlphaProof will become stronger and stronger and stronger.”
He continued, referencing the impressive results that AI had achieved at the International Mathematical Olympiad (IMO): “What we saw in the IMO is just the beginning. Once you have a system that can scale and can keep learning and learning and learning, really the sky’s the limit. What will these systems look like in two years or five years or 20 years? I personally would be amazed if AI mathematicians don’t transform the whole of mathematics. I think it’s coming.”
Silver said that mathematics was a field which was ideal for being worked on with AI systems. “Mathematics is one of the few areas where, in principle, everything can be done completely digitally by machines interacting with itself and just going and going and going. So there’s really no fundamental barrier to experience-driven AI systems mastering mathematics,” he said.
Mathematics, incidentally, wasn’t one of the fields that initial LLMs were particularly good at. Early LLMs like GPT-3 often struggled with basic arithmetic, and famously couldn’t count the number of ‘r’s in the word ‘strawberry’. Even today, the best LLMs struggled to multiply extremely large numbers. Some papers have speculated that LLMs don’t yet have a ‘mathematical model’, and only do things like simple addition through a complex process of going through previous mathematical calculations they’ve seen in their training data.
But LLMs are getting better at math all the time. GPT o3 had scored extremely well on the AIME, which is a feeder exam to the International Math Olympiad. An OpenAI executive has even predicted that an AI will win a gold medal at the IMO next year. ““So GPT 4 (scores) at essentially 0%, maybe, maybe 5%, maybe 10% (on the AIME),” Bubeck said. “o1-preview is already at 50%. You can see (the progress). It’s going to be at 90 percent next month in two months,” he had said.
Google Deepmind’s David Silver too seems to think that AI will revolutionize the field of mathematics. This could have massive implications, not just in mathematics, but all fields downstream of it, like physics. Mathematics is said to be the purest science, and if AI gets good at figuring it out, it can cause ripple effects that could accelerate all scientific progress in the years to come.
