Why should we care that AI is suddenly insanely good at math?
It’s a question that I’ve been asking mathematicians and AI researchers for months, and especially this week, when OpenAI published a solution to one of the notoriously complex Millennium Prize Problems.
And it’s the one math question I can answer without breaking into a cold sweat.
Math is many things to many people: beautiful, elegant, nightmare fuel. Over the past year, it has also become proof of artificial intelligence’s rapidly accelerating progress, showing how the technology is getting much smarter, much faster than most people anticipated.
On the day society learned about AI’s solution to a Millennium Prize Problem, I called another Fields medalist to make sense of the latest breakthrough.
“The flickering dream that there’s still a human role in mathematics isn’t completely killed off by this,” Timothy Gowers told me. “But my guess is that if you’ve got a model that’s capable of doing this, it will be capable of doing lots of other things.”
Which led him to an unsettling conclusion.
“I don’t want to say it’s all over,” Gowers said, “but I certainly don’t want to say it’s not all over.”
Whether it’s really over was the agenda when OpenAI hosted mathematicians in its San Francisco offices last month. The summit began with an assumption: AI will become superhuman at math.
Once they accepted that premise, they could reason through all of its practical implications. What does it mean for education, or collaboration, or the publication of ideas, or how we train the next generation? They didn’t come up with any solutions, but that wasn’t the point.
Their goal was selecting the right problems to work on, like the mathematicians who chose the Millennium Prize Problems.
And we should all care. Because this time, humans will have to solve them.