For Bill Gates, artificial intelligence is the most important invention since the internet or the personal computer. For Google boss Sundar Pichai, it will have a more profound impact than electricity and fire. Already, though, there are signs the AI revolution may be faltering. Since OpenAI released its landmark GPT-4 system in March 2023, new large language models like Google鈥檚 Gemini have offered only incremental improvements.
GPT-5 could change this tomorrow, of course. But progress can also come from unexpected places, like the esoteric world of pure mathematics.
Advertisement
Although maths provides the foundations on which AI systems are built, professional mathematicians have been sceptical that AI could ever discover new conjectures, say, or construct tricky proofs. That is because research-level maths requires highly sophisticated reasoning and creativity. As we report in our story 鈥淗ow AI mathematicians might finally deliver human-level reasoning鈥�, however, AIs are beginning to pull off some remarkable mathematical feats, matching humans in elite competitions and contributing fresh insights to decades-old conundrums.
There are question marks around the extent to which these systems are approaching human-like reasoning, and whether their combination of a component that generates solutions and another that evaluates them tells us anything new about how to build more generally intelligent AIs.
In any case, computer scientists are excited. Seemingly simple architectures have yielded progress before. Google鈥檚 Transformer model ushered in the current era of AI chatbots, for example. And it is possible that the reasoning and creativity we think of as uniquely human aren鈥檛 as difficult to replicate as we imagine.
Pure maths might seem an unlikely proving ground for generally intelligent AIs. But the logic is that if these systems can tackle the hardest problems in maths, they might also be capable of generalising those abilities. That would herald machines that reason and think like humans, and justify prophecies of a revolutionary technology.
