Geoffrey Hinton: AI Will Outstrip the World's Best Mathematicians Within a Decade — Hassabis Agrees

Fortune·medium signal

Hinton argues that in closed systems like mathematics, AI can generate its own conjectures, test them, learn from failures, and compound endlessly — the AlphaGo loop applied to reasoning — and that language models are on the same trajectory. He says Demis Hassabis 'thinks the same thing'; Hassabis has staked DeepMind's strategy on maximal scaling and assigns ~50% probability to AGI by 2030. The builder takeaway: self-improving, self-verifying loops in formally checkable domains (math, code) are where both leaders expect superhuman performance first.