Keep pulling the thread on Grant Sanderson.
AI has effectively "cold-solved" geometry problems within the International Math Olympiad (IMO) competition.
If an AI were to solve a problem like the Riemann Hypothesis by creating a fundamentally new mathematical theory, this level of intelligence would likely have broad impacts across the economy.
Grant Sanderson predicts that the most useful progress from AI models in mathematics in the next 5 years will be their ability to act as "supercharged connectors," finding and mapping connections between different fields.
Grant Sanderson suggests a key advantage of AI in research is the ability to systematically reset an agent's context or spawn multiple agents with different contexts, allowing them to escape cognitive ruts that humans struggle with.
Grant Sanderson suggests a major future benefit of formal systems like Lean is the ability to create an AI that endlessly explores and extends a formal mathematics library like Mathlib without human supervision.
Grant Sanderson predicts that teaching will be one of the most stable careers over the next 50 years, even in a post-AGI world, because it is fundamentally relational and involves mentorship.
Grant Sanderson previously held the view that an AI achieving a gold medal in the International Math Olympiad (IMO) would represent another benchmark being passed, rather than the achievement of Artificial General Intelligence (AGI).
An AI system solved the geometry problems on the 2024 International Math Olympiad (IMO) in approximately 19 seconds.
An AI system would have achieved a gold medal in the 2024 International Math Olympiad (IMO) if the competition had not featured two combinatorics problems, a category where the AI struggles.
A formula describing the statistical correlation between pairs of zeros of the Riemann zeta function, found by Hugh Montgomery, is identical to an expression Freeman Dyson recognized from the study of eigenvalues for random Hermitian matrices.
Grant Sanderson believes it is difficult to create a formal benchmark for an AI's ability to generate good conjectures or definitions because their quality is subjective and cannot be easily scored.
Grant Sanderson has observed a significant positive shift in how mathematicians discuss the utility of AI in their work between mid-2025 and 2026.