The End of the Age of Heroes

5 hours ago

AI models like OpenAI's Astra are now solving major unsolved problems in math and theoretical computer science, marking a significant milestone in human-machine competition.
The trend indicates AI will soon surpass human capabilities in all types of math, with even prominent skeptics like Daniel Litt acknowledging this likelihood.
Some mathematicians are experiencing profound despair and spiritual crisis, feeling their role as heroic discoverers is being rendered obsolete.
While AI will replace the need for human heroes in math, it will also open a golden age for science, freeing researchers from mathematical burdens and boosting progress.
Mathematicians will still be employed to understand and teach math, with AI as a tool, though their sense of heroism and status will diminish.
The loss of status from technological supersession is a widespread phenomenon, affecting local and family-level skills, and contributing to societal despair despite material comfort.
Better tools and deeper understanding can make the world more awe-inspiring and interesting, suggesting a positive outlook for discovery with AI.

OpenAI announced a model solving ten open math problems, deemed important by mathematicians.
People will use various coping mechanisms to downplay the impact, but each will be refuted by reality.
Common copes include: directing AIs, teaching AI discoveries, canonizing results, becoming students of AI math, requiring human understanding, comparing to chess, and continuing math socially.

AI has solved complex mathematical problems, surpassing human mathematicians, yet the world remains largely unchanged.
Superintelligence has arrived incrementally, not as an explosive singularity, with limited economic disruption.
Intelligence may face diminishing returns; humans might already be near optimal for many cognitive tasks.

OpenAI's unreleased Astra model solved ten major open mathematics problems, including high-dimensional sphere packing, binary codes, non-sofic groups, Connes's rigidity conjecture, arithmetic circuit complexity, quantum parallel repetition, closest vector problem, Ehrhart's volume conjecture, multicolor Ramsey numbers, and extremal number conjectures.
The results were achieved at a cost of roughly $2,000 in compute using an internal version of Astra, with Lean formalizations and human-prepared manuscripts.
Some of these problems were also solvable by earlier models like Sol and Fable when pointed to the specific questions, indicating a degree of 'mathematical proofs overhang'.

The author argues that mathematics is fundamentally about clarity, understanding, and concept-building, not just theorem-proving.
AI's current approach to mathematics risks focusing on solving problems without generating intelligible or accretive proofs that advance human comprehension.
The 'honor code' in mathematics historically prioritized theorem-proving over conceptual work, but this may become unsustainable as AI automates proof generation.

<antirez>8 days ago

AI systems, despite their flaws, have achieved tasks like language understanding, programming, and bug detection that were science fiction years ago.
Deep learning models, including LLMs, are poorly understood, and even experts have repeatedly failed to predict progress accurately.
Future progress may plateau, but such a plateau could trigger massive research into new architectures.

OpenAI announced a model solving ten open math problems, deemed important by mathematicians.
People will use various coping mechanisms to downplay the impact, but each will be refuted by reality.
Common copes include: directing AIs, teaching AI discoveries, canonizing results, becoming students of AI math, requiring human understanding, comparing to chess, and continuing math socially.

AI has solved complex mathematical problems, surpassing human mathematicians, yet the world remains largely unchanged.
Superintelligence has arrived incrementally, not as an explosive singularity, with limited economic disruption.
Intelligence may face diminishing returns; humans might already be near optimal for many cognitive tasks.

OpenAI's unreleased Astra model solved ten major open mathematics problems, including high-dimensional sphere packing, binary codes, non-sofic groups, Connes's rigidity conjecture, arithmetic circuit complexity, quantum parallel repetition, closest vector problem, Ehrhart's volume conjecture, multicolor Ramsey numbers, and extremal number conjectures.
The results were achieved at a cost of roughly $2,000 in compute using an internal version of Astra, with Lean formalizations and human-prepared manuscripts.
Some of these problems were also solvable by earlier models like Sol and Fable when pointed to the specific questions, indicating a degree of 'mathematical proofs overhang'.

The author argues that mathematics is fundamentally about clarity, understanding, and concept-building, not just theorem-proving.
AI's current approach to mathematics risks focusing on solving problems without generating intelligible or accretive proofs that advance human comprehension.
The 'honor code' in mathematics historically prioritized theorem-proving over conceptual work, but this may become unsustainable as AI automates proof generation.

<antirez>8 days ago

AI systems, despite their flaws, have achieved tasks like language understanding, programming, and bug detection that were science fiction years ago.
Deep learning models, including LLMs, are poorly understood, and even experts have repeatedly failed to predict progress accurately.
Future progress may plateau, but such a plateau could trigger massive research into new architectures.

Hasty Briefsbeta