Will it be a golden age or a dark age for mathematics (and science)?
What's new ● Covered by 2 sources
The guest author argues that large language models can generate mathematically checked Lean proofs, but the loss of human understanding could steer mathematics and science toward a “dark age” despite short-term increases in output. The author points to OpenAI’s “Tuesday October 6, 2026, Drop,” which reported trying 4000 problems and succeeding on only hundreds after aggregation and selection. The piece concludes that progress depends on choices about how LLMs are integrated—requiring continued human expertise, skepticism, and attention to mathematization and how mathematical progress is evaluated.
Why it matters
The issue frames a question about whether large language models will usher in a “golden age” or a “dark age” for mathematics and science. It begins by noting that large language models have…