Who Taught the Machines Mathematics?

Commenters debated whether modern AI changes what it means to do mathematics and who gets credit for discovery. Stabbles compared AlphaZero’s non-human training to extractive use of human-created math, arguing labs proved architectures can outperform humans without human game data. Tarsierscripter countered that games like Go differ from open-ended math, and bemtevi (a former pro mathematician) described a split between those who embrace AI-driven "problem miners" and those who value human exploration, pointing to the Association for Human Mathematics as an anti-AI stance. Glimshe asked whether it’s acceptable for people to outsource theorem discovery to AI and only provide human-readable verification; theplumber and others replied that AI is just a tool and verification is a legitimate role. Hypfer and ykonstant reported local communities oscillating between denial and despair about hiring and careers, while unknownian worried elders aren’t protecting future human thinkers and flagged plagiarism and scooping risks. Socializer quipped about AI-written articles.

Opinion divides over values and incentives: some see AI as a neutral, inevitable tool that accelerates work and should be normalized (theplumber, hypfer), while others see it as corrosive to the joy, originality, and community-building that motivate mathematicians (bemtevi, stabbles). Practical disagreements focus on whether safeguards or norms are needed, how tenure and publishing should adapt, and whether open weights versus cloud-hosted models change power dynamics; glimshe and others worried explorers will be outcompeted by AI-augmented miners.