Mathematicians Need to Act

3 hours ago

The author argues that mathematicians should not collaborate with AI companies for the foreseeable future, citing recent AI achievements like proving the existence of nonsofic groups for only $2,000.
AI threatens core mathematical values such as proper attribution, accountability for errors, and open-access knowledge sharing, as LLMs often misattribute work and produce nonsense proofs.
The use of LLMs in mathematics undermines human understanding, especially in education where students rely on AI to complete homework without learning, and in research where it raises the bar for early-career mathematicians.
The author believes that if AI and humans compete directly in theorem proving, AI will surpass humans within 2-5 years, making human-led mathematical activity obsolete unless structural protections are implemented.
The essay calls for mathematicians to refuse work at AI companies, reject gifts and partnerships, and avoid using LLMs to prove new theorems, instead advocating for collective decision-making and political action to preserve the mathematical community.

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.

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.

The author has minimal firsthand experience with AI and eschews its use, but feels compelled to write a blog post to defend against false accusations of writing 'AI slop' based on stylistic choices like em dashes.
AI is criticized for promoting ignorance and laziness by replacing the need to learn or struggle, with the goal of making humans 'super-dumb' rather than AI becoming super-intelligent.
AI's resource consumption exacerbates climate change, causes hardware shortages, and raises electricity costs, while vendors engage in intellectual property theft and regulatory capture.

The author observes widespread AI psychosis in companies, where rational discussions about AI's effectiveness are impossible.
Most AI projects are failing, with 0% success observed in over a year, despite claims of massive productivity gains.
Companies are often bad at software projects, and AI adds extra risk; internal and customer-facing chatbots frequently fail due to poor data or lack of use.

Richard Feynman's warning not to fool yourself applies especially to AI users, who may feel more productive without evidence.
No independent scientific study confirms that AI improves productivity; one study found developers were 19% slower using AI.
Personal anecdotes and layoffs attributed to AI are unreliable—many layoffs are 'AI washing' or cost-cutting for unproven technology.

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.

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.

The author has minimal firsthand experience with AI and eschews its use, but feels compelled to write a blog post to defend against false accusations of writing 'AI slop' based on stylistic choices like em dashes.
AI is criticized for promoting ignorance and laziness by replacing the need to learn or struggle, with the goal of making humans 'super-dumb' rather than AI becoming super-intelligent.
AI's resource consumption exacerbates climate change, causes hardware shortages, and raises electricity costs, while vendors engage in intellectual property theft and regulatory capture.

The author observes widespread AI psychosis in companies, where rational discussions about AI's effectiveness are impossible.
Most AI projects are failing, with 0% success observed in over a year, despite claims of massive productivity gains.
Companies are often bad at software projects, and AI adds extra risk; internal and customer-facing chatbots frequently fail due to poor data or lack of use.

Richard Feynman's warning not to fool yourself applies especially to AI users, who may feel more productive without evidence.
No independent scientific study confirms that AI improves productivity; one study found developers were 19% slower using AI.
Personal anecdotes and layoffs attributed to AI are unreliable—many layoffs are 'AI washing' or cost-cutting for unproven technology.

Hasty Briefsbeta