tech
December 23, 2025
How social media encourages the worst of AI boosterism
The era of hype first, think later.
TL;DR
- OpenAI's GPT-5 was initially claimed to have solved 10 unsolved math problems, but it was later revealed to have found 10 existing solutions that were not yet widely known.
- This incident is seen as an example of AI hype, where social media amplifies claims that may be exaggerated or misrepresented.
- While the hype overshadowed GPT-5's ability to find obscure references, LLMs are considered valuable tools for mathematicians to search existing literature.
- Recent studies indicate LLMs have limitations in medicine and law, providing flawed diagnoses and inconsistent advice.
- The rapid pace of AI development and competition fuels a culture of making bold claims on social media, making sober assessments difficult.
- Axiom Math's AI model, AxiomProver, has achieved success in solving open math problems and performing well in the Putnam competition, but the true capabilities of LLMs in mathematics require deeper investigation.
Continue reading
the original article