Opinion: Allocate $10M Inference Budget to Let AI Solve Open Math and Cyber Problems
kylekabasares · x · 2026-07-24
AI researcher Peter Gostev argues that the fact that random users can solve open math problems with ChatGPT Pro should be a wake-up call for universities. He highlights a massive gap in inference compute applied to open problems, suggesting the need for a dedicated unit with a $10M inference budget to continuously tackle open questions in math, physics, and cybersecurity. He noted that when AI found numerous vulnerabilities in Firefox, it wasn't just about model capability, but a concerted effort—something AI labs are currently not taking seriously enough.
More from AGI Musings
- Conference slide says Graeme Segal is the mathematician AI will replace last — AlexKontorovich · 2026-07-24
- Charliermarsh: It's Worth Learning Things Useful Today Even If Obsolete Tomorrow — charliermarsh · 2026-07-24
- A new AI-era bottleneck is the physical world, not idea generation — xiao_ted · 2026-07-24
- A model screenshot argues the real problem is turning intelligence into agency — MoonL88537 · 2026-07-24
- AI’s underrated third use case may be contract formalization and negotiation — curious_vii · 2026-07-24
- In Five Years, the Internet May Be Overrun by AI-Targeted Pollution — willdepue · 2026-07-24