Hook
Posted Sep 6, 2026Video · 0:36by @secondsightai
We didn’t find a product being promoted in this video. Its hook, format and structure below still show why it took off.
Browse products winning on TikTokThe video leverages the authority of a world-renowned mathematician to provide a counter-intuitive, high-level perspective on AI, making complex technical concepts accessible and intriguing.
Mathematician Terence Tao discusses his unconventional approach to using AI for mathematical proofs. He explains that while AI output is often flawed, iterative looping and verification allow for successful results.
Fields Medalist Terence Tao compares the AI systems now producing research-level proofs to a collaborator who knows a lot and is "slightly drunk." The proofs they generate are sometimes complete rubbish. That isn't his objection, it's the operating condition. Loop the process, check the results, and something with a dismal per-attempt success rate still comes out positive in aggregate. Tao calls this orthogonal to how we normally think about intelligence, which we picture as grounded and methodical, working from first principles. Orthogonal isn't a synonym for inferior. The drunk is right often enough to be worth the supervision. Credits: Video: https://youtu.be/svl_1upFpQo?si=SYJckVvN83T1SQPs Speaker: Terence Tao #ai #artificialintelligence #futuretech #mathematics #intelligentdrunk
The world's best mathematician says AI's proofs are often complete rubbish. He uses them anyway.
Introduction to AI's role in mathematics
Explaining the 'drunk' nature of AI proofs
The necessity of looping and verification
Comparing AI to traditional methodical intelligence
Conclusion on the effectiveness of the method

Playing this video’s own audio, so you’ll hear the creator over the sound.
Outliers and viral hits: their videos on Overshot, all time.