NVDA/long
AI's 1-2% success rate on hard problems is transformative when applied at massive scale, making compute the critical lever for scientific breakthroughs
on any given problem an AI tool has a success rate of maybe 1% or 2%
Peak+36.9%
Terence Tao discusses how AI is transforming mathematics and science: idea generation cost has collapsed to near zero, AI solves hard problems at 1-2% rate but scale makes it transformative, and the verification bottleneck is the new constraint on scientific progress.
originalon any given problem an AI tool has a success rate of maybe 1% or 2%
AI has driven the cost of idea generation down to almost zero