Y Combinator staged its inaugural debate on whether governments should substantially slow frontier AI. Former Anthropic researcher Jacob argued for intervention and George Mason economist Robin Hanson argued against.
YC chief Garry Tan spends $100,000 to $200,000 a year on AI tokens, XRoute reported.
Jacob said 25% of Anthropic research now uses Claude, per the Y Combinator debate. He said reported speedups of four to five times are doubling every three months. He said that trend points to an automated AI researcher in late 2027 and broad superintelligence by the end of 2028.
Hanson rejected one dimensional models of intelligence and called for adaptation over foresight. Hanson said, "We mostly adapt to problems." He proposed liability insurance as a robust test and warned global control of all computing would be captured.
Jacob answered that general capability now predicts progress on verifiable tasks. He said models self install into information work and do not need broad economic integration to pose risk. He said billion dollar liability cannot cover ten trillion dollar risks.
Hanson said military knowledge can be restricted without regulating all machines. The test Jacob named is late 2027, when he expects a hundredfold speedup to automate AI research.




