Will Miles Brundage win his bet with Gary Marcus?
Miles Brundage, former head of policy research at OpenAI, and Gary Marcus, NYU neuroscientist, have made a public bet about AI progress (details on Gary Marcus's Substack) that we will not see AGI by the end of 2027. They're operationalizing that as 8 out of the following 10 things being feasible in principle by AI: Understand everything about a new movie that a human does, just by watching it. Same but for novels. Write hallucination-free biographies/obituaries of New York Times quality. Master almost any new video game as fast as a human can. Write hallucination-free human-level legal briefs. Write 10k original bug-free lines of code from a spec or dialog with a nontechnical user. Write Pulitzer-caliber fiction and non-fiction books. Write Oscar-caliber screenplays. Make paradigm-shifting, Nobel-caliber scientific discoveries. For any math proof expressible in current Lean, convert it to lean from natural language. This market resolves however Gary Marcus and Miles Brundage resolve their wager. Related Links https://manifold.markets/LovkushAgarwal/possible-outcomes-of-the-bet-on-ai (version of this market with each of the 10 tasks predicted separately) https://www.metaculus.com/questions/31246/2027-ai-bet-winner-between-gary-marcus-and-miles-brundage/ (Metaculus's version of this market) https://milesbrundage.substack.com/p/times-up-for-ai-policy (article that prompted Gary Marcus to challenge Miles Brundage to a wager) https://garymarcus.substack.com/p/superhuman-agi-is-not-nigh (older post laying out Gary Marcus's AGI timelines) https://x.com/Miles_Brundage/status/1870642568121462791 (original hashing out of the wager) https://x.com/GaryMarcus/status/1873766399618785646 (summary of the wager) https://garymarcus.substack.com/p/where-will-ai-be-at-the-end-of-2027 (official specification of the wager) https://thezvi.substack.com/p/ai-97-4?open=false#%C2%A7wanna-bet
