Arnold Kling's Nov. 9 roundup brought a 15% U.S. output-per-person wager into a broader debate about AI's effects on growth, government staffing, AI benchmarking and future investment.
The wager asks whether real GDP per capita in the United States can post a one-year rise of at least 15% from its prior peak by the end of 2033. Kling read the odds as four to one against the target. Under Kling's description of the terms, Benjamin Moll would owe the affirmative side $80,000 if it wins; a win for Moll's side would pay him $20,000. Will MacAskill and Dwarkesh Patel are on the affirmative side, while Robin Hanson supports the opposite outcome.
Jason Furman said the Council of Economic Advisers (CEA) could function with roughly half its current staff and a more senior team. He proposed selecting staff to a greater degree than in the past for interpersonal skills and the ability to manage AI agents. Furman expects the council's optimal headcount to decline as AI advances and government operations adjust.
A SemiAnalysis post by Max Kan, Andrew Megalaa and Dylan Patel urged benchmark designers to test general agentic abilities in areas where frontier models may have the biggest effects. The authors said adoption differs by domain and some fields do not need the most advanced models.
Gad Levanon forecast a second boom in spending to build what AI capacity makes possible, saying improving AI will feed directly into digital products such as tutors and agents and physical products such as robots and driverless cars and trucks. Levanon said expectations of broad AI use have already increased infrastructure investment and stock-market wealth. He sees average annual GDP growth of 3% to 4% through 2031 if the economy avoids recession, above current projections. He also said recession looked less likely to him than it had one month earlier.
