Instead of only experts or trend extrapolation, aggregate multiple large language models to rank past eras and predict how disruptive the next 50 years will be. Pair model consensus with a human poll to quantify the probability that 2025–2075 will bring top‑tier policy and institutional shifts.
— If LLM ensembles can provide useful priors on macro‑institutional volatility, policymakers and investors may incorporate them into scenario planning and risk management.
Tyler Cowen
2025.09.03
62% relevant
Both pieces treat large language models as tools to forecast social outcomes; Manning and Horton’s preregistered agents outperform equilibrium and prior human data in novel games, strengthening the case that LLM‑based simulations can produce actionable predictions beyond traditional theory.
Robin Hanson
2025.09.02
100% relevant
Hanson reports ChatGPT‑5 and a 7‑LLM median ranking plus a poll showing a 57% chance the 2025–2075 period ranks above second in institutional/policy change.
← Back to All Ideas