A thread through the archive
Essays tagged “AI forecasting”. Follow an idea across the archive.
Scott examines forecasts about a plague death at a Russian lab, finding that top human and AI forecasters all assign very low probabilities (under 1%) to a significant outbreak, demonstrating the value of technical forecasting over mainstream reporting.
Scott analyzes whether AI forecasters will plateau near human levels or keep improving, arguing that even non-miraculous progress could yield 4-12 percentage point improvements in prediction market accuracy.
Scott examines the rise of AI superforecasters that now match or exceed top human forecasters, explores how they work and their current performance, and analyzes implications for decision-making, prediction markets, and the future role of AI opinions.
Scott reviews the state of prediction markets after explosive growth, finding that most volume is degenerate sports gambling rather than useful forecasting, and proposes both technical solutions and two potential futures for the field.
Announcement of an AMA session with the AI Futures Project team about AI, forecasting, and alignment.
Scott announces his collaboration with AI Futures Project's blog and their upcoming AMA, highlighting recent posts including one about AI time horizons that was validated by new OpenAI data.
Scott introduces a new AI forecasting project predicting rapid AI development and potential superintelligence by 2028, led by Daniel Kokotajlo, whose previous 2021 predictions proved remarkably accurate.
Scott examines a new AI forecaster, discusses Polymarket's success, and reviews recent developments in prediction markets and forecasting.
Scott Alexander reviews competing methodologies for predicting AI timelines, focusing on Ajeya Cotra's biological anchors approach and Eliezer Yudkowsky's critique.
Scott Alexander examines GPT-3's capabilities, improvements over GPT-2, and potential implications for AI development through scaling.
Scott analyzes a new survey of AI researchers, showing diverse opinions on AI timelines and risks, with many acknowledging potential dangers but few prioritizing safety research.