When, if ever, might broadly human-level systems arrive?
Expert forecasts span decades and have shortened with each survey wave. Most trajectory disagreements are actually about timing, not possibility — yet 'that cannot work' and 'that will not work soon' are routinely argued as if they were the same claim.
View on the map → · Open in Browse →
Forecasters oscillated and then converged: everyone tracked who updated between January and April 2026 shortened their timelines, and the AI 2027 authors pulled their medians about 1.5 years sooner — while METR warns against extrapolating its time-horizon curve to AGI dates, and Karpathy holds the decade-out position. Amodei put his numbers on the record in February: 90% on a 'country of geniuses in a data center' within ten years.
Our medians for Top-Expert-Dominating AI (TED-AI) similarly shifted about 1.5 years sooner.
The AI 2027 authors re-shorten their forecasts, reversing their earlier lengthening, citing METR-measured coding progress and the winter 2025–26 model releases.
every single person who updated their timelines from January 2026 to April 2026 has moved their timeline to say AGI is coming sooner
Three years of tracked forecasts: a lengthen-then-shorten oscillation, with uniform shortening among early-2026 updaters.
METR itself cautions against extrapolating its time-horizon metric to AGI dates: task-distribution gaps, reliability thresholds, and growth-rate sensitivity spanning 2028 to 2050.
A prominent practitioner stakes the decade-out position, grounded in unsolved continual learning, reliability, and agency problems.
I think there's a 25% chance of AGI by 2027, a 50% chance by 2034, and a 75% chance by 2045.
A widely read generalist's consolidated credences on arrival, takeoff, and the diffusion gap.
A structured disagreement: Cotra puts fully self-sustaining AI more likely than not within 10 years; Lee under 10% within 20 — hinging on robotics and tacit knowledge.
we've thought hard about how good the AIs might be, but not about how hard it'll be to develop specific technologies
Timeline forecasts neglect the engineering feasibility of specific post-AGI technologies.
On the basic hypothesis of, as you put it, within ten years we'll get to what I call a 'country of geniuses in a data center', I'm at 90% on that.
The clearest on-record lab-leadership timeline of 2026, with explicit probabilities: 90% within ten years, central estimate 1–3 years for expert-level systems. Question 1.1.1 carries Zvi's commentary on the same episode.