Estimates of when AGI might arrive range from a few years to many decades — or never. Understanding why helps you read forecasts critically.
Different Definitions
Forecasts depend heavily on what counts as AGI. A system that can do most remote knowledge work is very different from one matching humans at everything.
Different Models of Progress
- Extrapolating trends in compute, benchmarks and task length suggests fast progress.
- Emphasising unsolved problems — reliability, continual learning, physical skills — suggests longer timelines.
Incentives
Companies raising investment, researchers seeking funding and commentators seeking attention all have reasons to lean one way or another.
Track Record
Historical AI predictions have often been wrong. Some recent capabilities arrived faster than experts expected; others have been slower.
Reading Forecasts
- What definition is used?
- What evidence and assumptions underlie it?
- How confident is it, and does it give a range?
- Who's making it, and why?
A Practical Stance
Rather than betting on a date, prepare for a range of outcomes: plan for continued rapid improvement while keeping realistic expectations of current systems.