The Progress of AI is Officially
According to analysis of METR data, the progress of AI is accelerating at a super-exponential rate, with autonomous work capacity expected to increase from 1 hour in 2023 to 10,000 hours by 2030.
Why it matters
This analysis suggests that the pace of AI progress is accelerating at an unprecedented, super-exponential rate, with major implications for industries and society as a whole.
Key Points
- 1METR data shows AI progress is not just exponential, but super-exponential with accelerating doubling rates
- 2Autonomous work capacity is projected to reach 4.13 hours in 2026, 23.5 hours in 2027, 155 hours in 2028, 1,171 hours in 2029, and 10,234 hours (426 days) in 2030
- 3This super-exponential growth is described as
- 4 and exceeding previous conservative forecasts about AI progress
Details
The analysis of METR data shows that the progress of AI is not just exponential, but is actually accelerating at a super-exponential rate. The implied doubling time of AI capabilities has fallen from around 265 days in 2021 to just 113 days by 2030, with the proportional growth rate rising from 0.95/yr in 2021 to 1.53/yr in 2025. This super-exponential trend means that autonomous work capacity is projected to increase from crossing the 1 hour threshold this year, to 10 hours in 2026, 100 hours in 2027, 1,000 hours in 2029, and an astounding 10,000 hours (426 days) by 2030. The author notes that this pattern of AI progress exceeding conservative forecasts is well-established, and that Jensen Huang has already described it as
No comments yet
Be the first to comment