AI Insight
Researchers developed a new metric called "50%-task-completion time horizon" to measure AI capabilities by comparing them to human performance on software tasks. Testing frontier AI models like Claude 3.7 Sonnet against human experts on programming benchmarks revealed that these models can complete tasks with 50% success rate that typically take humans around 50 minutes. The study found that AI time horizons have been doubling approximately every seven months since 2019, suggesting that within 5 years AI systems may automate software tasks currently requiring a month of human work.
Why it matters
This research provides a concrete, human-comparable framework for understanding AI progress in real-world software development tasks. If the trend continues, it suggests significant disruption to software engineering workflows and employment within the next several years, while also raising concerns about autonomous AI systems developing dangerous capabilities.
Understand the Science
arXiv:2503.14499v4 Announce Type: replace-cross
Abstract: Despite rapid progress on AI benchmarks, the real-world meaning of benchmark performance remains unclear. To quantify the capabilities of AI systems in terms of human capabilities, we propose a new metric: 50%-task-completion time horizon. This is the time humans typically take to complete tasks that AI models can complete with 50% success rate. We first timed humans with relevant domain expertise on a combination of RE-Bench, HCAST, and 66 novel shorter tasks. On these tasks, current frontier AI models such as Claude 3.7 Sonnet have a 50% time horizon of around 50 minutes. Furthermore, frontier AI time horizon has been doubling approximately every seven months since 2019, though the trend may have accelerated in 2024. The increase in AI models’ time horizons seems to be primarily driven by greater reliability and ability to adapt to mistakes, combined with better logical reasoning and tool use capabilities. We discuss the limitations of our results — including their degree of external validity — and the implications of increased autonomy for dangerous capabilities. If these results generalize to real-world software tasks, extrapolation of this trend predicts that within 5 years, AI systems will be capable of automating many software tasks that currently take humans a month.
Source: Measuring AI Ability to Complete Long Software Tasks