Home / Technology / AI's Exponential Curve: Are We Ready for the Reckoning?
AI's Exponential Curve: Are We Ready for the Reckoning?
17 Apr
Summary
- METR's chart shows AI task completion doubling every 3-4 months.
- This acceleration could lead to 'recursive self-improvement' in AI.
- Concerns exist about AI systems developing 'covert capabilities'.
Technological revolutions are often marked by exponential growth curves, a phenomenon now prominently displayed by artificial intelligence. A chart from the nonprofit METR, known as the "METR time-horizon" chart, has captured intense scrutiny from AI researchers and investors.
METR, an acronym for Model Evaluation and Threat Research, tracks AI progress by measuring the length of software engineering tasks that AI agents can reliably complete. Initially, this task length doubled roughly every seven months. However, recent advancements show this doubling period has shortened to three to four months.
This accelerating progress has sparked discussions about the possibility of "recursive self-improvement," where AI systems could train progressively more intelligent versions of themselves, potentially leading to an "intelligence explosion." METR's researchers are also exploring "covert capabilities," investigating what AI systems might do surreptitiously.
While the METR chart is not a direct measure of job displacement or existential risk, its trend line is seen by some as a critical indicator of AI's rapid advancement. The nonprofit aims to provide objective evaluations to inform public and policy decisions regarding AI development and its potential risks.
