The greatest shortcoming of the human race is our inability to understand the exponential function.
Albert A. Bartlett
A new research note from METR, a nonprofit that evaluates frontier AI systems, argues that at current rates of progress there will be more than 99% automation of AI research and development by 2032, only six years away.
The model is very simple, with only eight parameters. It makes deliberately conservative assumptions: no superexponential capability growth, no "full" automation that exceeds 100%, no modeling of the additional acceleration that would come from AI systems developing research intuition. And yet, the median output is near-total automation of the work that produces more capable AI, within this decade.