My current median for AIs fully automating AI R&D is maybe Jan 2031 or perhaps a touch earlier. My views have updated towards towards shorter timelines over 2026. My median for AI R&D parity is around early/mid 2030 (in part due to an update towards thinking we hit parity significantly before “full” automation). By full automation, I mean removing humans from AI R&D would slow things down by <10%.
I discuss something more like my 30th/35th percentile here (roughly my modal scenario).
My current median for AIs fully automating AI R&D is maybe Jan 2031 or perhaps a touch earlier. My views have updated towards towards shorter timelines over 2026. My median for AI R&D parity is around early/mid 2030 (in part due to an update towards thinking we hit parity significantly before “full” automation). By full automation, I mean removing humans from AI R&D would slow things down by <10%.
I discuss something more like my 30th/35th percentile here (roughly my modal scenario).