Leopold Aschenbrenner - Footnote 34
The claim of a 100,000x increase in effective compute between 2024 and 2027 is not explicitly stated in the source. It is implied through the discussion of orders of magnitude (OOMs) of progress per year, but not directly quantified as 100,000x. The claim that AI systems will surpass college graduates on many benchmarks by 2025-26 is not explicitly stated in the source. The source mentions that by 2027, models will be able to do the work of an AI researcher/engineer, which is a related but distinct claim.
Our claim
entire recordNo record data available.
Source evidence
1 src · 1 checkNoteThe claim of a 100,000x increase in effective compute between 2024 and 2027 is not explicitly stated in the source. It is implied through the discussion of orders of magnitude (OOMs) of progress per year, but not directly quantified as 100,000x. The claim that AI systems will surpass college graduates on many benchmarks by 2025-26 is not explicitly stated in the source. The source mentions that by 2027, models will be able to do the work of an AI researcher/engineer, which is a related but distinct claim.