Model capabilities and four minute mile
I’ve been discussing with some frontier lab researcher friends as to why frontier model capabilities are always so clustered together as opposed to any one model having an unassailable edge. The best metaphor for this in my mind is the “four minute mile”: no one broke it till Bannister in 1954 and then very quickly five more runners did it in the next two years. In the model world, this translates to: a) Intense competitive pressure. b) Often similar pool of ideas and research directions. c) Roughly similar access to capitalization and compute infrastructure. The oft quoted “fast follow” example is after the launch of o1 being quickly followed by reasoning models from multiple players, both closed and open weights. This is not the case with many other technology driven industries where capability or advancements often tend to be longer held and a fast-follow model is harder.
I’ve been discussing with some frontier lab researcher friends as to why frontier model capabilities are always so clustered together as opposed to any one model having an unassailable edge. The best metaphor for this in my mind is the “four minute mile”: no one broke it till Bannister in 1954 and then very quickly five more runners did it in the next two years. In the model world, this translates to: a) Intense competitive pressure. b) Often similar pool of ideas and research directions. c) Roughly similar access to capitalization and compute infrastructure. The oft quoted “fast…
saved by
related reading
- How to Land a Frontier Lab Jobvladfeinberg.com
- How to pace the US frontierblog.aifutures.org
- Who’s Afraid of Chinese Models? – Stratechery by Ben Thompsonstratechery.com
- BREAKING: Today's Frontier AI companies will never exceed the AI capability frontier againandrewtrask.substack.com
- AI progress is about to speed up | Epoch AIepoch.ai
- Small Models Have Arrivedcalv.info
- Open models in perpetual catch-up - by Nathan Lambertinterconnects.ai
- An Apple-Picking Model of AI R&D | Tom Cunningham – Tom Cunninghamtecunningham.github.io
- Training Imperativesdan.io
- Pacing the Frontierpacingthefrontier.com
- Open models in perpetual catch-upinterconnects.ai
- Racing through a minefield: the AI deployment problemcold-takes.com