Specifically, this tweet makes the assumption that open-source AI is basically as good at search, especially automated search for algorithms as the big labs, and if that assumption is violated, the prize won’t be claimed even if big companies have much better AI than the general public.
This is short, but IMO this implies it’s measuring open-source’s ability to get good AI, not whether any AI can be very general.
Francois Chollet inadvertently limits his claim on ARC-AGI
Link post
Specifically, this tweet makes the assumption that open-source AI is basically as good at search, especially automated search for algorithms as the big labs, and if that assumption is violated, the prize won’t be claimed even if big companies have much better AI than the general public.
This is short, but IMO this implies it’s measuring open-source’s ability to get good AI, not whether any AI can be very general.