[haswell]

Language is just a variable resolution abstraction on top of non-linguistic thinking and information processing.

I think it’s a mistake to equate the language ability of these models with general intelligence, even if the language produced is excellent.

I think some people see AGI that way (some nebulous unfalsifiable thing), but that’s not what I’m arguing. I think there’s a strong case to be made that at a minimum, AGI’s core “knowledge” will need to be made up of far more than just a lossy textual representation of the world.