[paytonjjones]

The low level foundations of intelligence, both human and machine, are just not well understood in general. It's totally plausible that different architectures could be capable of superintelligence at far fewer parameters than current LLMs. We just don't know.

All we have to go on is the empirical trajectory, and I can see reasonable arguments either way.

[morpheos137]

this assumes general intelligence as an abstract principle vs an adapation to specific constraints is even a thing. we also dont have a low level understanding of how many angels can dance on the head of a pin.