Ah, I assumed that by “linear returns” you mean, if we can create an IQ 150 robot for $N whether we can get twice as much “IQ 150 thoughts per second” for $2N… which we trivially can, by building two such robots.
If you meant whether we can get an IQ 300 robot for $2N, yeah, that is a completely different question. Probably no. Maybe “first no, and then yes”, if let’s say at IQ 500 we figure out what intelligence truly is and how to convert resources to intelligence more effectively, rather than just running the same algorithms on stronger hardware.
Ah, I assumed that by “linear returns” you mean, if we can create an IQ 150 robot for $N whether we can get twice as much “IQ 150 thoughts per second” for $2N… which we trivially can, by building two such robots.
If you meant whether we can get an IQ 300 robot for $2N, yeah, that is a completely different question. Probably no. Maybe “first no, and then yes”, if let’s say at IQ 500 we figure out what intelligence truly is and how to convert resources to intelligence more effectively, rather than just running the same algorithms on stronger hardware.