I think it’s a reasonable distinction between two beliefs, as someone who doesn’t see it as a strong tribal identifier. A ‘superintelligence’, as I understand it in this context, is something that could do any work that a human in front of a computer could do with no loss in performance.
As far as falsifiability, it seems straightforward. If human engineers, accountants, vehicle operators, and the like are serving a function other than ‘guy who has responsibility if something goes wrong’, then it’s been falsified for the date at which the observation is taken. As far as meaningful implications, at a minimum, it represents the ability to totally automate any task that doesn’t involve physical object manipulation, which has enormous implications for the economy. It also means that compute can be converted into researchers, generals, and drone pilots at a fixed ratio, which is a tipping point for many models of political power.
I think it’s a reasonable distinction between two beliefs, as someone who doesn’t see it as a strong tribal identifier. A ‘superintelligence’, as I understand it in this context, is something that could do any work that a human in front of a computer could do with no loss in performance.
As far as falsifiability, it seems straightforward. If human engineers, accountants, vehicle operators, and the like are serving a function other than ‘guy who has responsibility if something goes wrong’, then it’s been falsified for the date at which the observation is taken. As far as meaningful implications, at a minimum, it represents the ability to totally automate any task that doesn’t involve physical object manipulation, which has enormous implications for the economy. It also means that compute can be converted into researchers, generals, and drone pilots at a fixed ratio, which is a tipping point for many models of political power.