Hey Gordon! My economic model is that individual worker productivity only acts as a ceiling on wages/compensation through impact on labor demand (eg not many businesses can pay someone more than their productive value to that business).
Actual compensation is a dance between labor demand and supply. In practice, compensation appears to increase along with productivity, but that's because higher productivity raises demand. Moving from $1 in productivity to $2 increases the number of businesses who can get a positive return on employing that labor, so demand rises.
But if supply rises exactly to match, compensation wouldn't change. Labor supply is very laggy in software though, because it takes years to train.
Sharing for feedback!
The hard question for me to answer is about the demand for additional software in the world. How much software is currently not economically viable to build because building software is too expensive? I don't even know how to guess at that
But if a lot of software becomes economically viable at 2x productivity, then demand raises and wages could go up. Even at 100x productivity it could happen as long as that 100x productivity required a rare level of skills / capability / effort.
One thing that's different about AI is that it allows us to dramatically increase the supply of programming by giving lots of non-programmers access to programming and giving programmers the ability to do a lot more programming. So this isn't quite a normal change in supply and demand because it's not strictly tied to the supply of workers, and in the near future AI might get even better at programming and need less human oversight.
I do think we'd see a lot more software if it were cheaper to make it. In fact, I think we're already starting to see this, and is similar in nature to, say, the amount of financial forecasting done before vs. after spreadsheets. Actually, spreadsheets are probably one of my primary analogies for AI impacts on labor, and expect AI to do something similar to programmers that spreadsheets did in finance (at least until AI less needs humans to deliver results).
Hey Gordon! My economic model is that individual worker productivity only acts as a ceiling on wages/compensation through impact on labor demand (eg not many businesses can pay someone more than their productive value to that business).
Actual compensation is a dance between labor demand and supply. In practice, compensation appears to increase along with productivity, but that's because higher productivity raises demand. Moving from $1 in productivity to $2 increases the number of businesses who can get a positive return on employing that labor, so demand rises.
But if supply rises exactly to match, compensation wouldn't change. Labor supply is very laggy in software though, because it takes years to train.
Sharing for feedback!
The hard question for me to answer is about the demand for additional software in the world. How much software is currently not economically viable to build because building software is too expensive? I don't even know how to guess at that
But if a lot of software becomes economically viable at 2x productivity, then demand raises and wages could go up. Even at 100x productivity it could happen as long as that 100x productivity required a rare level of skills / capability / effort.
I'm thinking along similar lines (or think I am).
One thing that's different about AI is that it allows us to dramatically increase the supply of programming by giving lots of non-programmers access to programming and giving programmers the ability to do a lot more programming. So this isn't quite a normal change in supply and demand because it's not strictly tied to the supply of workers, and in the near future AI might get even better at programming and need less human oversight.
I do think we'd see a lot more software if it were cheaper to make it. In fact, I think we're already starting to see this, and is similar in nature to, say, the amount of financial forecasting done before vs. after spreadsheets. Actually, spreadsheets are probably one of my primary analogies for AI impacts on labor, and expect AI to do something similar to programmers that spreadsheets did in finance (at least until AI less needs humans to deliver results).
If AI continues to automate and the owners of the business capture all that value, it feels essential to own as much of these companies as possible
Agreed!