Yeah by one-time-gain i don’t mean to imply it would literally all happen at once—i think spread out over years.
But i do think there’s a cap there (holding technology levels fixed) based on only having so many humans to direct.
Perhaps innovations in company organisation will take years for superintelligence to discover and will allow massive gains to output though, like >100X. That would be a way in which “one time gain” is more misleading than helpful and your idea of continued productivity growth more fitting
That’s fair, I think I might’ve just gotten the wrong impression from the graph. Personally, I wouldn’t think there would be a hard cap, as the IE would naturally boost technology levels instead of holding them fixed. However, I do agree that, either way, there will eventually come a point where a generalized machine laborer will be more efficient and more productive than a human laborer.
I just read your ‘Three Types’ essay, and I thought it was also really good! Particularly interesting to me was the idea of how, as the IE cascades towards the full-stack IE, the concentration of power becomes more decentralized. I’ve been working on a model to anticipate the geopolitical and social impacts of AI development (please check it out!), and I hadn’t previously considered how the IE itself could have centralizing/decentralizing vectors.
Great work! I’ll definitely be keeping an eye out for more stuff from you guys.
Yep agreed you might need energy expansion for the IE. You might be interested in our post on this: https://www.forethought.org/research/three-types-of-intelligence-explosion
Yeah by one-time-gain i don’t mean to imply it would literally all happen at once—i think spread out over years.
But i do think there’s a cap there (holding technology levels fixed) based on only having so many humans to direct.
Perhaps innovations in company organisation will take years for superintelligence to discover and will allow massive gains to output though, like >100X. That would be a way in which “one time gain” is more misleading than helpful and your idea of continued productivity growth more fitting
That’s fair, I think I might’ve just gotten the wrong impression from the graph. Personally, I wouldn’t think there would be a hard cap, as the IE would naturally boost technology levels instead of holding them fixed. However, I do agree that, either way, there will eventually come a point where a generalized machine laborer will be more efficient and more productive than a human laborer.
I just read your ‘Three Types’ essay, and I thought it was also really good! Particularly interesting to me was the idea of how, as the IE cascades towards the full-stack IE, the concentration of power becomes more decentralized. I’ve been working on a model to anticipate the geopolitical and social impacts of AI development (please check it out!), and I hadn’t previously considered how the IE itself could have centralizing/decentralizing vectors.
Great work! I’ll definitely be keeping an eye out for more stuff from you guys.