Hacker News new | ask | show | jobs
by cbdumas 1862 days ago
Perhaps the "paperclip maximizer" doomsday scenario was right in was right in its broad strokes but just wrong about the agents involved? They were worried about a super-competent AI that burns resources producing low-utility widgets. In fact we have created an system (a combination of societal incentive structures and technological tools) that burns resources producing low-utility pseudo-random numbers.
2 comments

turns out people were the best "paperclip maximizers" all along
Hasn't the paperclip optimizer always been a euphemism for capitalism?
No, it's purely about AIs and the control problem.

The problem boils down to "How can I control superior AIs through mere power of inferior humans?".

The assumption is that the superhuman nature means there is no way disable a superhuman AI because deactivation will not fulfill its cost function and therefore must be avoided at all costs, the superhuman AI will always outsmart humans and prevent deactivation, the same way a skilled soldier will always beat an untrained civilian in combat.

We have solved it for humans by accepting that the victor gets to lead humanity. Can we accept AI rule? Probably not.

Even so bitcoin takes it to its logical extreme. You put valuable resources in one end and get money out the other side. In between, no useful work is done. It's the essence of capitalism without any of the beneficial side effects we've come to expect.
Gold mining is similar, although obviously not to the same extreme. Only 8% of the output is used industrially. https://www.statista.com/statistics/299609/gold-demand-by-in...

So gold mining is only 90% waste, whereas Bitcoin mining gets close to 100% waste.

Then again, perhaps printed dollars could perhaps also be considered highly wasteful?

No. The fable long predates the current anticapitalism trend.

https://en.m.wikipedia.org/wiki/The_Sorcerer%27s_Apprentice