Seems like "peak oil" is happening, but not at all like we originally thought.
My main worry in 2010 was that it is very difficult to balance an exponential increase in demand for a thing that is a big employer (such as oil) with some convenient replacement that employs a similar number of people at a similar cost.
The economics of the transition to PV and wind are much smoother than I expected, because I am not an economist and did not have a remotely realistic model for how any of this would work in practice. (Hopefully my similar concern about the upcoming economic transition due to AI automation is misplaced for exactly the same reasons).
How to deal with that is a political problem, not particularly an economic one.
Even narrow AI is more of the same — one that can drive, would rapidly put millions of drivers out of work, but not much else. That's a huge one-off transition to manage, but not inherently radical.
General-purpose AI? Depends on the details. Let's say it's got the all the capabilities and limitations of an exactly average human, with a 24/7 equivalent power requirement of 17.2 kW — that will cost around $15.1k/year to run: https://www.wolframalpha.com/input?i=17.2+kW+*+1+year+*+%240...
I picked that power requirement to get an annual cost comparable to the USA federal minimum wage: https://www.wolframalpha.com/input?i=%247.25+per+hour+*+40+h...
This means it's economically neutral to pick an average human or a machine for those roles and with that assumption about electricity cost… provided that humans are willing to accept minimum wages for that role.
This lowers the wages for the average human to minimum wages.
But! The USA's electricity supply is about 1.3 TW: https://www.publicpower.org/resource/americas-electricity-ge....
This is only 3.8 kW/capita: https://www.wolframalpha.com/input?i=1.3TW%2Fusa+population+...
This suggests that prices may rise rapidly, at least before the AI can be put to work guiding robots to install solar power systems (there's a startup on Y Combinator dedicated to exactly this).
How much might prices rise? To whatever level where the AI can still be a neutral replacement for human labour. If the median capacity human earns $45k, then that assumption for power requirement would raise electricity prices to $0.30/kWh. If the AI is as energy efficient as a human brain, and uses 30 watts, then it doesn't have any impact on prices because the US doesn't run out of power in the first place; if the AI needs megawatts, then the effect is also reduced, because 1 MW * 1 y * $0.1/kWh = $876,000 and therefore it isn't economically sensible to displace even one worker.
So what I'm worried about is the transition, the time between now and the steady-state long-term future where the robots can put in as much PV as they need. Do we have a risky period where robots could use all the available electrical power and might be able to out-bid the average human on both sides of the equation, the income side by being as cheap as it's legally possible for a human to be, and the expenses because now the human isn't earning any money and can't pay for electricity?
Or does this dynamically fix itself because I'm doing an over-simplified toy model in my head and I've missed a critical part? (Probably).
I guess of you really think robots are gonna bid up electricity, residential solar makes even more sense.
I may be simply wrong, but I can assure you it's definitely not a hand-wave.
You may think it absurd that an economic choice would be made over a human value, but that kind of behaviour is exactly what happened in the Irish potato famine — there it was exporting food while starving to death.
I am also planning to get domestic PV as soon as possible precisely because I anticipate this dynamic.
Again, I'm aware I might simply be wrong, but "wrong" and "hand wave" are different things.
> What's driving the infinite demand if humans don't have wages to buy things with?
1) I didn't say infinite demand
2) I did say "out-bid the average human on both sides of the equation".
"Average" human.
Humans are not identical, but rather are a distribution.
The ones who are more economically viable can continue to compete while the rest starve.
This is also what we saw in the Irish potato famine, though "economically viable" had a different meaning.