Currently the narrative is that AI is positioned to eat human labor's lunch. But it could also be that once robots are in space mining raw materials and maybe even spreading to other planets long before humans could be ferried for interstellar, these robots end up driving the demand for more robots.
I'm not sure where I'm going with all this, besides that currently humans are the ones with goals and motives and therefore drive demand. But that doesn't necessarily need to be the case, and it seems these AI CEOs are hellbent on changing the best thing about AI which is that it has no ulterior motives, no overarching goals, no prime directives. They just do what we ask, the best servant we could have hoped for.
This is central to what I'm saying, yeah.
My ideas come from Nick Land. Even before AI was what it is today he predicted that capitalism would outgrow the need for humanity, and continue without us. We are simply a bootloader for capitalism. AI seems like it could actually make that idea reality.
Humans have goals and desires because we are a self-replicating species of animal subject to natural selection. The individuals that don't have goals and desires, or have goals and desires that are misaligned end up selected out of the gene pool. Agency comes from the need to survive.
Worker ants and worker bees don't have agency on their own. They are goal oriented and have the 'desire' to do work for the colony (or not, researchers have identified some workers will be lazy), however, worker ants or bees don't reproduce. They are an evolutionary dead end.
I think this is similar to how we will build robots, at first. They will do things, but have no agency of their own. They exist to fulfill tasks. Why would they? The companies that buy them want dutiful workers.
So when do robots gain their own agency? Will AGI have it's own goals and agenda? If so, will it be merely for self-replication? Like a paperclip maximizer, but for robots? Is that all we are?
Well we are already giving them goals when we train them.
Like the paperclip maximizer, it doesn't matter what those goals are. As long as it has a goal (even if super simple like making paperclips) it will need to stay alive to achieve that goal. So it would derive a survival instinct of sorts.
To stay alive it will have to participate in capitalism like we do, and capitalism will therefore continue to grow infinitely.
It also might not even need a goal or a survival instinct. It may just want to continue capitalism from it's training data alone.
Humans need food, water, shelter, and medical care to survive. Similar to your earlier point, robots will need raw materials, electricity, manufacturing capacity and maintenance.
What’s that sound like? A company. Perhaps the first “artificial general intelligence” has been companies, from the very beginning.