Spencer Greenberg No, I don't assume robot supply is fixed. Why I...
· Comment — Archer T. Ships replied to his own comment. · View on Facebook · Markdown source
Spencer Greenberg No, I don't assume robot supply is fixed. Why I would I do that? What I do assume is that like all previous technical advances, robots will have some cost of construction, fuel supply, etc. Those costs are going to be high at first. For example, the currently most competent LLM's have massive power requirements. Many of the major investors in AI firms (Sam Altman, Bill Gates, etc) have also invested in a nuclear power firms for this reason. It takes 4+ years to deploy a nuclear power plant under the best circumstances, and it can take much longer. The US didn't start _any_ new nuclear power plants for 35 years, and the most recent new plant took 13+ years and had huge cost overruns. So, even under the most optimistic assumptions, it's going to be at least 4 years, just to build the power plants required for widespread deployment of human level AI. There are many other problems as well: * No humanoid robots are currently in widespread production. It will likely take years for them to make it to market in an affordable manner. * I'm skeptical that AI's will escape the uncanny valley in the next 10 years. As a result, there are many human facing jobs that humans will still perform better. * Chip fab plants required to build AI's take years + billions to build. Yes, AI's will be able to accelerate progress, but there are physical limits to what can be done. It takes time to mine rare earth metals, to place concrete and rebar, to construct high tolerance chip fab factories. I think it's highly implausible that AI's will wholly replace humans in the interim. In fact, I think the most plausible rollout timeline is on the scale of decades, during which time humans and AI's will be working in tandem, and we'll have a better sense of the risks AI's actually pose and potential remedies. (As opposed to ahistorical "foom!" scenarios. You also seem to think that humans will remain static. But it seems to me that humans will augment themselves. First with drugs (steroids, GLP-1 antagonists, modafinil), then with machines (artificial kidneys, titanium hearts, eyes, neurolinks), then with full-on brain simulations / robot bodies. What does it matter if AI's take over, if humans become competitive AI/human hybrids themselves? Also, biological humans can only operate well in a narrow temperature/pressure/gravity band. AI's will be able to operate in a much broader set of environments, from the deep ocean to deep space. IMO, there will relatively brief window (a few decades) during which AI's and humans will compete for similar resources on earth. After that, AI's will rapidly expand into space, the solar system, then the galaxy. I think AI's at that point will regard biological humans the same way we regard great apes today: sympathetic primitive creatures whose habitat we try preserve for historical / nostalgic reasons. It seems quite plausible to me that AI's will set aside planet sized nature preserves for biohumans can live in luxury if they want.