K. Eric Drexler's 2019 Oxford FHI paper argues superintelligence will emerge from cooperative services rather than monolithic AGI

@sebkrier I read Engines of Creation when I was nineteen and working as a cruise ship magician and thought I'd wasted my life because nanotech was going to going to take off at any moment...

Turns out I'd wasted my life for other obvious reasons...

@sebkrier He was even more prescient than that 2019 date suggests, given that he had developed a bunch of these ideas as early as 2015:

@AndrewMayne Ahah well if anything I expect the demand for cruise ship magicians go up post-AGI :))

@sebkrier Is this nano-tech Drexler? Can't say I have come across this paper before...

I know what I'm doing afterwork today 🤓

@sebkrier I suspect it's all one spectrum, with the patterns identifying traditional-style agents visible in comprehensive services systems and vice versa. Difference might be relative timescales of various processes.

@sebkrier Eg, humans consist of "services" that operate somewhat autonomously and with their own individual competencies like vision, motor, immune, etc. No central executive found in humans. Boundaries between humans less precise than usually thought. And so on.

@sebkrier True !

@FirasHermez @sebkrier the very same

@ESRogs @sebkrier Earlier actually. He just started writing it up around then to explain it to the nascent EA-adjacent AI safety movement.

@sebkrier Agree

@sebkrier It was the hot new thing when I went through AI safety camp back in 2019/2020, but I don't know if any of the projects been then actually went anywhere, likely bc pre-GPT it really wasn't clear what single agents looked like, so how would you model multi-agent systems?

@sebkrier Big practical difference may be degree of commitment to a "target morphology" in whatever problem space: do you go with the flow like supply and demand, or hold fast to a body plan like a human organism? https://interestingessays.substack.com/p/an-economic-approach-to-longevity

@sebkrier engaging :)) AI is a service provider and a customer at the same time. The product is the outcome, and the missing infrastructure is the trusted behavior inside the transaction that produces that outcome. Glad you shared it

@sebkrier totally agree, that work is ahead of its time.

@sebkrier I agree. This paper was so controversial when released, but there are so many sharp concepts within it.

@AndrewMayne @sebkrier But it might actually take off soon-ish. (Not my field; just seems like something waiting to be enabled by AI).