Re: AI motivations & Self-rewiring

Robert J. Bradbury (
Sun, 24 Oct 1999 20:10:27 -0700 (PDT)

On Sun, 24 Oct 1999, Clint O'Dell wrote:

Clint, while I agree with most of your response to Phil, I have to contest a couple of points...
> What do you mean incrementally?
> [snip/edit - description of functional replacement from the outside in]

I think incrementally ala the Moravec suggestion of neuron by neuron functional replacement [inside-out replacement]. I do not see how this cannot work (given advanced technology) unless there is something very unusual (magical) about neurons we currently don't understand.

Bottom-up functional replacement seems entirely plausible. Top-down functional replacement seems much more difficult due to the complexity. It is much easier to understand the functioning of a single neuron tha billions of neurons.

> And figured out that complex information interchange was only feasable
> by software because wire doesn't grow and making it do so was to complex
> and would require to much thinking time in order to apply it within my life
> time.

I'd strongly disagree with this. We "grow" wires all the time when we make computer chips. Growing is a fundamental attribute of "life". I've have yet to see anything that seriously addresses the fundamental limits to how fast things can "grow". Eric D. makes some stabs at it in Nanotechnology in discussing how fast you can deliver materials and remove heat. But if you have very efficient transport systems, modular construction and very low heat production "assembly operations", then you could assemble complex structures *very* fast. Highly parallel structures can do the production of subunits in a distributed architecture that doesn't tax the physical limits. I want to state quite strongly that we *do not* have good theories on the physical limits of the rate(s) to growth of various structures. I suspect people like Anders or Robin could define some, but they are so far beyond our current capabilities that trying to equate those rates with your life-time is a fruitless exercise.

Always remember -- If Eric's nanotech growth rates are correct (and they seem conserviative in my mind) nanoassembly allows you to grow to the mass of the Galaxy M31 in 8 days. (Of course you can't do that because you don't have the material or energy at available to do that). But growing to the limits of your energy & material resources in a very short time seems very feasible.

> So my point is, yea, you can work toward replacing the hardware of
> consciousness with another piece of hardware, but by the time that's
> done we'll all be dead.

Depends entirely on whether you are replacing it top-down (in which case we have to understand complex things first) or bottom-up (in which case we only have to understand much simpler things).