Re: Complexity of AGI

From: Michael Roy Ames (
Date: Sun May 19 2002 - 23:10:50 MDT

Eliezer S. Yudkowsky wrote:
> I'm not sure I agree with this.

Which part don't you agree with? The difficulty of correctly connecting
new hardware to an uploaded mind, or the need to understand exactly how the
brain's mental processes work in order to make the connections?

> Create a virtual environment, raytrace it, and simulate the impinging
> photons on the rods-and-cones layer
> of the retina. Seems relatively straightforward by comparison with
> simulating the rest of the brain.

The estimate of difficulty would seem to depend on where the simulation ends
and reality begins in the target hardware. If we were going to simulate the
brain, plus the entire nervous system, plus all the senses up to the last
layer of cells (without understanding how the brain's mental processes
work)... then, yes I can see how that would be a straightforward
implementation. Otherwise, not straightforward at all.

Create a virtual environment. That sounds easy when you say it fast... I
wonder how difficult it really is? I fear it may be very hard indeed to
create a totally convicing reality simulation... much easier to simply
interface with reality directly. I suspect the same thing will be true for
AI's - that interfacing directly with reality will be less difficult that
creating virtual environment. What say the rest of SL4?

Michael Roy Ames

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT