Re: Complexity of AGI

From: Eliezer S. Yudkowsky (
Date: Sun May 19 2002 - 18:36:52 MDT

Michael Roy Ames wrote:
> I have not seen anyone address this issue substantively... that is scary.
> *I* would like to upload eventually, and thus avoid death, but I want to
> upload into a device that provides at least the same quality sensory I/O as
> I have in my current body. The recreation of sensory modalities in a
> non-biological substrate seems to be a solvable engineering problem, even
> with today's level of scientific knowledge. But connecting that new
> hardware to the uploaded mind in a way that will make sense to that mind,
> would seem like a much more difficult problem. To do *that* right, would
> require the human brain be simulated in advance of upload, and that the I/O
> channels for all the senses be mapped from the wetware to the hardware. In
> order to do this, you are going to have to understand exactly how the
> brain's mental processes work. There is no short-cut available here.
> Without this detailed mapping, the process of uploading will be horrific
> beyond anyone's worst nightmare.

I'm not sure I agree with this. The brain's visual brainware is the single
best-understood supersystem in the brain. The closer a system is to the
beginning of visual processing, the better we understand it. There are very
detailed computational models of the retina and the rods-and-cones layer is
the most straightforward part of the retina. 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.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT