From: Keith Henson (email@example.com)
Date: Fri Feb 24 2006 - 07:49:15 MST
At 11:01 PM 2/23/2006 -0800, you wrote:
>I still don't see how biophysics at all relates to development of a seed
>AI's world modelling or incorporation of goal systems. An adult human
>brain is all hardware while an AGI is slated to run using software.
Unless you have figured out some way to process information without
material, an AGI has to run on some kind of hardware. While any computer
can simulate any other computer, the cost of doing so is high. I.e., an
AGI that solves problems much slower than we do has limited uses. It may
be that AGI will turn out to partly or even mostly be a hardware problem
>And a human baby cannot recursively self improve.
It takes outside help (consider badly neglected or extremely rare wild
raised children) but most people would consider the progress between a
newborn and a PhD self improvement. How is your see AI going to acquire its
knowledge of the world?
And true, the process gets limited by the hardware, but don't you think an
AGI is going to be limited by hardware too? If not I have this spiffy
Apple II I want to run your AGI software on. (I know, a seed AI is
expected to improve its hardware too.)
It also might be useful to consider recursive self improvement as a
separate problem from AGI. I would be impressed enough by an intelligence
as smart as an average human if you could bring that off. Insisting on
both smarts and recursive self improvement and friendly in the same package
is a Xanadu. (Xanadu was trying to do so much at once that events--the
Web--ran over them.)
>This is the wrong place to look for seed AI size estimations. Wheat can't
Humans can, or at least we think they think. (All too often I have my
However, if you don't think a DNA estimate has merit, propose another.
>Keith Henson <firstname.lastname@example.org> wrote:
>At 08:22 AM 2/23/2006 -0500, Richard wrote:
> >Ditto for the amount of information in the DNA that triggers the
> >development of a brain.
>I don't see (prior to birth anyway) where more information comes from.
>If you don't have other information input, then a zygote is a NI (natural
>intelligence) seed. I can't think of an AI researcher in the world who
>would not be *delighted* with an A! I that had the potential of a baby.
>If you look at it in this light, then the information content of a seed AI
>with human level potential should not be more than the information content
>of the human genome and is probably a good deal less.
>It just a question of bits, of course the question is *which bits.*
>Relax. Yahoo! Mail
>scanning helps detect nasty viruses!
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT