Any progress towards AI at all? was Re: [extropy-chat] Futures Past

Eugen Leitl eugen at leitl.org
Mon Oct 10 06:57:30 UTC 2005


On Mon, Oct 10, 2005 at 01:36:34AM +0100, Russell Wallace wrote:
> On 10/10/05, Dirk Bruere <dirk.bruere at gmail.com> wrote:
> >
> > There are two things needed.
> >  Raw computing power, which is on track to deliver Human performance

Not raw computing power. Integration density, which is something different.
(Potential computing power, if arranged in the right configuration for
the job).

> > equivalent within a decade, and software.
> > IMO the big breakthroughs will come by doing crude brute force simulations
> > of real brains - the data derived from slice-and-scan. First insects, then
> > small mammals, then people.
> >
> 
> Well okay, that's an idea we had long before 1998, though I tend to regard

That's an idea which goes back to the middle of the last century. Probably
before, just thought of Campbell.

> uploading as something separate from AI. It'll be a lot more than a decade
> before we have the computing power for that, though. (By 2015 we might have

I agree. But the point is that you don't need detailed simulations,
just seeding an evolutionary optimization process from good guesses taken
from scanned critters neuroanatomy.

> the computing power to match human performance in a significant spread of

Please show me the benchmark by which you see a human equivalent by 2015.
(Could it be? Somebody is making a specific prediction, again?

> domains _if_ we had appropriate AI software optimized to make efficient use
> of electronic hardware - though I doubt we will have such software by then -
> but _not_ to run an uploaded human brain.)

You can't afford software in human equivalent AI. You have to actually route these
spikes represented as packets, though a high-dimensional network
implemented directly in semiconductor structures, if you want to achieve
a realtime performance. Do the math.

The bits are only there for the state, and how the state influences
other state. And simulating prototypes for bootstrap, of course.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07100, 11.36820            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20051010/3d4442b6/attachment.bin>


More information about the extropy-chat mailing list