[ExI] Computer power needed for AGI [WAS Re: Hard Takeoff-money]

[Samantha Atkins]

On Nov 16, 2010, at 9:41 AM, Richard Loosemore wrote:

> Samantha Atkins wrote:
>>>> But wait.  The first AGIs will likely be ridiculously expensive.
> 
>> Keith Henson wrote:
>>> Why?  The programming might be until someone has a conceptual breakthrough.  But the most powerful super computers in the world
>>> are _less_ powerful than large numbers of distributed PCs.  see http://en.wikipedia.org/wiki/FLOPS
>> Because: a) it is not known or much expected AGI will run on
>> conventional computers; b) a back of envelop calculation of
>> equivalent processing power to the human brain puts that much
>> capacity, at great cost, a decade out and two decades or more out
>> before it is easily affordable at human competitive rates; c) we have
>> not much idea of the software needed even given the computational
>> capacity.
> 
> Not THIS argument again!  :-)
> 
> If, as you say, "we do not have much idea of the software needed" for an AGI, how is it that you can say "the first AGIs will likely be ridiculously expensive"....?!

Because of (b) of course.  The brute force approach, brain emulation or at least as much processing power as step one, is very expensive and will be for some time to come.

> 
> After saying that, you do a back of the envelope calculation that assumes we need the same parallel computing capacity as the human brain..... a pointless calculation, since you claim not to know how you would go about building an AGI, no?
> 

Not entirely as human beings are one existence proof of general intelligence.  So looking at their apparent processing power as a possible precondition is not unreasonable.  This has been proposed by many including many active AGI researchers.  So why are you arguing with it?

> Those of us actually working on the problem -- actually trying to build functioning, safe AGI systems -- who have developed some reasonably detailed architectures on which calculations can be made, might deliver a completely different estimate.  In my case, I have done such estimates in the past, and the required HARDWARE capacity comes out at roughly the hardware capacity of a late 1980s-era supercomputer...

Great.  When can I get an early alpha to fire up on my laptop?

This is a pretty extravagant claim you are making so it requires some evidence to be taken too seriously.  But if you do have that where your estimates are reasonably robust then your fame is assured.

- samantha