Re: Hard takeoff date projections

From: Martin Striz (
Date: Mon Jan 24 2005 - 15:17:32 MST

--- Dani Eder <> wrote:

> I agree with the first comment and disagree with the
> second. We know to some extent how the human brain
> functions, at least at the level of neurons and
> synapses. A sufficiently accurate simulation of
> 10^11 neurons and 10^14-10^15 synapses should
> produce a human level intelligence by brute force.
> Clever AI software design may require less than
> this, but I claim that it is an upper limit.
> The exact computation required to simulate a neuron
> sufficiently accurately is not known exactly, but
> we can put some reasonable estimates to it. I use
> 1 synapse firing = 1 bit +- factor of 30, which
> leads to a human equivalent = 3,000 Tflops (range
> 100-100,000 TFlops).

You're getting into a discussion that I've been wanting to get into for some
time now. I've been wanting to write a thorough critique of Bostrom's and
Moravec's estimations of the computational capacity of the human brain, coming
from a neuroscience perspective. At the present time, and at least until the
end of May, I won't have the time, but I can summarize by suggesting that most
estimations assume that the entire brain is involved in the kinds of
computation that are relevant to human abstract thought (simplistically
multiplying out the estimated number of synapses times the guesstimate of their
computational value in silico over time). However, most of the brain is not
dedicated for this purpose, and much of it is redundant, so that I think the
computational power necessary to run human-equivalent AI could be several
orders of magnitude less than Bostrom's 10^17 flops.

Your lower end of 10^14 is probably closer to the computational demands of
human minds on human wetware, but the theoretical needs of an efficient AI
could be several orders of magnitude less than that still.

That's IF we know how to streamline it. To brute force it, you might need much

Martin Striz

Do you Yahoo!?
Meet the all-new My Yahoo! - Try it today!

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:50 MDT