From: David S. Hansen (haploid@haploid.com)
Date: Mon Jan 20 2003 - 16:28:38 MST
On Mon, 20 Jan 2003, Dani Eder wrote:
...snip...
> To put some numbers to this theory, I estimate
> 'economic crossover' for general design tasks based
> on when computer power is cheaper than brain power.
> Human brain power is estimated to be equivalent to
> 100 TFlop to 100,000 Tflop. A computer running
> 24x7 and working at 1 human design engineer equivalent
> rate would be worth about $3 million. So my
> estimate is that when you can buy computers for
> $30,000 to $30 per TFlop plus 7 years (to allow
> time to usefully deploy the computers) is the
> expected date for the Singularity. We are 5-20
> years from reaching the cost threshholds, so that
> gives a prediction of 2015 to 2030.
That's a nice model that gives some interesting numbers, but I would tend
to disagree that human-brain computing parity will be required. Consider,
for a moment: the cerebellum, the motor cortices, the limbic system, all
of which contain billions of neurons which are used in most estimations of
the "power" of the human brain, but which are almost completely irrelevant
for abstract design tasks.
Assuming that artificial design engineers will not require processing
power necessary for motor control, proprioception, audio or perhaps even
visual processing, I believe it would be safe to say that many teraflops
can be shaved off the computational capacity neccessary for complex design
tasks such as self-reengineering. It would not be unreasonable to move
that time window up by several years.
- Dave
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:41 MDT