Re: extropians-digest V1 #110

Eugene Leitl (
Mon, 18 Nov 1996 15:27:19 +0100 (MET)

On Mon, 18 Nov 1996, Michael Lorrey wrote:

> Eliezer Yudkowsky wrote:
> >
> > Computer power doubles every two years.

Not true. We're in the saturation zone already. The development is
discontinous. After maspar becomes mainstream, we'll see a jump, until
we'll hit semiconductor photolitho limits. then we have to wait for
wait for the advent of nanotech.

Have a look how current machines become. Memory bandwidth is rising
hardly at all. Due to cache hierarchy paradigm, pipelining, branch
prediction, and all that happy shit our machines have become finicky,
unpredictable in respects to optimization. Progressively catankerous.

Read "High Performance Computing" by Kevin Dowd, O'Reilly. It will make
your hair stand up.

> > Computer power doubles every two *subjective* years.

I dunno. Monster applications hog resources quite efficiently. Very
little, if at all subjective speedup is left.

How long it takes your machine to boot? Two generations back my
home machine booted in about 1 seconds. It was slow, yet I landed right in
a relatively powerful Basic environment. One could hack algorithms with
it, it was a useful tool.

On my home unix machine, I require X, shell, xemacs, make, gcc & Co,
needing lots of ugly incantations before I even can begin with solving my
problem. Though Java is 20 times slower, being an interpretive language,
it still offers faster turnaround cycles.

Sorry, the usefulness has decreased imnsho.

> > Computer power goes to infinity four years after computer researchers
> > reach human equivalence.

Current PC/workstation equivalent to 100 biologically realtime neurons
(_not_ wet neurons).

> > Mathematically, I believe it's called a "singularity".

Mathematical singularities are not identical with the technological
Singularity. It's just a mental, fuzzy picture. Sufficiently different to
become unpredictable. Way beyond the predictability horizont.

Ask the maestro himself: