Re: The Spike, nanotech, and a future scenario

Hal Finney (
Mon, 6 Oct 1997 08:36:57 -0700

Eliezer S. Yudkowsky writes:
> Hal Finney wrote:
> > Actually I find that the Singularity stays about 20-30 years in the
> > future. I'd like to see some evidence that anyone in 1980 predicted it
> > to be in 2035, 55 years in their future. I don't think anyone even knew
> > about the Singularity back then.
> Going from memory (books lent out):
> Vernor Vinge gave birth to the concept in 1979.
> The first calculation with respect to human-equivalent AI was performed by
> Hans Moravec and gave a result of 2035. I believe this was 1980 or
> thereabouts, although it could be 1984.

Some people trace the Singularity concept to Vinge's story "Bookworm,
Run". This was published in 1966 and predicted intelligence enhancement
in 1984 (according to the notes published with the story in True Names).
The story ends with the expection of a transformed world.

Hans Moravec's Mind Children was published in 1988. He predicts that
computing power capable of human-level AI "would be available in a $10
million supercomputer before 2010 and in a $1,000 personal computer
by 2030." This would range from 22 to 42 years away.

> > When the movie 2001 was made, it did not seem implausible
> > that a space station, lunar bases, AI, and a manned mission to Jupiter
> > could occur in that time frame.
> Yeah, but that was wild optimism, not an extrapolation based on previous rates
> of progress.

If you plotted such things as the fastest a human had ever gone,
or the highest a human had ever gone, you would undoubtedly have
people at Jupiter by 2000 if you extrapolate that curve from ca 1970.
Gregory Sullivan wrote here here last December about predictions by G.
Harry Stine in 1961. Stine extrapolated based on previous rates of
progress and found that human speeds should become infinite in 1982.
Instead we saw a levelling off of the rate of progress.

> > Personally, I think we are going to hit a big wall in computer technology
> > in the next ten years.
> Yawn. Copper circuits, qubits, quantum transistors, STM, nanotech, diamond
> disk drives, parallel processing, networked computing, etc, etc, ad Singularitum.

Diamond disk drives, parallel processing and network computers
are irrelevant to the technological limitations I was referring to.
I don't know much about copper conductors, but I suspect that they will
provide only a modest improvement in performance, and don't help with the
limitations of the silicon gates themselves. Qubits, quantum transistors,
STM, and nanotech are unlikely to be providing commercial-quality and
-cost computer components in this time frame. For the next decade it is
unlikely that we will see a commercially viable move away from silicon,
and if we do, it will probably be to something like gallium arsenide,
which at least has been proven to be usable to build computer circuits.

The real question is this: will we see a failing of Moore's law in the
next ten years? Will the performance per price of computers continue
to grow by a factor of two every two years? This would require five
doublings in the next ten years. I don't think it will happen.

> > But I am always suspicious of that 20-year prediction horizon. We can
> > guess what will happen technologically in the next ten years, but beyond
> > 20 we really have no idea. "Here there be dragons," and we are inclined
> > to put our wonders safely in the 20-30 year period. In practice though
> > things often take much longer than we expect.
> 2008? How safe is that?

2008 is not within the 20-30 year prediction horizon, but few people will
agree that it is a reasonable prediction for the Singularity.