Re: What is Intelligence?

Dan Clemmensen (dgc@shirenet.com)
Mon, 30 Sep 1996 20:33:00 -0400


Robin Hanson wrote:
>
> Dan Clemmensen writes:
> >I feel strongly that speed of access to information and speed of
> >disemmination of new information are very important contributore to
> >the equation. ...
> >Computers are a new phenomon. Their impact is only now being seriously
> >felt. So far, the contribution to knowledge has been mostly in the
> >infrastructure: Communications, word processing, publishing, etc. More
> >... Comuputers are now essential elements in the development of new
> >digital circuits, including new computers. Thus, I feel that the
> >situatin you describe is changing very rapidly. ...
> >I prefer not ot predict the exact nature of the way computers will
> >augment intelligence, since IMO it's not central to our discussion. I
>
> Sure computers are new, and will new kinds of impacts. And at one
> time cars were new, trains were new, radio was new, etc. But the
> question here is, do they fundamentally change the nature of economic
> growth any more than these did? You have stated your belief in this,
> but have not offered any reasons.
>
Cars, Trains, and radio are not intelligence-augmemnting technologies
in the same sense that computers are likely to be. This is the crux
of my argument.

> >If uploads are similar to other types of algorithms,
> >faster will be preferable to more, because multiple processors spend a
> >lot of time on cummunication of intermediate results.
>
> Sometines a few fast processors are best, sometimes lots of little
> ones are best. That's the way it is now, and should remain.
>

To the extent that I have any area of expertise, it's in the area of
distributed processing, computer communicatins, and multiprocessor
architectures. There is a very large body of research in this area.
Your statement is correct in the abstract, but in practice you start
with the fastest reasonable processor you can get. You don't use
multiple processor unless a single processor is very expensive.
For simple communications-intensive problems such as communications
switching, the power of a multiprocessing system goes up with the
log of the number of processors: for equal processors, a four-processor
system is twice as fast as a single-processor system. Different
problems have different characteristics, and the interconnection
topology can be optimized for the problem, but the algorithms get
complicated very quickly.

> >I feel that "knowledge-filled" will be superceeded by the ability to
> >access the entire knowledge base at a very high data rate.
>
> You lost me here.
>

That was a bit cryptic, wasn't it? Your post mentioned knowledge and
processing power as contributors to intelligence. I assumed from context
that you were modelling an intelligent entity as containing a processor
and
a data store, along with other stuff. My point is that the data store
can be distributed and shared via communications. If it is, the complete
body of knowledge contributes to each intelligent entity. I propose
that the size of this database is not the only factor in intelligence:
the speed at which it can be accessed is also important.
Consider: the WWW currently contains about 2 terabytes of accessible
information: so does a reasonable University library. However, I can
access the whole web rapidly. Thus the web amplifies my intelligence
more than all of the world's university libraries do.