RE: rambling question on AI

From: Colin Hales (colin@versalog.com.au)
Date: Sat Jan 05 2002 - 22:09:10 MST


Happy new year, folks.

I just got back from a brief holiday. While the wet suits are drying, I
thought I'd check out the >1000 extropian emails. I swore I'd go lurking
full time in Y2K+2. So much for that new year's resolution :). This whole
area of finance is very close to home for me and I am faced with this very
dilemma right now (as is Peter Voss). I think I have something to add to the
discussion: Specifically

Observation #1 ___There's $$$ in *not* doing AI____
I have been cruising the WWW in the AI and neural area for so long and into
so many little corners my head spins. What I have noticed is that the human
brain (as opposed to mind), as source of incredible complexity, justifies
money to do apparently very rewarding and totally obscure projects .eg.
"Quantitative analysis of a directed behavior in the medicinal leech".

There is always the justification that something is learned with respect to
human pathology .ie. somehow we are better off knowing <insert subject
matter> because we can help people with <insert condition/disease>. When
human pathology is involved (eg cancer) $$$ seem to flow differently. Just
try and formulate an argument _NOT_ to do something to directly help human
health. Unpopular indeed.

To me, it looks like there's a industrial scale business in decoding the
brain and if some aspect of 'mind' pops out, then it's a nice/interesting
accident but, so what?. From the point of view of AI I liken it to trying to
debug a word processor by looking at the silicon in the computer running it.
It's a long and apparently lucrative endeavor. You can write lots of papers,
publish books, go to workshops and lectures, actually be seen to do
something to apparently directly help humanity. Careers can be run along
these lines, and there's very few people to argue alternate courses of
action.

Am I seeing something here, or am I imagining things?
I'd be interested in your thoughts.

Observation #2 ___IMHO this is not a computer science problem____
I can relate to Mr Reason's gripe....big time.....:

> Now if you're walking into a VC's office to pitch an AI company -- one of
> the few places you're guaranteed not to be the strangest project they've
> seen -- they're going to want to hear how you're going to be generating
> revenue within six months from an initial investment. [Since you're a
> software company, that is, and producing those easy-to-make software
things,
> as opposed to an infrastructure company making those hard-to-make lumps of
> iron, but I digress -- lambasting VCs is far too easy].

....except in the tacit assumption that this is a software exercise. It
implies that all you have to do is pull a few comp sci folk in a room and
".....there shall spring forth a gazelle". This I liken to identifying a
group of carpenters with all their own tools and then saying to
them....."build a communications satellite". It'll end up made of wood and
the solution to all problems will be "more carpenters!!!". Anyone who
believes this is a software exercise alone will, I believe, suffer the
consequences of project bloat and end up the slave of a self-serving monster
that will die, having eaten all of its food chain. {Can you tell I have run
a few software teams in my time :-) - It's called aversion therapy}.

The comp-sci parts of this are a component of the the AI end-game. Go to
them with a detailed specification, get competitive quotes and hold them to
it! From that perspective it's probably a good thing that VCs want payback
so soon - it'll put limits on the length of the food chain from the start
and kill it appropriately early.

Observation #3 ___A Camel is a Horse designed by a commitee____
The other side of this argument/observation is that AI becomes, by it's own
nature, a nebulous thing to sell to a VC. How do you convincingly sell an
endeavor that needs Neurophysiology, automation engineering, mathematics,
computer science, electrical/electronics engineering, psychology,
philosophy? Well IMO you can't. The VC has to share the vision - now there's
a challenge. A contradiciton in terms?

observations done....
----------------------------------------
My ideal VC with truly prodigious gonads....
1) Hear me out on work to date.
2) Give me A$300,000 per year for 2 years. (defined goal, no promises)
3) Negotiate extensions after 1 year based on progress.
4) Let me get on with it (I will grudgingly agree to monthly reports on
progress)
:)
*dream on, Colin*

In my own AI quest I have been unable to avoid Mr. Yudkowsky's "life is what
happens while you are making
other plans" syndrome. It's a reality that gets a bit tiresome when you have
to take your whole family along for the ride. It ...tests your resolve, to
put it mildly.

What the hell. Just do it, I say! Life's too short.
(how about that, an extropian joke into the bargain)

cheers,
Colin
*returning to cash cow construction with renewed vigour and, yes, dry wet
suits (now there's an odd phrase)*



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:32 MST