Robots again

From: Doug Yowza <yowza_at_yowza.com>
Date: Mon Mar 16 04:34:49 1998

> Come to think of it, whatever happened to Prolog? I had always planned to
> pick up Borland's Prolog and learn it, but I guess it's too late for that now.

Why too late? I found it hard to do anything very useful in Prolog, but
the idea of a "predicate calculus" engine is still very powerful. You
probably write a small Prolog program every time you do a complex search
in a search engine. I'm sure you can find a few public domain prologs on
the net.

> Was there just no demand for AI, or has the market just taken the useful
> aspects and abandoned the remainder? Isn't the Neural Net technology being
> used in various pattern recognition applications (i.e. OCR) descended from
> the AI research of the the 70's and 80's?

Exactly. All of the useful ideas from AI that became mainstream are no
longer considered AI. Today, OCR, speach recognition, machine
translation, and predictive analysis are off-the-shell apps or embedded in
products like Microsoft Word to help catch your spelling and grammar
errors.

There are still interesting problems, though. Machines can kick your
chess-playing butt, but you won't find one nearly coordinated enough to
hit a baseball and run around a few bases. IMHO, AI researchers have
overestimated the brainstuff and underestimated the sensor and actuator
stuff. Here's my theory of how you learn to speak, for example:

1) When you're a little robot with a blank EPROM, you get bombarded with
meaningless sights and sounds from the environment.

2) Quite by accident, you generate some random motor impulses that allow
you to produce your own sounds.

3) Your richly cross-linked RAM network allows you to make associations
between the sounds you hear and the motor activity used to produce them.
When the associations get strong enough, your EPROM gets programmed (look
for a book called Neural Darwinism for a good discussion of the biological
mechanisms). Add a little non-linear feedback to get your nets from
getting stuck in simple linear relationships..

4) Eventually you can correlate sounds made by others with your own sounds
(imitation), then you associate sounds your mom makes with visual input
("doggy"), and associating sounds and mouth movements with abstract ideas
isn't far off....

Ob classicmp: Lisp Engines :-)

-- Doug
Received on Mon Mar 16 1998 - 04:34:49 GMT