?

Log in

No account? Create an account

Previous Entry | Next Entry

lecture

I am trying to decide if I should go to this lecture tomorrow. It looks like it could be extremely cool, has a very transhuman feel to it. I've always enjoyed hearing about ways that humans could improve or circumvent evolution.


"Beyond Competition: Towards a principle for
Self-organization and continuous learning"
Jordan B. Pollack
Brandeis University
http://www.demo.cs.brandeis.edu <http://www.demo.cs.brandeis.edu/>
Thursday, March 20, 2003
3:00 p.m.
MaRC Auditorium

How does life progress to such complexity as to eventually achieve
cognition? The biological matrix dissipates energy from the sun leading to a
local reversal of entropy, which results in ever more complex order.
Mechanisms of astonishing beauty emerge from interactions between
replicating forms in a non-cognitive process of open-ended knowledge
discovery. For the past decade my students and I have been studying
"artificial co-evolution" as a computational approach to self-organization
and automatic design. Co-evolution is easiest envisioned as a population of
game-playing agents who learn better and better strategies simply by playing
the game (Self-learning).

We have had many successes in optimization, language, games, problem
solving, and - most publicly - robotics. Yet these systems - organized
using Darwinian notions of competition - always seem to run out of steam.
The agents discover collusive equilibrium, which halt progress. They
disengage into winners and losers who cannot learn from each other. They
engage in boom and bust cycles. Achieving a fundamental understanding of why
co-evolution stalls had led to discovery of a new principle of
self-organization, which is neither competitive nor altruistic. It may have
numerous applications, including peer-to-peer educational technology.

Biographical Sketch

After an early career in the software industry Jordan B. Pollack received
the PhD from University of Illinois in 1987. He taught at Ohio State
University from 1988-1994 prior to moving to Brandeis University in 1994
where he is on the faculties of computer science and complex systems. He was
named one of MIT Technology Review's "TR 10" in 2001.

Jordan has advised many PhD's and Postdoc's who have contributed to a broad
range of different fields including neural networks, dynamical systems,
evolution, machine learning, cognitive science, artificial life, robotics,
and educational technology. His lab, the "Dynamical and Evolutionary Machine
Organization" (DEMO) has been partially funded by ONR, NSF, DOE, and DARPA.
Beyond academics, http://jordanpollack.com <http://jordanpollack.com/> is a
prolific inventor, founded a company called Thinmail, advises startups,
incubators, and VC's, develops new theories of intellectual property and
works for world peace.


--
Charmion Richards
Administrative Manager
College of Computing
Georgia Institute of Technology
801 Atlantic Drive
Atlanta, GA 30332-0280

Comments

( 1 comment — Leave a comment )
spoonless
Mar. 20th, 2003 04:18 pm (UTC)
Well, I decided to go. And I'm glad I did, although I was a bit disappointed.

Apparently, Jordon Pollack was one of the two guys in charge of the Golem project, which I participated in a couple years ago (a distributed project over the internet where lots of computers work simultaneously on helping to evolve virtual robots by mutating and adapting to their environment). I'm definitely a fan of his "mind-body coevolution" approach to AI, but it's not really a new idea.

There were two main points to his lecture. One was this mind-body coevolution, which just basically says that instead of trying to evolve a robot brain to fit a pre-existing robot body; or to evolve a body to fit a mind you've already programmed for it--we need to evolve them both together, so that its body can adapt to the environment and it learns to use its body as it grows
and learns.

The other main point to his lecture was about a new technique he calls "incursion," which is an alternative to pure competition for approaching teaching and learning. This was the idea for which I really came to the lecture, although it turned out far less promising than I'd anticipating. I was hoping he was taking it to a level of replacing evolution, but it turns out he is after much more mundane things even though I think he implied that in the description to get people to come. In the end, he figures we may find out that evolution has a lot of incursion built in to it already. And that our purely competitve models of evolution weren't accurate to begin with. Almost all of the questions asked by the audience were stupid, except for one guy behind me who asked several good ones. Pollack's answer to the question he asked about whether he saw incursion of a means of supplementing evolution or replacing it was "Well, if I were Stephen Wolfram that's what I'd say." I admire him for his humility, and his unwillingness to say too much at this point, but as it turns out they really haven't done much work at all and what they have learned didn't seem all that promising to me.
( 1 comment — Leave a comment )

Profile

blueshirt
spoonless
domino plural

Latest Month

May 2017
S M T W T F S
 123456
78910111213
14151617181920
21222324252627
28293031   

Tags

Page Summary

Powered by LiveJournal.com
Designed by Lizzy Enger