Source

Automatically imported from: http://commons.somewhere.com:80/rre/1996/Toward.a.Model.of.Mind.a.html

Content

This web service brought to you by Somewhere.Com, LLC.

Toward a Model of Mind as a Laissez-Faire Economy of Idiots

``` [I'm deleting the original header on this message so that nobody replies by mistake to the person who distributed this announcement, which is for a talk this Friday at the Salk Institute next door to UCSD. Freidrich Hayek is to capitalism what Karl Marx is to communism, and the program described here seems to have been modeled on Hayek's economic theory. Does that make the program a test of Hayek's theory?]

---

This message was forwarded through the Red Rock Eater News Service (RRE). Send any replies to the original author, listed in the From: field below. You are welcome to send the message along to others but please do not use the "redirect" command. For information on RRE, including instructions for (un)subscribing, send an empty message to rre-help@weber.ucsd.edu

---

SALK VISION SEMINAR (Presented by the Salk-Sloan Center for Theoretical Neurobiology)

"Toward a Model of Mind as a Laissez-Faire Economy of Idiots"

Eric Baum NEC Research Institute 4 Independence Way Princeton NJ 08540 eric@research.nj.nec.com http://www.neci.nj.nec.com:80/homepages/eric/eric.html

Friday, February 9, 1996 12:00 Noon

Trustees Room East Building The Salk Institute

Host: Dr. Terrence Sejnowski

(Note: This seminar room is located below ground in the area between the North and South wings of the Salk's new East Building. Enter through the glass doors located on the north side of the South wing, descend the stairs, and proceed to the right (eastward ) across the large travertine lobby. The "Trustees Room" is located along the eastern margin of this underground complex.) ABSTRACT A learning machine called ``The Hayek Machine" is proposed and tested on a simulated Blocks World planning problem. Hayek learns a set of agents from reinforcement. The agents interact in a market economy. A price mechanism is proposed and seen to have four desirable effects. First, the market price learns to estimate Hayek's future reward from using a given agent. Second, the market automatically selects the agent with highest estimate to act next. Third, new agents can enter the market if and only if they have greater expected utility than direct competitors. Fourth, by utilizing condition-action agents, and keeping exactly those agents which are profitable, Hayek dynamically addresses the critical ``curse of dimensionality". Hayek learns by gradual accretion of useful agents and elimination of poor ones,and by refinement of its price estimates. Many agents act in consort to solve problems. Our Blocks World (BW) problems, which involve discovering the abstract goal of copying a stack, and necessitate solving Towers of Hanoi-like problems, are far more complex than any BW problems previously addressed by a learning algorithm. Starting from tabula rasa, and with reinforcement only upon completion of an instance, Hayek learns to solve eight block problems. These are larger than can be solved by a simple handcrafted program utilizing the same impoverished representation, and of comparable complexity to those solvable by state of the art general purpose planners, which utilize human generated and learned heuristics. Given intermediate reinforcement or the simple features: ``top of stack", Hayek does much better. Given both, Hayek produces a program which solves arbitrary instances, and is efficient on almost all instances. My goals in studying Hayek are threefold. First, Hayek represents progress toward a useful program for learning to reason. Second, it is intended as a model of mind. It shows how human-like mental capabilities can be autonomously and usefully broken into simple components nowhere invoking a homunculus. Third, it is a model of economics. Utilizing very limited agents and starting far from equilibrium, Hayek achieves a productive and stable economy. It thus gives insight as to why and how far mathematical economic predictions based on assumptions of computationally powerful and rational agents, and on convergence to equilibrium, can be expected to hold in the real world. A full paper can be found at my web site: http://www.neci.nj.nec.com:80/homepages/eric/eric.html ```

This web service brought to you by Somewhere.Com, LLC.