**From:** Emil Gilliam (*emil@emilgilliam.com*)

**Date:** Tue Jan 04 2005 - 12:52:12 MST

**Next message:**Eliezer S. Yudkowsky: "*Annual* SL4 Chat, Wed Jan 5th @ 9PM Eastern"**Previous message:**Billy Brown: "RE: Definition of strong recursive self-improvement"**Messages sorted by:**[ date ] [ thread ] [ subject ] [ author ] [ attachment ]

This month longtime member and general inspiration Ray Solomonoff will

be delivering three lectures on

Algorithmic Probability and Artificial Intelligence

When: 12, 19 and 26 January 2005 -- 7 to 10 pm

Where: Bldg 32, room 144

MIT

No enrollment limit, no advance signup.

Participants are welcome at individual sections.

The lectures will be about an hour followed by

questions and discussion. See below for information

on individual sessions.

Lecture notes will appear at the URL

http://world.std.com/~rjs/ These notes will be

periodically updated and will eventually

include text of the entire set of lectures

Contact: G.J.Sussman, gjs@mit.edu

Sponsor: Electrical Engineering and Computer Science

Cosponsor: Engineering Systems Division

Lecture 1: Algorithmic Probability

Definitions of induction, prediction.

Motivations for redefinition of probability: Heavy and

light models -- over and under fitting.

Algorithmic Probability and the Universal distribution.

Three kinds of probabilistic prediction: How

Algorithmic Probability gives very small expected error

in all of them.

How algorithmic probability differs from minimum description length

(MDL),

Rissanen's Stochastic complexity, Kolmogorov complexity.

Incomputability of Algorithmic Probability -- How it enables good

probability

estimates despite its incomputability.

Subjectivity in algorithmic probability, in other Bayesian methods,

and in non-Bayesian methods.

Basis of subjective a priori. The Anthropic principle.

Explanation based learning. Placebo explanations.

Is an Anthropic principle necessary?

Lecture 2: Applications of Algorithmic Probability

Linear and nonlinear prediction.

Neural Nets and Genetic Programming.

Lecture 3: Systems for Strong Artificial Intelligence.

A definition of Strong Artificial Intelligence. Training

sequences. The role of Levin's Search Algorithm and Enhanced

Genetic Programming in preliminary and advanced Artificial

Intelligence systems.

www.pobox.com/~fhapgood

_______________________________________________

Nsg mailing list

Nsg@polymathy.org

http://polymathy.org/mailman/listinfo/nsg_polymathy.org

**Next message:**Eliezer S. Yudkowsky: "*Annual* SL4 Chat, Wed Jan 5th @ 9PM Eastern"**Previous message:**Billy Brown: "RE: Definition of strong recursive self-improvement"**Messages sorted by:**[ date ] [ thread ] [ subject ] [ author ] [ attachment ]

*
This archive was generated by hypermail 2.1.5
: Wed Jul 17 2013 - 04:00:50 MDT
*