Difference between revisions of "Updateless decision theory"

From Lesswrongwiki
Jump to: navigation, search
(Blog posts)
(writing stub)
Line 1: Line 1:
 
{{stub}}
 
{{stub}}
The best description of Updateless Decision Theory to date is in paper by Daniel Hintze, [http://intelligence.org/files/ProblemClassDominance.pdf Problem Class Dominance in Predictive Dilemmas], section 3.4.
+
 
 +
==Motivation==
 +
Updateless Decision Theory (UDT) is a new decision theory meant to deal with a fundamental problem in the existing decision theories: Treating the agent as a part of the world. In contrast, the most common decision theory today, Causal Decision Theory (CDT), the decision is not part of the world model--it is the output of the CDT, but in the context of the world the agent's decision is "magic": It is uncaused, like a dualist version of a soul with free will.
 +
 
 +
Getting this issue right is critical in building a self-improving artificial general intelligence. Such an AI must analyze its own behavior and that of a next generation that it may build.
 +
 
 +
==Content==
 +
 
 +
UDT specifies that the optimal agent is the one with the best algorithm--the best mapping from  observations
 +
to actions--across a probability distribution of all world-histories. ("Best" here, as in other decision theories, means one that maximizes a utility/reward function.)
 +
 
 +
This definition may seem trivial, but in contrast, CDT says that an agent should choose the best option at any given moment based on the effects of that action.
  
 
==Blog posts==
 
==Blog posts==
Line 14: Line 25:
 
*[http://lesswrong.com/lw/334/another_attempt_to_explain_udt/ Another attempt to explain UDT] by cousin_it
 
*[http://lesswrong.com/lw/334/another_attempt_to_explain_udt/ Another attempt to explain UDT] by cousin_it
 
*[http://lesswrong.com/tag/udt/ All posts tagged "UDT"]
 
*[http://lesswrong.com/tag/udt/ All posts tagged "UDT"]
 
*[https://formalisedthinking.wordpress.com/2010/08/18/an-introduction-to-decision-theory/ An introduction to decision theory] (series of posts)
 
  
 
==Relevant Comments==
 
==Relevant Comments==
Line 27: Line 36:
 
*[http://dl.dropbox.com/u/34639481/Updateless_Decision_Theory.pdf Formal description of UDT] by Tyrrell McAllister
 
*[http://dl.dropbox.com/u/34639481/Updateless_Decision_Theory.pdf Formal description of UDT] by Tyrrell McAllister
 
*[http://intelligence.org/2014/10/30/new-report-udt-known-search-order/ UDT with known search order] by Tsvi Benson-Tilsen
 
*[http://intelligence.org/2014/10/30/new-report-udt-known-search-order/ UDT with known search order] by Tsvi Benson-Tilsen
 +
[http://intelligence.org/files/ProblemClassDominance.pdf Problem Class Dominance in Predictive Dilemmas], section 3.4. (The best summary to date.)
 +
*[https://formalisedthinking.wordpress.com/2010/08/18/an-introduction-to-decision-theory/ An introduction to decision theory] (series of posts)
 +
  
 
==See also==
 
==See also==

Revision as of 06:27, 25 February 2016

Motivation

Updateless Decision Theory (UDT) is a new decision theory meant to deal with a fundamental problem in the existing decision theories: Treating the agent as a part of the world. In contrast, the most common decision theory today, Causal Decision Theory (CDT), the decision is not part of the world model--it is the output of the CDT, but in the context of the world the agent's decision is "magic": It is uncaused, like a dualist version of a soul with free will.

Getting this issue right is critical in building a self-improving artificial general intelligence. Such an AI must analyze its own behavior and that of a next generation that it may build.

Content

UDT specifies that the optimal agent is the one with the best algorithm--the best mapping from observations to actions--across a probability distribution of all world-histories. ("Best" here, as in other decision theories, means one that maximizes a utility/reward function.)

This definition may seem trivial, but in contrast, CDT says that an agent should choose the best option at any given moment based on the effects of that action.

Blog posts

Relevant Comments

In addition to whole posts on UDT, there are also a number of comments which contain important information, often on less relevant posts.

External links

Problem Class Dominance in Predictive Dilemmas, section 3.4. (The best summary to date.)


See also