Difference between revisions of "Timeline of decision theory"
From Timelines
(→What the timeline is still missing) |
(→What the timeline is still missing) |
||
| Line 83: | Line 83: | ||
* [https://philpapers.org/rec/CAMTPD symmetry argument]? I found this paper linked in [https://qchu.wordpress.com/2013/06/30/cantors-theorem-the-prisoners-dilemma-and-the-halting-problem/] | * [https://philpapers.org/rec/CAMTPD symmetry argument]? I found this paper linked in [https://qchu.wordpress.com/2013/06/30/cantors-theorem-the-prisoners-dilemma-and-the-halting-problem/] | ||
* https://wiki.lesswrong.com/wiki/Parfit%27s_hitchhiker https://arbital.com/p/parfits_hitchhiker/ | * https://wiki.lesswrong.com/wiki/Parfit%27s_hitchhiker https://arbital.com/p/parfits_hitchhiker/ | ||
| − | * http://fennetic.net/irc/finney.org/~hal/udassa/summary1.html -- in a post on everything-list, hal also mentions that he at one point had udassa.com. it looks like wayback didn't capture it in time (by the time it got to it, the domain was parked and for sale) | + | * http://fennetic.net/irc/finney.org/~hal/udassa/summary1.html -- in a post on everything-list, hal also mentions that he at one point had udassa.com. it looks like wayback didn't capture it in time (by the time it got to it, the domain was parked and for sale); i'm not sure if udassa.com had anything different from finney.org/~hal/udassa |
* http://lesswrong.com/lw/gu1/decision_theory_faq/ | * http://lesswrong.com/lw/gu1/decision_theory_faq/ | ||
* http://lesswrong.com/lw/aq9/decision_theories_a_less_wrong_primer/ | * http://lesswrong.com/lw/aq9/decision_theories_a_less_wrong_primer/ | ||
Revision as of 23:15, 6 March 2020
This is a timeline of decision theory, with a focus on updateless/timeless/acausal/functional/logical decision theories.
Contents
Big picture
| Time period | Development summary | More details |
|---|---|---|
| up to 2006 | the bad old days | |
| 2006-2010 | Drescher, TDT, UDT | |
| 2010-2012 | more progress | |
| 2012 | stagnation? |
Full timeline
| Year | Month and date | Event type | Details |
|---|---|---|---|
| 1969 | Newcomb's problem is discussed by Robert Nozick. | ||
| 1980 | Brian Skyrms's Causal Necessity: A Pragmatic Investigation of the Necessity of Laws discusses the smoking lesion problem (or a similar problem that becomes called the smoking lesion problem in later publications).[1]:128–130 Yudkowsky and Soares cite Skyrms for the smoking lesion problem.[2]:3 | ||
| 1985 | The idea of superrationality is introduced by Douglas Hofstadter in his Metamagical Themas. | ||
| 1997 | The Sleeping Beauty problem is first formally analyzed. | ||
| 1997 | The absent-minded driver problem is introduced (in the same paper as the sleeping beauty?).[3][4] | ||
| 1999 | January 21 | Wei Dai posts the first description of what would later be called UDASSA is posted to everything-list.[5] UDASSA seems to be a precursor to UDT.[6] | |
| 2002 | July 17 | Hal Finney, in a mailing list discussion, brings up ideas that according to Wei Dai come "pretty close to some of the ideas behind TDT".[7][8] | |
| 2006 | March 29 | On the Theory of Everything Mailing List (everything-list), Wei Dai sends an email with subject "proper behavior for a mathematical substructure". He would later call this "a 2006 proto-UDT".[9] | |
| 2006 | May 5 | Gary Drescher's Good and Real is published.[10] | |
| 2007 | | ||
| 2007 | May 30 | Philosopher Kenny Easwaran blogs about his discussions with Joshua Von Korff. Korff has apparently devised a decision-theoretic protocol that one-boxes on Newcomb's problem but smokes in the Smoking Lesion problem. The post does not make clear when Korff came up with his ideas or whether he wrote them up anywhere.[13][14] | |
| 2009 | February | Eliezer Yudkowsky starts LessWrong using as seed material his posts on Overcoming Bias.[15] During the following years LessWrong would become the locus of discussion about timeless/updateless decision theory. | |
| 2009 | March 19 | Vladimir Nesov introduces counterfactual mugging.[16][17] | |
| 2009 | August 13 | Wei Dai publishes the post "Towards a New Decision Theory" on LessWrong. The post does not use the term "updateless decision theory" (UDT), but describes what would later be known as UDT1.[18][9] | |
| 2009 | August 20 | Gary Drescher proposes Metacircular Decision Theory (MCDT) in a comment on LessWrong.[19] | |
| 2010 | Timeless decision theory is published in paper form by Eliezer Yudkowsky.[20] | ||
| 2010 | February 18 | Wei Dai publishes "Explicit Optimization of Global Strategy (Fixing a Bug in UDT1)" on LessWrong.[21] This post introduces the decision theory UDT1.1, which improves on UDT1 by iterating over policies (observations-to-actions mappings) rather than iterating over actions. | |
| 2010 | April | Gary Drescher proposes the "agent simulates predictor" decision problem to the decision-theory-workshop mailing list (a private mailing list for discussing decision theory).[22] The problem would be published publicly by Vladimir Slepnev in May 2011. | |
| 2011 | Wei Dai proposes UDT2 in a post to the decision theory workshop mailing list.[23] The idea behind UDT2 would be described in a comment by Wei Dai in January 2014,[24] and by Vladimir Slepnev in a blog post in September 2013.[25] | ||
| 2014 | April 23 | Daniel Hintze publishes "Problem Class Dominance in Predictive Dilemmas".[26] The paper compares evidential decision theory, causal decision theory, timeless decision theory, and updateless decision theory (specifically, UDT1.1) on the decision problems Parfit's hitchhiker and the curious benefactor (equivalent to counterfactual mugging?). | |
| 2014 | November 4 | Project | The Intelligent Agent Foundations Forum, run by MIRI, is launched.[27] |
| 2017 | March 18 | "Cheating Death in Damascus" by Nate Soares and Ben Levinstein is announced on the Machine Intelligence Research Institute blog.[28][29] | |
| 2017 | October 13 | "Functional Decision Theory: A New Theory of Instrumental Rationality" by Eliezer Yudkowsky and Nate Soares is posted to the arXiv.[2] The paper is announced on the Machine Intelligence Research Institute blog on October 22.[30] | |
| 2018 | July 10 | The Alignment Forum beta is announced.[31] The forum is a website intended for discussing research in AI alignment. (Decision theory is sometimes motivated by AI alignment concerns.) |
Meta information on the timeline
How the timeline was built
The initial version of the timeline was written by Issa Rice.
What the timeline is still missing
- History of the concept of decision theory
- More on decision theory in academia, journals related to it, where it fits in with the rest of academia
- symmetry argument? I found this paper linked in [1]
- https://wiki.lesswrong.com/wiki/Parfit%27s_hitchhiker https://arbital.com/p/parfits_hitchhiker/
- http://fennetic.net/irc/finney.org/~hal/udassa/summary1.html -- in a post on everything-list, hal also mentions that he at one point had udassa.com. it looks like wayback didn't capture it in time (by the time it got to it, the domain was parked and for sale); i'm not sure if udassa.com had anything different from finney.org/~hal/udassa
- http://lesswrong.com/lw/gu1/decision_theory_faq/
- http://lesswrong.com/lw/aq9/decision_theories_a_less_wrong_primer/
- http://lesswrong.com/lw/5rq/example_decision_theory_problem_agent_simulates/ and more about the decision theory mailing list
- "the paper by Piccione and Rubeinstein that introduced the absent-minded driver problem" "p19 Piccione, Michele, and Ariel Rubinstein. “On the interpretation of decision problems with imperfect recall.” Games and Economic Behavior 20.1 (1997): 3-24." [2]
- Something about Spohn; see e.g. this comment
- "This idea follows in the wake of Gauthier (1994), who advocated making decisions using global policy selection, and Arntzenius, Elga, and Hawthorne (2004), who applied this idea to an infinite decision problem similar to the “Procrastination Paradox” of Yudkowsky (2013). Another decision procedure similar to that of Dai was proposed by Meacham (2010)" [3]
- when was TDT "officially" declared obsolete?
- I think several of cousin_it's posts should be included. i don't know enough yet to know which ones though.
- "the time they tried to hire a philosophy prof to write up TDT?" [4] See [5], [6], [7], [8]
- Will MacAskill has a Meta Decision Theory that's supposed to take into account uncertainty about which decision theory to use (sounds kinda similar to his approach to moral uncertainty?) [9]
- https://news.ycombinator.com/item?id=9321984
- UDT1.5/UDT2
- roko's basilisk
- https://ea.greaterwrong.com/posts/tDk57GhrdK54TWzPY/i-m-buck-shlegeris-i-do-research-and-outreach-at-miri-ama/comment/byH8abnt5RnPMunts
- interesting historical bit about UDT and two-boxing on newcomb: https://www.greaterwrong.com/posts/Kr76XzME7TFkN937z/predictors-exist-cdt-going-bonkers-forever/comment/afyRSrYtx8nP6kCs3 (i think wei dai has an older comment on LW saying a similar thing but giving less detail)
- vanessa kosoy and abram demski's discussion.
Timeline update strategy
See also
- Timeline of Machine Intelligence Research Institute
- Timeline of Center for Applied Rationality
- Timeline of AI safety
- Timeline of Wei Dai publications
External links
- "A comprehensive list of decision theories" by Caspar Oesterheld and Johannes Treutlein
- "Comparison of decision theories (with a focus on logical-counterfactual decision theories)" by Issa Rice
References
- ↑ Skyrms, Brian (1980). Causal Necessity: A Pragmatic Investigation of the Necessity of Laws. Yale University Press.
Suppose that the connection between hardening of the arteries and cholesterol intake turned out to be like this: hardening of the arteries is not caused by cholesterol intake like the clogging of a water pipe; rather it is caused by a lesion in the artery wall. In an advanced state these lesions will catch cholesterol from the blood, a fact which has deceived previous researchers about the causal picture. Moreover, imagine that once someone develops the lesion he tends to increase his cholesterol intake. We do not know what mechanism accounts for this effect of the lesion. We do, however, know that the increased cholesterol intake is beneficial; it somehow slows the development of the lesion. Cholesterol intake among those who do not have the lesion appears to have no effect on vascular health. Given this (partly) fanciful account of the etiology of atherosclerosis, what would a rational man who believed the account do when made an offer of Eggs Benedict for breakfast? I say he would accept. He would be a fool to try to "make it the case that he had not developed the lesion" by curtailing his cholesterol intake. […] Examples could be multiplied. R. A. Fisher once suggested that the correlation between smoking and lung cancer might be due to them both being effects of a common genetic cause. Fisher's hypothesis has not fared well, but if, contrary to evidence, it were true and you knew it to be true, and smoking were consistently pleasurable and not harmful in other ways, you would be foolish to refrain from smoking in order to lower the probability of having smoking-cancer gene. You either have it or not, and you can't influence your genetic makeup by abstinence.
- ↑ 2.0 2.1 Yudkowsky, Eliezer; Soares, Nate. "[1710.05060] Functional Decision Theory: A New Theory of Instrumental Rationality". Retrieved October 22, 2017.
Submitted on 13 Oct 2017
- ↑ "The Absent-Minded Driver". LessWrong. September 16, 2009. Retrieved September 10, 2017.
- ↑ "Absent-Minded driver - Lesswrongwiki". LessWrong. Retrieved September 10, 2017.
- ↑ Wei Dai (January 21, 1999). "Re: consciousness based on information or computation?". everything-list. Retrieved March 6, 2020.
- ↑ https://www.greaterwrong.com/posts/SkXLrDXyHeekqgbFg/shock-level-5-big-worlds-and-modal-realism/comment/yMCxvHCpBqsYEorpt
- ↑ "Wei_Dai comments on Common mistakes people make when thinking about decision theory - Less Wrong". LessWrong. Retrieved September 10, 2017.
- ↑ Finney, Hal (July 17, 2002). "self-sampling assumption is incorrect". Google Groups. Retrieved September 10, 2017.
- ↑ 9.0 9.1 "Wei_Dai comments on Taking Ideas Seriously - Less Wrong". LessWrong. Retrieved January 10, 2018.
- ↑ "Good and Real: Demystifying Paradoxes from Physics to Ethics (MIT Press): Gary L. Drescher: 9780262042338: Amazon.com: Books". Retrieved September 10, 2017.
- ↑ "Andy Egan, Some counterexamples to causal decision theory". PhilPapers. Retrieved September 10, 2017.
- ↑ "Smoking lesion - Lesswrongwiki". LessWrong. Retrieved September 10, 2017.
- ↑ "Different Ideas About Newcomb Cases". Thoughts Arguments and Rants. May 30, 2007. Retrieved September 10, 2017.
- ↑ "CarlShulman comments on Counterfactual Mugging". LessWrong. June 21, 2013. Retrieved September 10, 2017.
- ↑ "FAQ - Lesswrongwiki". LessWrong. Retrieved June 1, 2017.
- ↑ Nesov, Vladimir (March 19, 2009). "Counterfactual Mugging". LessWrong. Retrieved September 10, 2017.
- ↑ "Counterfactual mugging - Lesswrongwiki". LessWrong. Retrieved September 10, 2017.
- ↑ "Towards a New Decision Theory - Less Wrong". LessWrong. Retrieved January 10, 2018.
- ↑ "Gary_Drescher comments on Ingredients of Timeless Decision Theory - Less Wrong". LessWrong. Retrieved September 10, 2017.
- ↑ Yudkowsky, Eliezer (2010). "Timeless Decision Theory" (PDF). Retrieved September 10, 2017.
- ↑ Dai, Wei (February 18, 2010). "Explicit Optimization of Global Strategy (Fixing a Bug in UDT1)". LessWrong. Retrieved July 25, 2018.
- ↑ Slepnev, Vladimir (May 19, 2011). "Example decision theory problem: "Agent simulates predictor"". LessWrong. Retrieved July 25, 2018.
- ↑ "Comment on "Updatelessness and Son of X"". Intelligent Agent Foundations Forum. Machine Intelligence Research Institute. November 6, 2016. Retrieved July 26, 2018.
This does seem to be the “obvious” next step in the UDT approach. I proposed something similar as “UDT2” in a 2011 post to the “decision theory workshop” mailing list, and others have made similar proposals.
- ↑ Dai, Wei (January 15, 2014). "Comment on "Functional Side Effects"". LessWrong. Retrieved July 26, 2018.
- ↑ Slepnev, Vladimir (September 15, 2013). "Notes on logical priors from the MIRI workshop". LessWrong. Retrieved July 26, 2018.
- ↑ Hintze, Daniel (April 23, 2014). "Problem Class Dominance in Predictive Dilemmas" (PDF). Machine Intelligence Research Institute.
- ↑ Benja Fallenstein. "Welcome!". Intelligent Agent Foundations Forum. Retrieved June 30, 2017.
post by Benja Fallenstein 969 days ago
- ↑ Bensinger, Rob (March 18, 2017). "New paper: "Cheating Death in Damascus"". Machine Intelligence Research Institute. Retrieved September 10, 2017.
- ↑ Soares, Nate; Levinstein, Benjamin A. "Cheating Death in Damascus" (PDF). Retrieved September 10, 2017.
- ↑ Matthew Graves (October 22, 2017). "New paper: "Functional Decision Theory" - Machine Intelligence Research Institute". Machine Intelligence Research Institute. Retrieved October 22, 2017.
- ↑ Raemon (July 10, 2018). "Announcing AlignmentForum.org Beta". LessWrong. Retrieved July 25, 2018.
