[visad] Compression/prediction inference in intelligence, Matching pennies, elusive model paradox

• To: billh@xxxxxxxxxxxxx
• Subject: [visad] Compression/prediction inference in intelligence, Matching pennies, elusive model paradox
• From: "David Dowe (Infotech)" <david.dowe@xxxxxxxxxx>
• Date: Sat, 16 Jul 2011 04:19:32 +1000

Dear Bill,

Hi.   I'm not sure whether I've e-mailed you before re some similarities
and/or overlaps of
interest in our work.    I do see that you cite my joint work with Jose
Hernandez-Orallo
(who I'm currently visiting in Valencia) Hernandez-Oralllo & Dowe (AI
Journal, 2010).

I've read your 3 papers below:
*Measuring Agent Intelligence via Hierarchies of
Environments<http://www.ssec.wisc.edu/%7Ebillh/g/hibbard_agi11a.pdf>
,* paper at The Fourth Conference on Artificial General Intelligence
(AGI-11<http://agi-conf.org/2011/>
). August 2011

* Societies of Intelligent
Agents<http://www.ssec.wisc.edu/%7Ebillh/g/hibbard_agi11b.pdf>
,* paper at The Fourth Conference on Artificial General Intelligence
(AGI-11<http://agi-conf.org/2011/>
). August 2011

*Matching Pennies <http://matchingpennies.com/>* is Tim Tyler's web site for
a tournament of algorithms competing at the game of "matching pennies." I
recommended such an algorithm competition in my
AGI-08<http://agi-conf.org/2008/>paper
*Adversarial Sequence
Prediction<http://www.ssec.wisc.edu/%7Ebillh/g/hibbard_agi.pdf>
*, which is equivalent to matching pennies. February 2011.

Might I invite (or ask) you to please have a look at (at least) two sections
of my
D. L. Dowe (2011a [was to be 2010]),
"MML, hybrid Bayesian network graphical models, statistical consistency,
invariance and 
uniqueness<http://www.csse.monash.edu.au/%7Edld/Publications/2010/Dowe2010_MML_HandbookPhilSci_Vol7_HandbookPhilStat_MML+hybridBayesianNetworkGraphicalModels+StatisticalConsistency+InvarianceAndUniqueness_pp901-982.pdf>
",
Handbook of the Philosophy of
Science<http://japan.elsevier.com/products/books/HPSorg.pdf>- (
HPS <http://japan.elsevier.com/products/books/HPS.pdf> Volume 7) Philosophy
of Statistics <http://www.johnwoods.ca/HPS/#Statistics>, P.S. Bandyopadhyay
and
M.R. Forster (eds.), Elsevier, [ISBN: 978-0-444-51862-0 {ISBN 10:
0-444-51542-9 / ISBN 13: 978-0-444-51862-0}],
pp901-982<http://www.csse.monash.edu.au/%7Edld/Publications/2010/Dowe2010_MML_HandbookPhilSci_Vol7_HandbookPhilStat_MML+hybridBayesianNetworkGraphicalModels+StatisticalConsistency+InvarianceAndUniqueness_pp901-982.pdf>,
1/June/2011
http://www.csse.monash.edu.au/~dld/David.Dowe.publications.html#Dowe2011a
?

On the issue of agents in social environments and two-part (MML) inference
vs
prediction as a measure of intelligence, you should at least possibly find
sec. 7.3
(pp956-958) of abovementioned Dowe (2011a) [and references therein] and
D.L. Dowe, J. Hernandez-Orallo, P.K. Das (AGI, Aug 2011)
"Compression and intelligence: social environments and
communication"<http://users.dsic.upv.es/proy/anynt/paper5-compression.pdf>
(Artificial General Intelligence, Mountain View, San Francisco, August
2011<http://agi-conf.org/2011/>)
:
[This paper discusses the choice between multiple-model inference and
single-model inference, and the two-part code vs. one-part code, and the
implications for intelligence, social environments and communication.]
to be relevant and/or useful.

Re (iterated) matching pennies (which I have called the ``elusive model
paradox''),
you might possibly be interested in my discussion of (variations of) this
problem from
[Dowe 2008a, footnote 211][Dowe 2008b, p455] and (abovementioned) [Dowe
2011a, sec. 7.5]
and also sec. 2.2 of (abovementioned) Dowe, Hernandez-Orallo & Das (AGI,
2011).

I was a good friend of Ray Solomonoff's (and am chairing his memorial
conference
in Melbourne [Australia] in Nov/Dec 2011) and I worked very closely with
Chris
Wallace for 13+ years.  The comments on Solomonoff towards the end of my
Dowe (2008a, ``Foreword re C. S. Wallace'') reflect my take (loyal to them
both)
on (two-part {Wallace} [MML]) inference vs (Solomonoff) prediction.

Bill, do feel most welcome to let me know what you think and to keep in
touch.

[Paramjit, as a next step in your research, I encourage you to read Bill
Hibbard's
three abovementioned papers.]


Best to all,

David.

---------- Forwarded message ----------
From: Jose Hernandez-Orallo <jorallo@xxxxxxxxxxx>
Date: 6 July 2011 06:16
Subject: Elusive model paradox
To: "David Dowe (Infotech)" <david.dowe@xxxxxxxxxx>


Hi David,

I already sent you this link some time ago:

http://www.ssec.wisc.edu/~**billh/g/hibbard_agi11a.pdf<http://www.ssec.wisc.edu/%7Ebillh/g/hibbard_agi11a.pdf><
http://www.ssec.wisc.edu/%**7Ebillh/g/hibbard_agi11a.pdf<http://www.ssec.wisc.edu/%7Ebillh/g/hibbard_agi11a.pdf>
>

I've been taking a look at it, and in section 2 it describes de "elusive
model paradox" in the context of adversarial learning.

Best,

Jose.