Intelligence from Causal Entropic Forces

Posted 24 Apr 2013 at 19:01 UTC (updated 24 Apr 2013 at 19:01 UTC) by steve Share This

You may recall our 2012 story, The Universe, The Internet, and the Brain about a paper identifying similarities between the structure and dynamics of the brain and the universe. Now, a new paper from Dr. Alex Wissner-Gross proposes a more dramatic connection: universes that produce more entropy (or disorder) over their lifetimes tend to have more favorable properties for the existence of intelligent beings. The paper describes how two attributes of intelligence, tool use and social cooperation, emerge spontaneously in simple physical system due to "causal entropic forces". According to the paper, these forces provide the motivation behind adaptive behavior. From the paper:

Recent advances in fields ranging from cosmology to computer science have hinted at a possible deep connection between intelligence and entropy maximization. In cosmology, the causal entropic principle for anthropic selection has used the maximization of entropy production in causally connected space-time regions as a thermodynamic proxy for intelligent observer concentrations in the prediction of cosmological parameters. In geoscience, entropy production maximization has been proposed as a unifying principle for nonequilibrium processes underlying planetary development and the emergence of life. In computer science, maximum entropy methods have been used for inference in situations with dynamically revealed information, and strategy algorithms have even started to beat human opponents for the first time at historically challenging high look-ahead depth and branching factor games like Go by maximizing accessible future game states

For more details, including the math, see the short paper, which was published in the 19 April 2013 issue of Physical Review Letters, Causal Entropic Forces (PDF format). You can also read a longer Inside Science article about the ideas in the paper along with comments from other researchers, Physicist Proposes New Way To Think About Intelligence. Or read on to see a short video showing software demonstrations of simple systems driven by entropic forces that spontaneously learn pole balancing, a simplified type of tool use, and social interaction.

Interesting Take on Intelligence and Entropy, posted 12 May 2013 at 01:30 UTC by MDude » (Journeyer)

I hadn't thought of entropy as relating to the idea of keeping your future options open. Also, I would have thought keeping the stick upright would have been the lower entropy action compared to dropping it, given that you could extract useful work from a falling stick but would have to use it to pick it back up. Though it does make more sense when looking at it in terms of information rather than classical physics. I wonder if how it would conserve or expend a limited fuel supply, especially if it was required for balancing the stick.

It would be interesting to compare and contrast with other learning systems. How would it handle the body of the starfish self modeling robot compared to the AI made for it? Would it try to model the body, or can it only decide how to use a body it already has a model of? If the latter, I suppose the two systems would be rather complementary to each other in a way. It'd be interesting to see if ENTROPICA could be easily integrated with other AI as a kind of subsystem/advisor. I also wonder how it would utilize other bodies, including modular and swarm robots, and whether it would tend to build replicas of and/or augment itself if given the opportunity.

Another thing I suppose you could contrast this with is Conway's game of life, which is irreversible due to each state having only one future but many potential pasts (or none, for some starting patterns), and could be seen as internally anti-entropic. Patterns in it are seen as somewhat life-like, but tend to die out easily and don't readily reproduce. I've wondered what would happen if you built a probabilistic reverse Conway's life that takes all possible previous states and selects while one while attempting to avoid those that themselves have no preceding states. Perhaps Conway's game is life-like due to being a kind of reverse life?

See more of the latest robot news!

Recent blogs

30 Sep 2017 evilrobots (Observer)
10 Jun 2017 wedesoft (Master)
9 Jun 2017 mwaibel (Master)
25 May 2017 AI4U (Observer)
25 Feb 2017 steve (Master)
16 Aug 2016 Flanneltron (Journeyer)
27 Jun 2016 Petar.Kormushev (Master)
2 May 2016 motters (Master)
10 Sep 2015 svo (Master)
14 Nov 2014 Sergey Popov (Apprentice)
Share this page