What are you?

A collection of information processing units designed through millennia of evolutionary trial and error to effectively model the environment and simulate interactions between the environment and these units.

Effectively?

Receptive fields of V1 neurons. These basic structures are the most effective way  to encode natural images.

Receptive fields of V1 neurons. Combining these basic structures is the most effective way to encode complex natural images.

Yes, one buzz word would be ‘sparse coding’.  Only a minority of units are needed to encode a typical stimulus. This is well documented in the low level V1 visual system in the brain. The receptive fields (the stimulaus that causes the neurons to fire) of V1 neurons that create a basis for all images are optimized for ‘natural images’.

It seems plausible that the rest of the units work in the same manner. Take emotions for instance. Wouldn’t they just be ‘sparse coding’ efficiently encoding interactions with the environment?

Fear, love, hate, disgust…they are all just a basis set, optimized to encode ‘naturalistic’ interaction.

To what end?

To perform actions on the environment that will ensure survival.  Any complex pattern requires a local decrease in entropy in order to different between inside and out, between environment and organism. In life forms the whole metabolic system couples exothermic reactions, that is, energy producing processes with endothermic processed that need that energy. Most life on earth arises from the energy delivered from the sun’s photons exciting electrons to a higher level, or as the biologist Albert Szent Györgyi put it “Life is interposed between two energy levels of the electron” (Although some life depends on deep sea thermal outlets or volcanos).  The survival of life depends on slowing the downwards cascade of this energy; maintain this local minimization of entropy. Any interaction with the environment is very costly although some interaction is needed to obtain more energy/food. This is why the the end goal of these units is to simulate possible action outcomes with the environment and choose which of them will minimize the loss of entropy/maximize survive chances. Another buzz word in this context would be, minimizing ‘free-energy’, or surprisal which is the information theory equivalent of entropy. The free energy principle tries to explain how (biological) systems maintain their order (non-equilibrium steady-state) by restricting themselves to a limited number of states. So if you are one of those who is constantly searching for a ‘goal’, minimization of entropy/free-energy seems to be as good as it gets. Striving for complex knowledge structures and creating just about anything and minimizing waste fit well with this idea but so does bringing more life into this world and cleaning my room.

Is that all that you are?

Me chronologically

Me chronologically

me chronologically inverted

me chronologically inverted

me, least information on top

me, least information on top

 

 

 

 

 

 

 

 

 

‘I’ am a model within a model, projected out into the world allowing others with a similar algorithm to effectively model me. Or perhaps one could turn the tables. The effective modeling others conduct when encountering ‘me’ is mirrored back into these units as a simplified singular version of itself.

One could postulate that there are as many ‘I’ as the people who know me, although they too are probably composed of an effective sparse basis.

 

For more reading:

Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences36(03), 181–204. http://doi.org/10.1017/S0140525X12000477

Cook, N. D., Carvalho, G. B., & Damasio, A. (2014). From membrane excitability to metazoan psychology. Trends in Neurosciences37(12), 698–705. http://doi.org/10.1016/j.tins.2014.07.011

Friston, K., & Kiebel, S. (2009). Predictive coding under the free-energy principle. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences364(1521), 1211–1221. http://doi.org/10.1098/rstb.2008.0300

Friston, K., Thornton, C., & Clark, A. (2012). Free-energy minimization and the dark-room problem. Frontiers in Psychology3(MAY), 1–7. http://doi.org/10.3389/fpsyg.2012.00130

Olshausen, B., & Field, D. (1996). Emergence of simple-cell receptive field properties by learning a sparse code for natural images Nature, 381(6583), 607-609 DOI: 10.1038/381607a0

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *