Organic Learning

Organic Learning: Exploring Dichotomies

Carlos Castellanos & Steven J. Barnes
Keywords: emergence, self-organization, cybernetics, autopoiesis, autonomy, learning, self-construction, associative learning, coincidence detection, artificial intelligence, artificial life, autonomous agents
The Organic Learning experiment allows us to explore the extent to which a quasi-organic electrochemical solution and electrode assembly can manifest features of associative learning.1 If there is sufficient simultaneous activation of motion-gated anodes and a sound-gated cathodes any resulting plasticity and current fluctuations would constitute a bioelectrical record of sensory-sensory learning. The purpose of this particular experiment is not merely to demonstrate the presence of associative sensory learning but is more broadly to explore the classic dichotomies of inorganic vs. organic and non-biological vs. non-biological. Our system is, by standard scientific definitions, inorganic. Yet, we have commonly observed patterns of bifurcated growth and dissolution that have qualities classically reserved for organic biological systems.
This experiment utilizes several properties related to theories of biological learning as a conceptual framework for the design of self-constructing systems capable of evolving associations among various stimuli from their environment. Our method explores if and how dendritic thread plasticity might serve as a coincidence detector.2 If our system does have such a capability, then it should manifest itself in the network as both a plasticity of the dendritic processes and as a long-term potentiation or depression of current flow between the respective anode and cathode.3
An electrochemical solution [img] functions as an adaptive self-constructing, self-repairing entity capable of evolving its relationship with the environment so as to give rise to behaviors characteristic of biologic learning. Like the Emergent Relations experiment, of particular interest in this experiment is the adaptive interaction of the unstable dendritic network with the many uncertainties of an external environment such as an architectural space.
The gating of current through each of the individual electrodes is controlled by features of a gallery environment. Each of nine electrodes (the anodes) is gated by motion in one zone near the test apparatus, while each of the remaining four electrodes (the cathodes) is gated by the presence of sound within a particular frequency range (i.e., low, low-mid, high-mid, and high range) in the gallery. In short, the circuit through the stannous chloride solution would only close when at least one anode and one cathode were active at the same time.
Organic Learning set-up
The Organic Learning setup. Coincident sound and motion information was fed into the stannous chloride solution

This experimental set-up allows us to test the boundaries of the inorganic and organic, the non-biological and the biological, by attempting to show that our "inorganic non-biological" system can manifest properties comparable to those associated with a biological system that is learning about aspects of its environment (e.g., neuronal and glial plasticity, or long-term potentiation/depression of synaptic communication).
1. Balsam, Peter D., Drew, Michael & Gallistel, C.R. "Time and Associative Learning," Comparative Cognition and Behavior Reviews 5, 2010, 1-22.
2. Stuart, Greg J., & Häusser Michael. "Dendritic coincidence detection of EPSPs and action potentials", Nature Neuroscience 4, 2011, 63-71.
3. Abraham, Wickliffe C. & Robins Anthony, "Memory retention--the synaptic stability versus plasticity dilemna," Trends in Neurosciences, 28, 2005, 73-78.