Control: an emergent biological property

Abstract

This paper contains the speculation that associative memory neural circuitry evolved from control neural circuitry. As yet, no one been able to train a jellyfish to any associations. The jellyfish however, can attack prey, feed, swim, reproduce and perform all other functions necessary for life. Since we know that slightly more advanced creatures can form associations, a reasonable speculation would be that the neural circuitry involved in forming associative memories evolved from the neural circuitry involved in control tasks. In this paper this speculation is further explored and its consequences discussed. Our explorations have given further credence to this speculation. We have been experimenting with simple biologically plausible artificial neural networks, or ANNs, which consist of two input neurons and one output neuron and are called three-neuron controllers (TNC). The biological plausibility of our ANNs arises from the fact that the activity level of the receiving neuron is determined by integrating over time the transmissions of the two input neurons. The transmission rate of the two input neurons is determined by their activity levels, which we have assumed constant over each integration. The integrating of the two inputs achieves a form of control which resembles that obtained by a fuzzy logic controller.