Logic gates are the logical constructs that make up the framework for path generation in computer processing. The use of logic gates in computers predates any modern work on artificial intelligence or neural networks. However, the logic gates provide the building blocks for machine learning, artificial intelligence and everything that comes along with it.
A logic gate facilitates the choice of outputs depending on input in a computing system. Early on, this led to comparisons between a microprocessor and the human brain.
As work on neural networks started to evolve years later, a philosophy called “connectionism” came into play. Connectionism, which in some ways dates back to the 1940s, is the idea that complex behavioral patterns are generated through the combined work of individual small units – for example, in the brain, neurons.
All of this led to the idea of using programming, and in turn the underlying logic gates, for more complex processes. One of the definitions of machine learning is that the computer program evolves beyond the limits of what it was originally given as an input. In other words, the machine learns as it goes. It still uses the logic gates for processing given inputs and outputs, but the use of the logic gates for computing works in a fundamentally different way.
By continuing to study the human brain, and the performance of the neurons and synapses, scientists are getting closer to being able to model some of this activity with computing systems. Here, the logic gate will do the work of a human neuron.
Consider this excerpt from a scholarly paper on the design of various logical gates in neural networks:
“It is apparent that the neuron performs the equivalent of a logical OR operation on the excitatory inputs – if the presence of pulses represents a logical value of ‘1,’ then the behavior of an OR gate may be realized by a neuron with two excitatory inputs and the output fed back as an inhibitory input. The latter ensures that the neuron returns to a relaxed state when the excitation ceases, corresponding to a logical value of ‘0.’ The OR-gate neuron exhibits distinct ‘turn-on’ and ‘turn-off’ delays that change, depending upon past and present inputs.” – Suryateja Yellamraju, et. al., “Design of Various Logic Gates in Neural Networks”
It’s evident from this reading that close correlations can be made between the performance of an OR logic gate and the performance of a neuron working on binary excited or relaxed inputs.
With this in mind, artificial intelligence work often includes the use of logic gates in computing systems to model the types of behavior that are exhibited by neurons in the human brain. The extent of this modeling success will determine the future capabilities of strong artificial intelligence – whether by extremely advanced modeling, we can create sentient technologies, or whether the human mind proves sufficiently complex and elaborate to restrict or limit this kind of technological development.
In an article on Medium, V.V. Preetham talks about teaching logic to neural networks through the use of applied logic gates. This detailed tutorial shows how to represent the use of logic gates, and code, in ways that simulate the work of human neurons.
In this way, logic gates, which featured early on in the development of yesterday’s computing systems, continue to be the underlying resources for very advanced work in neuronal networks and the adoption of ever stronger machine learning and artificial intelligence tools that will dramatically change our interactions with technology in the years to come.