Enabling AI to Learn through the Viable Systems Model


More in this series:
Machine Learning and Analytics in the High Performance Cloud

The Viable Systems Model was proposed and refined by Stafford Beer and others (John von Nemann, Norbert Weiner, W. Ross Ashby, Alan Turing, and many others) between the 1930’s and 1970’s.  What these theoreticians were attempting to model were the universal mechanisms for a system, any system, to be self-governing and self-perpetuating – Viable. The field of study Beer and his associates were operating in was called Cybernetics, the science of communication and automated control.  You could think of Cybernetics as a contemporary and companion to Systems Science, Control Theory, and all of the other disciplines that seek to describe and understand how things (gene regulatory networks, organisms, groups, corporations, political bodies, societies, etc.) adapt and change over time.  Beer was primarily concerned with corporations and economic governance, but, his models and theory is general enough and universal enough to be applied within any discipline including art and sports training. If you have a background in biology or psychology, this system will look very familiar as it makes use of positive and negative feedback loops .  

I’ve found the best way to describe the Viable Systems Model and its subsystems is to use a proxy.  I like to use a pride of lions as my proxy, but you could substitute wolves or any other social group of animals or insects (ants and bees work well too).  But, lions are big, furry, and roar, so lions it is.  

System 1 – Is the activity itself that describes the system, a living system, for example, metabolizes and respires (burns food, produces waste).  Lions do this, right?

System 2 – These are communication systems between within the living body, a nervous system or signally system of some sort.  Lions have nervous systems.

System 3 – This is the monitor and control system for System 1, in a living system, this system regulates simple activities like metabolic rate and respiration rate.  In humans and mammals, this can be thought of as the autonomic or involuntary nervous system. Depending upon the complexity of the organism, System 3 can also encompass circadian rhythms and other innate behaviors.  Lions sleep, are awake, hunt, mate and have other certain behaviors.

System 4 – This is the first set of outwardly looking systems that take in input from the external world or milieu.  These systems can be thought of as external sensors, touch, sight, hearing, and so forth. With these sensors are also rules that allow for self-preserving behaviors.  For example, System 2 communicates to System 1 that the organism is running low on energy (is hungry). System 4 identifies the communication as hunger and identifies a food source and begins to eat.  Lions do this very frequently, we can think of this as typical individual lion behavior.

System 5 – This is the component of the system that governs or balances System 4 activities against Systems 1-3.  For example, if we look at the pride of lions, we see System 5 activities taking place with feeding priorities, adult males first, young, just weaned cubs are higher up the feeding ladder (eating with their mothers who did the hunting) eat second, and older cubs/juveniles, who eat last.  This is done to assure the next generation of cubs can nutritionally make it to adulthood, provided there is enough food and resources in the environment while maintaining the social order of the pride. In this case, System 5 is the lion pride dynamics that govern a group of lions and modulates their behavior as they impact their environment.  

So, how does all of this map into our “Ideal Model”?  

First, a few, ground rules, system mapping does not necessarily have to correspond with the order of execution, but rather function.  So it is possible to start with something in System 5 or even System 4, rather than System 1.

viable systems model

Well, in our ideal architecture, we can easily map functions like “Semantic tagging” and “Model training” to System 1 as these are the basal “metabolic” functions of the architecture.  System 2 processes map to the internal mechanisms that divide and move data from stage to stage. We see these behaviors manifest in “Semantic aggregation of training data”, and “Convert Model to RTL….”  System 3 is the connectivity between the cloud and edge as shown by the dotted blue lines and the red lines leading back to “Semantic tagging and metadata creation” from “Inference testing” and “Push RTL to edge….”  System 4, the nervous system that controls the rate of Systems 1 and 2 can be thought of the rate of change inside of the “Public and private repositories of electronic media.” System 5, are those forces that operate on the “Public and private repositories of electronic media” that are influenced by our AI.  In short, System 5 deals with the issue of the architectures impact upon its environment or community, in short, how the AI impacts the community rules, policy, or ecosystem of the data being consumed. In many cases, the impact of System 5 should be negligible, but, for completeness, it is included and is often considered an outward facing policy layer.

At this point, you are either confused or are thinking, “Great, we just mapped one theoretical construct to another.  Big deal!” Well, it kind of is a big deal. By satisfying the Viable Systems Model we have now enabled our AI to learn and adapt over time in response to its environment and the data it is receiving.  In short, we are making the transition from training to learning. Learning will be important when we begin to look at fields that are non-traditional applications for AI.