Artificial Life - Next gen AI - Engineering Karma
In Artificial Life (next-gen AI), behavior emerges when the different agents make choices with each other. Good karma is making right choices. What can enable this karmic synthetic world?
AI now - Neuron mimicking Algorithms
Present day AI works by mimicking human neural Network. Every node is connected to each other and the connections are fixed. Learning is managing connections, by changing the connections weights. The system is trained by exposing it to data and control the weight so that it provides the right answer when it is queried with an input.
Artificial Neural Network is just a Function approximation Algorithm that attempts to mimic neurons in us
Training this network is very expensive and it does not adapt (a.k.a Drift) to the environment without retraining the network. But real life, constantly changes so retraining every-time environment changes is not very practical.
Time to advance AI to overcome these challenges by mimicking life and not neurons.
Next AI - From imitating Neurons to imitating Life
Life, is a biological process, that does signaling and self-sustaining processes. Each biological entity (human being, animal, etc) communicate with others thru the environment. For an individual cell in our body, the environment is the human body. The cells do not communicate with cells that are considered foreign (i.e. cells that do not have the same genesis), ensuring human being integrity. How is this organized?
Let us explore this with a simple example of relationship between me and Bubbles. I experience him when i play with him. He responds by wagging his tail, a hint that he is enjoying the experience. This observation of wagging the tail, changes me and i respond by repeating the behavior. Technical term, for this is shared transition. i.e It is when two biological entities do a handshake/interact and continue to be themselves. Like the interaction between cells in our human body.
Next Gen AI is artificial Life. I.e modelling shared transitions or interactions
Generalizing and expanding on this, Artificial life (AL) is many processes (aka Agent) that interact with each other. Agents adapt by listening to signals from other agents and communicating the same to other agents that are listening. To the outside observer, this agent configuration changes, appears as learning. The learning is immediate, solving the present day machine learning drift problem.
Taking our instance of bubbles and me, it is not possible for me to influence him intrusively, we can only learn from each other by interacting with each other by signals presented. This level of non-intrusiveness makes the agent autonomous (bubbles and me in this case). Our biosphere is full of living autonomous agents.
Synthetic life i.e next generation AI, is shared transition or interactions of autonomous agents.
Why AL & not AI?
For a AI system to be trusted it should have mathematical principles behind it. Mathematical rigor, introduces explain-ability & repeatability. Success of present day programming is because of its Lambda calculus mathematical base. Present day AI, does not have a mathematical model which explains it and hence cannot explain its output and all the other attributes expected out of a AI system.
Unlike present day AI, AL (Artificial life) operates on the premise of interacting autonomous agent creating processes. Pi calculus , theory of processes, give AL rigor and makes technology explainable and repeatable. Synthetic life, is a process that move forward, by the choices taken by these agents. It also offers privacy and security.
AI systems based on “Artificial Life” is trustworthy because of mathematical rigor
What is karma? and what is good karma?
Karma, is the philosophy of eastern thought. It is the great law of "cause and effect", of "action and reaction", which controls the destiny of all living entities. Synthetic life consists of Society of Autonomous agents (much like a human society). The synthetic life is what we observe when all these Agents in the society interact with each other to create “cause effect” sequences.
A Synthetic life is said to be operating in Good karma if all the agents in that society operate to the benefit of the society. Some autonomous agents in the society can be a destroyer (like the killer T cells of the human body) to keep the good karma in the society.
Ethical AI is all the autonomous agents in society operate in good karmic manner.
What is this tech that drives this emergent Causal world.
Nature does not compute. It is not world of algorithms with fixed input and outputs, but a world in which everything is a cause effect sequence.
The above figure is extracted from my article in ACM. It brings a case of algorithms as cause and effect rather than the present day sequential computation. It shows how two algebraic expressions run in parallel with two expressions which share a common shared cause effect channel A. First expression raises A and second expression listens to A and when they come into existence they interact and move the expression forward.
As you can see, even computation is a cause effect sequence
Whole systems is a network or society of autonomous agents that interact with each other as Cause and effect sequences. This new processor that models this system cannot be sequential. It should operate on interactions rather than the present day computation paradigm. This new processor that enables synthetic life is based on this new processor i.e Interactions.
Enterprise switch from axon is one such Karmic (Interaction) processor. Built to create this causal karmic world.
Conclusion
Artificial life based systems will transform your enterprise to next generation enterprise. For example,supply chain is not viewed not as a chain of many supplier organized in a linear fashion, but is re-imagined as a smart space of autonomous agents that interact with each other, lowering working capital required and increasing top-line by reducing stock-outs.
These systems will be have all properties of nature like Self healing, Self governing, Self executing, and self growing attributes.