AI IN CONTEXT OF LIFE

Recently, I was discussing with a friend, who had a view, that AI has limited applicability, as in most cases we do not know why AI is giving the results that it is giving. That is, lack of explanatory power will be a nemesis for widespread use of AI.

At the first glance, this argument sounds very powerful and hence correct. However, here I try to show a deeper, albeit very brief analysis, to reveal that AI has very long runway to go, maybe spanning many human generations, before we run out of applications for AI. Let us start with basics

Life as Entropy Reduction Machine

For a very long period now, I have held view that the reason of life is to reduce local entropy of the universe. More on that in some other post, let us accept this premise now and move on.

Computational Reducibility: Mechanism to Reduce Entropy

If we subscribe to the view of Wolfram that the universe is computation machine based on very simple rules; then there exists a very large set of algorithms which can help you make predictions without any reference to underlying rules.

Some examples of computational reducible algorithms are Ptolemy’s Epicycles, Newtons Laws of Motion and Einstein’s General Theory of Relativity. Each one of them predictions about positions of planets, while none of them really explain how gravity really works. Each new algorithm is better than the previous algorithm in the sense that the newer algorithm explains more of the reality. But none of these “scientific” laws tell us the actual mechanics of gravity,which still eludes humankind.

Computational Reducibility and Need for Story Telling

Why do humans need to tell stories? The reason is pretty simple. We have access to wide variety of computationally reducible algorithms which help us make seemingly correct predictions for most part, without really understanding the reality. At the same time, we cannot accept that we do not understand the true nature of universe. Hence, we need to create stories where actors with intentions act on the universe and shape it.

AI as Additional Tool-set for Expanding Computational Reducibility

Since life emerged on earth about four billion years ago, it is acquiring newer skills which helps it to access an expanding set of computational reducible algorithms. In the last seven thousand years, humans have further expanded those algorithms with emergence of logic, and debating skills in Vedic period in India which later were polished by the Greeks. Later, development of concept of zero and decimal computation algorithms led to explosion of mathematical toolkit which led to multiple industrial and now digital revolutions.

Now, in the last one decade, AI is creating access to newer algorithms which will help life, mainly through humans, access wider sets of computationally reducible algorithms. This will enable life to reduce entropy in the universe with more effectiveness. The process that began on earth four billion years ago, has just now acquired a new tool. So, opportunities and possibilities are enormous for application of AI.

Finally, it is worth remembering that in terms of explaining how exactly gravity works on planets, the now discredited Ptolemy’s Epicycles fails as much as Eisenstein’s Relativity. Both belong to the category of computationally reducible algorithms. Of course, Relativity makes far better predictions than Epicycles. But none explain the ground truth yet. AI also belongs to the same category. We accept the story told by Relativity now after 105 years of predictions so much that we have forgotten that it does not explain. So, eventually, we will accept power of AI when its predictions help us in our lives, while forgetting that AI does not explain reality.