# Artificial Intelligence: It is not MAGIC. It is MATHEMATICS.

Surely you have ever wanted to teach a skill or knowledge to a co-worker, a friend, or your children.

If it is explicit knowledge (that which is structured) you can transmit it orally or in writing through specific instructions. As if it were a kitchen recipe, describing the steps orderly and providing all the necessary information for each one of them. They can also be guidelines of the cause-effect style: “if this happens, do this, if instead, the situation is this other, then do that.” The instructions can be numerous and complicated, but you are able to give a solution for each of the possible situations.

This knowledge is easy to “teach” also to machines. It is the traditional coding of computers. Programmers write the code with all the instructions the machine needs, covering all possible scenarios. And machines are much faster than people executing these instructions, without errors or breaks.

At other times, it is tacit knowledge (diffuse and not easy to formalize) that is difficult for us to explain and communicate to others. The teacher needs to share time with the apprentice to teach. This knowledge is acquired based on examples, observing, deducing or imitating, and above all practicing. And once assimilated it arises when we need it without knowing how.

How can we transmit this fuzzy knowledge to machines? In a similar way to how we do it between people.

We feed the computer system with large amounts of examples (Data). And we ask it to find common characteristics or patterns that allow us to propose rules (Algorithms). Then we can test the algorithm with new Data, and adjust it to improve the result. It is an ongoing trial and error process that never ends. This is basically how Machine Learning (ML) works.

Let’s take a simplified example. If I want an ML to identify photos of cats, I will give it thousands of photos with cats of different races and colors (Data). This is the part of the explicit knowledge that I can pass on: “these are pictures of cats.” And I will ask it to discover patterns in the images using statistical and mathematical methods. It will begin by discovering that in all the images 2 triangular shapes that are the ears are repeated, it will also discover a repeated shape similar to a double opposite “Y” (one up and one down) that is actually made up of the nose and mouth of the cat.

That’s it. With this, the ML can build its first Algorithm proposal: if there are 2 triangular shapes and a double “Y” then it is a cat. We will test the algorithm with more Data (photographs) and review the result to verify that it works correctly (in the same way that a teacher would do). Perhaps we will discover that in some cases the hypothesis fails and mistakes a Tiger for a Cat. The algorithm needs to be improved. With more statistics and mathematics, you will find out that in the photos of tigers the pattern of the stripes on their fur is also repeated, so we will modify the Algorithm including this new characteristic: “If it has triangular ears, double” Y “ but it has stripes, it is not a cat”. The system learns from experience, adapts, and improves. It evolves.

This learning is a long and costly process of continuous trial and error until a sufficiently low degree of failure is achieved. Traditional programming offers concrete and precise results. ML offers a probability of success (in tasks in which traditional programming cannot be applied).

And the current growth of ML is possible thanks to 3 factors: new and better statistical and mathematical tools have been developed to find patterns in the data, we have much more powerful and cheaper chips (Moore’s Law) to run these algorithms, and we have access to large amounts of data that we can use to feed and teach ML systems.

People assimilate knowledge (Tacit. Unstructured) without knowing very well how or how to transmit it. And we call it experience and instinct. In reality, they are patterns that our psyche has recognized with time and effort.

Machine Learning does the same to learn: discovers patterns.

It is not magic. It is mathematics.

(although as Arthur C. Clarke said: “Any sufficiently advanced technology is indistinguishable from magic”)

Notes:

Nonaka and Ikujiro (The Knowledge-Creating Company: How Japanese Companies Create the Dynamics of Innovation)