THE FUNDAMENTAL NEED FOR LIFELONG LEARNING

This drive for reinvention led Walter Gropius, an architect from Berlin, to found a school that would serve and modernise society by fusing two disciplines: art and design. It would be the reflection of his ideas about architecture and creation, and was to be called Bauhaus (“Building House”). Today, it’s over 100 years old and remains an avant-garde symbol of how, through a new vision, a need spawned by a context was met through accessible design. Its innovative spirit continues to inspire us, as it’s a reflection of how an entire society, sunken and depressed, was able to reinvent itself and participate collaboratively, contributing talent to achieve reconstruction.

In a context similar to that of the Bauhaus, as a society we also feel this need for renewal. Our context is advancing as fast as Usain Bolt, and yet we’re moving slowly, gently, smoothly.

AND, BEING HONEST, WHEN WAS THE LAST TIME YOU LEARNED SOMETHING COMPLETELY NEW?

Our latest revolution goes hand in hand with digitalisation: a new origin. The fourth industrial revolution is not limited to adding new concepts, but rather to adding them in all areas of our lives. Some are more aware than others of the benefits of automation, sending automatic email responses thanks to RPA (Robotic Process Automation).

Ever since Klaus Schwab, Executive Chairman of the World Economic Forum (WEF), officially introduced this term in 2016, it has been understood as an essential strategy to further the advancement of all nations and people. For the rest of us mortals, however, it sounds a bit out there, or even scares us, because this is the first time that we’re seeing how a revolution in our socio-economic and production model is beginning to compete with our cognitive abilities.

WHY DOES THIS SCARE US SO MUCH?

At the end of the 1950s, coinciding with the first computers (what a coincidence!) cognitive psychology began to emerge, based on an interest in the mental processes involved when we process information. At that time, it began to be understood that people – like computers but with two eyes and a mouth – learn by acquiring and processing knowledge and understanding through thought, experience, and the senses. They were trying to figure out how our “human CPUs” perceived our environments and, consequently, developed problem-solving abilities and stored memories.

Everyone learns differently. There are as many learning theories as there are personalities in the world, all with their adherents and their detractors.

What is learning, then? Simplifying the matter, we might say that we learn by detecting patterns. When we observe a phenomenon (we receive information) we try to understand it by looking for the relationship between the event’s different variables (we transform it into knowledge).

According to Goel Gaurav, storyteller and expert in data science and engineering:

“All human learning consists of observing something, identifying a pattern, building a theory (model) to explain this pattern, and testing this theory to check whether it fits in most or all observations”.

It’s always been said that if you know more than one language, it’s much easier to learn additional ones… but why? What could German possibly have to do with Chinese? At first glance, it seems like apples and oranges. However, according to the article “Picking up a second language is predicted by ability to learn patterns” in the journal Psychological Science, put out by the Association for Psychological Science, it has to do with the ability to detect patterns between languages.

One of the findings of psychologist and researcher Ram Frost (Hebrew University of Jerusalem) was that students who scored better on pattern identification tests (including visual pattern ones) had better language learning results at the end of the semester (in this case, Hebrew).

Our brains also need shortcuts; basically, as Gaurav reminds us in his article:

“Every learning, fundamentally, is a model expressing a pattern in a set of observations. If there’s no conceivable pattern, there will be no learning”.

When we look at it this way, we humans are not that different from AI, right? Returning to 50s nostalgia to understand the cognitive processes of these clever little machines, we realise that, when all is said and done, Machine Learning (ML) follows learning paradigms that allow the computer to process all the information that will eventually be recorded as knowledge – and the process is really similar to ours.

Knowing how these mechanisms work is essential in the field of ML. In fact, all the algorithms and techniques within this field can be classified into three large groups, depending on the learning paradigm they are applying: supervised, unsupervised, and reinforcement learning.

In terms of supervised learning, this is a type that is based on discovering the relationships between input and output variables. Like people, this learning model needs observations; that is, lots of data, labelled and classified. Thus, we teach the technology the results we want to obtain for a certain initial value.

Let’s imagine that we have a series of numbers in which for a value of 1, the result is 2. For a value of 2, the result is 4. For 4, we get 8 as a result. For 8, the result is…? Exactly, 16. Here, your brain has generated knowledge through observation. Much like learning a new language, you’ve generated a relationship between the input and output variables, noting that the repeating pattern is to multiply the input value by 2. And this, something so simple and commonplace for our grey matter, is the basis of supervised learning.

At this point, it seems that AI with ML acquires knowledge by going to summer school while everyone else is off having fun at the swimming pool, such that it will soon surpass us in every field.

SO, WHY SHOULDN’T WE BE AFRAID?

We won’t compete against machines. They constantly learn and relate information, much faster and with fewer errors than us. That is true. Even so, we’ll become more important, to maximise their potential. This is one of the paradoxes of automation: the more efficient the automated system is, the more relevant human contribution is.

To be honest, not even good Will Hunting or Katherine Johnson, at NASA, would be able to surpass AI in terms of calculation speed, data processing skill, and data analysing agility while maintaining the same level of accuracy.

Luckily, humanity always manages to pull an ace out of its sleeve and gain an advantage. Reviewing what’s known about learning psychology, one can see how Bloom’s Taxonomy sheds some light on the previous statement. Although Bloom also based his work on behavioural and cognitive theories, he understands the learning process as something more than just receiving information, processing it, and storing it on our “hard drive” in the form of knowledge.

In fact, he proposes a model in which learning is classified into three domains: the cognitive, affective, and psychomotor domains (though it focuses on the first of the three). His cognitive taxonomy seeks to get the best out of the learner in order to achieve certain cognitive objectives (6 to be exact) based on intellectual, affective, and psychomotor abilities:

1. Knowledge, the ability to remember the information we have processed (a good hard drive).
2. Comprehension, transforming this information (data) into something we can understand and interpret (labelling and classifying).
3. Application, using what has been learned in a practical way (automatically detecting whether an email is spam or not, detecting images, or diagnosing a disease).
4. Analysis, breaking down what has been learned and developing hypotheses and testing them (making models, detecting patterns, and extrapolating knowledge).
5. Synthesis, briefly developing a model that puts together the information received and creates something different from what has been learned.
6. Evaluation, being critical based on certain criteria or opinions built around a system of values.

As can be seen, up until number 4, AI could be the classic class “know-it-all”. It takes human nature, however, to create and evaluate. Going back to the automation paradox, our transhuman companion may process information much faster, more accurately, and with fewer errors than us, but if we know how to turn that around and take advantage of AI’s full potential, we’ll be able not to compete with the machine (one can’t compete with something that is always learning) but rather to use its capacities to be better professionals and, yes, to put it to work for us.