Artificial intelligence, the science behind machines that think

Artificial intelligence was the protagonist of Telefónica’s first live broadcast on Twitch. Guanma and Nolia, from the Communications Department, unearthed the characteristics of this innovative technology from the hand of 20-year-old Pablo Gomez. we discovered What is artificial intelligence and what is its purpose.

This young man who hopes for artificial intelligence is a student in Campus 42, of Fundación Telefónica, and has met with the Study of the Data Science Profession, IA at the Polytechnic University of Madrid. As Pablo explains, the machine “thinks and thinks in a certain way, and helps us grow as a society.”

What is artificial intelligence?

Paul brings us all closer to the true meaning of this very modern system, which is only a century old. It’s not about magic, it’s about science. This technology goes beyond the concept of “machines can think, reason, and feel, and they are already defined”, although we must not forget that “these definitions will certainly evolve in the future”. Despite all the suggestions, Pablo takes the definition created by scientist Michael Mitchell to make us understand that it is not magic, and says “A system is smart if its performance improves based on experienceThis can be compared to the way we humans learn: How do we learn? The young expert asks, based on failed attempts, the experience provided by previous failures. “And the machine will try to reproduce this method of learning.”

There are three types of AI that are applied depending on the type of problem the hardware has to face. Did you know? Supervised, Unsupervised and Reinforcement Learning. Artificial intelligence algorithms are predictive algorithms, they take one or several input values ​​(data) to analyze and, accordingly, give the output results, i.e. prediction. It is a reliable predictor because, before using it, it is pre-trained, which leads to the accumulation of experience gained.

How do you train the algorithm? It gives you millions upon millions of bits of data so you can get patterns. An obvious use case is use in the analysis of medical examinations to obtain a fast and reliable prognosis prediction. Companies like Google are already using this type of AI to analyze and improve crops.

For its part, unsupervised learning trains the algorithm in another way. In this case, it presents a pattern that helps classification rather than prediction. It is widely used in social networking platforms, where it finds patterns of behavior among users of the platform, and categorizes the content they consume according to their tastes, among other applications.

While previous algorithms are trained on data sets and learn from experience, reinforcement learning does it differently than humans do: trial and error. In this case, in order for the algorithm to learn, a system of “rewards” is created according to the decisions you make, whether they are correct or not. This can be applied, for example, to cleaning robots that we can use in our homes.

AI replicates human behavior, but how good is that?

Gwanma and Noulia have sparked philosophical controversy over these operations, an aspect of interest to society, with risks such as Deep Fake or identity theft. To talk about it, Pablo asked the question: Is technology good or bad? Or does it rather depend on the use we give it?

But what if artificial intelligence goes even further in the future? Will we stop functioning as a society? According to the expert, “Replacing humans with technology will not help us advance as a society if people are not productive, and we will stop understanding how this technology works.” On the contrary, technology can help us a lot, and he gives an example of the support it gives when predicting diseases, “but for this we will not do without doctors.”

As Paul says, Understanding how they work helps demystify the ethical and moral debates they provoke. It’s a very useful innovation but we shouldn’t get too excited. However, a lack of regulation can harm communities. For example, the presence of prejudices (unequal propensity), which is one of the biggest problems they suffer from, and this is something that societies have to combat.

How do we benefit from artificial intelligence?

The medical sector was one of the sectors that benefited the most from the development of these technical developments. They’re helping clinicians find patterns, thanks to analyzing vast amounts of data, which might otherwise go unnoticed by health professionals. There is artificial intelligence that is able to detect emotional states through the use of social networks.

Other great AI apps can be found at Mobility and driving. Parking aids, or more advanced uses such as autonomous driving that improve safety on the road, aim to reduce fatalities while driving, or problems caused by mobility such as traffic congestion and the pollution that it entails, and solve problems globally, as it is expected that in 2025 there will be more of 500 million connected cars.

Speaking of using languages ​​and machine translation (instead of working with a professional translator), this led Noelia, Juanma and Pablo to another discussion: Will machines replace us at work? Paul mentioned to Jose Ignacio Latorrean expert on the subject, shares his opinion: Technology should complement people’s work, not replace them, because machines are unable to understand the emotional nuances of humans.

However, due to its great potential, it is applied in many, more and more fields.

In the footsteps of Touring, the father of Ai

The work of Alan Turing is present in shaping this young promise of artificial intelligence. “His life was complicated and despite his misunderstanding, he knew how to implement a lot of projects, among which we can consider him one of the fathers of artificial intelligence.”

We owe it to the British mathematician for designing an electronic digital computer known by its acronym ACE, the Automatic Computer Engine, and another computer, called Manchester Mark I, in 1947. With his interest in replicating the functions of the human brain in these machines, he laid the foundations for this technology around 1950.