• Innovation GPT-4, OpenAI's new language model, is here
  • Don Felipe at CILE "This is the century of Spanish"

David Carmona, director of the Department of Artificial Intelligence and Innovation at Microsoft, returned this week to his city, to Cádiz, to participate in the International Language Congress. On Monday, he was one of the professionals who explained to King Felipe how to work so that Artificial Intelligence speaks the Spanish language, how it is nourished by phrases and computers are trained to complete, also in our language, the next great technological revolution.

Find out more

Leaders.

Why linguists are the new heroes of Artificial Intelligence

  • Writing: LUIS ALEMANY Madrid
  • Writing: PHOTOGRAPHS: CARLOS GARCÍA POZO

Why linguists are the new heroes of Artificial Intelligence

Culture.

Disappointment at the slowness of the Perte de la Lengua: "We no longer have him"

  • Writing: LUIS ALEMANY Madrid

Disappointment at the slowness of the Perte de la Lengua: "We no longer count on him"The fact that he now works with natural language, is it a surprise in his life? Not so much. I have the personal feeling that the two worlds meet again after 30 years. I had a very fortunate upbringing where they instilled in me a love of mathematics and a love of language. It is one of the strengths we have in Spain, our education does not completely separate those two worlds and that allows us to have a different touch. It has helped me all my life, I have always considered myself a storyteller as well as a technology professional. It has always been very helpful. In reality, the reunion of these two worlds is what is enabling the technological revolution we are experiencing. It all boils down to the fact that we have gone from having an Artificial Intelligence (AI) that reasoned about numbers and helped us to be more efficient to another that reasons about human language. It's a huge change. First, because human behavior is encoded in words. If we can train computers in words, we can train them in human knowledge forged over centuries, we lead them to reason like humans and not just act on machines. The second change is that, since your new way of working is language, using it is much easier. We just have to ask them what we want them to do with our words. Many more people will have access to Artificial Intelligence going forward. I see how online translators have improved and I wonder if it will not be unnecessary to work so hard for our children to learn languages... It is an example to say that perhaps it will change our relationship with knowledge. I wouldn't recommend anyone to abandon that effort to learn languages. Exposing oneself to the culture of others is such a valuable experience in itself and so beautiful that it will not be questioned by Artificial Intelligence. No algorithm will give our children that experience. What does change is that languages are no longer a barrier to access to knowledge.Are we not heading then to an intellectually and professionally idle society? There is a very powerful concept that we always use: joint reasoning. Technology, which helped us to be more efficient, will henceforth help us reason. It is a great transformation, it allows us to expand human reasoning around all the knowledge of humanity, and so on at work, in education, in research, in personal life ... But we always talk about Artificial Intelligence as a co-pilot: the human has to remain in possession of control. Next to you, we put an assistant that helps you reason. What can limit the scope of that revolution? There is one obvious: lack of confidence can frustrate many changes. Society has to rely on new technology. For that we have to develop it together and in a responsible way institutions and companies. Well, the first thing we all think when listening to ChatGPT is "and what will happen to my work?". I guess there's an instinct of distrust. But we must convey the idea of before: we are creating an enlargement, not a substitution of the human. Always, anywhere, they ask me what's going to happen to my work. Now they ask me also in jobs related to knowledge. And I always say that nor we have to worry about Artificial Intelligence doing our job better than us. That the concern should be that someone else does our job better than us next to an Artificial Intelligence. This revolution will impact all jobs and it is important that there are protection mechanisms so that this change is assumed with serenity. But it will not replace us. And what can public administrations do well? Companies often ask me what to do and I always tell them the same thing: to offer a platform for the whole company to succeed with the new technology. Better than centralizing the effort in a department of excellence would be to allow all employees, who are the ones who know what processes can be improved, to be enabled to work with Artificial Intelligence. Well, that reasoning also applies to me for the public: that the administrations make things easier for citizens, that they give them access to data, computing, training ... How does the PERTE of Language sound to you? There is a chapter that talks about AI and language, which I find especially relevant, because it is a very important moment for Spanish to be at the forefront. Our task is to explain to our partners and customers how to act. At the moment, what we see are many transfers between administrations and public or semi-public companies. I don't see it that way. Microsoft does a lot of collaboration with the public sector. We are in LEIA with the RAE, we have integrated their services in our products, their criteria of use correto the Spanish. On this issue, the public and the private have to go together, because, if we go separately, some challenges are very complicated. People listen to ChatGPT4 and think that it is done, that the machine already speaks well. It has been a very important step. The interesting thing is that these algorithms have been trained on a corpus of many languages simultaneously and that has helped them. The transfer of knowledge works, what the machine learns in English, it also knows in Spanish. It is the same as bilingual children who learn two languages, who study mathematics in Spanish but also know how to subtract in English. There is another fact: algorithms are not only trained with huge databases. They also do it in dealing with users. That is why it is important that there is a true use. Yesterday it was denounced in Spain that the algorithms used by VTC vehicles of the Uber, Cabify and Bolt type make trips that begin or end in low-income neighborhoods more expensive. . And that shows that it's not just innovation that matters, it's the spectacular part. Responsible development matters as much or more. This is a textbook example of a case of bias. There are other challenges that always arise: transparency, privacy, accountability... Every company has to have a conversation to decide how to deal with them. It's not as easy as putting a checkup at the end of the process. From the moment a scenario is projected, you have to think about the possible biases. And so on, even the user interface. You have to guarantee the person that you are going to be the pilot.

According to The Trust Project criteria

Learn more

  • Spanish language
  • Artificial intelligence
  • Articles Luis Alemany