Pedro Gonçalves. “Technology has always been used for good and for evil”

“Lexology, in association with the Association of Corporate Counsel, has been rewarding the best in-house lawyers around the world for 17 years. Considered one of the most prestigious awards available to in-house lawyers, the European Counsel Awards represent a true standard of excellence in the profession. (…) Although based on thorough and coherent research, the results remain the subjective views of the European Counsel Awards team and the advisory board”. This is how the organization presents its awards. This year, Portugal received three “gold medals”. Pedro Gonçalves, lawyer and Legal and Compliance Director at Unbabel, in Information Technologies; Duarte Abrunhosa e Sousa, from Sonae MB of the ABB team, in the Labor Law category; and Valeria Ferraz, from Central de Cervejas, in the Commercial Law category.

O i interviewed Pedro Gonçalves:

He was nominated, among four lawyers, as the best lawyer in information technologies. What does this mean?

This means that it is a recognition from my peers, at European level, of what I have developed. In this case, in the field of artificial intelligence (AI).

And what has been developed?

I help with all new AI projects and ensure that there are no barriers to the entry of these new projects.

Are we talking in legislative terms?

No, the product itself. A company creates new products and, legally, there are no barriers to their entry and, at the same time, these products assume transparency and effectiveness among the people they are aimed at.

And is this approved by Brussels?

The company is American, but is based in Portugal. The founders are Portuguese. And they created the company in the USA, in California. And my role in the company is to ensure that all procedures, in terms of security, legal and compliance are fulfilled so that the company can develop its activity naturally.

The company, from what I understand, is characterized by accelerating translations.

It is a company that does translations, but it does more than that: it aims, in essence, to avoid language barriers between the various companies. And, for this, it has customer service, people dedicated to supporting each company in its translations… Basically, it prevents language from being a barrier to communication between people.

Give me a concrete example.

In video games, for example. The company ensures that the language is adapted to children, so that it is not a merely literal translation. The translation is done according to the recipient. It takes into account a series of factors, not only human but also cultural. If they are Chinese customers, the translation is adapted to China. The company does AI, but it is not a Google. It also uses the human part. If the translations do not ensure a certain level of quality, the human side intervenes. So, the company is also made up of hundreds of linguists who edit the texts. When the machine doesn’t work, so to speak, linguists come in and improve what the AI ​​did. With this improvement, the AI ​​itself is trained and becomes increasingly effective. For example, if we hire the company and we are in Thailand, the Thai cultural reality is different from the European one. “Good morning”, perhaps, is said differently. And the linguist adapts the intonation, adds two As, etc. That is why there is a language operation: it destroys any linguistic barrier that may exist and promotes the cultural adaptation of the translation that is created.

Do you think a chip will be found? Does a person put a chip in and can translate everything into the brain?

The company is involved in the largest European responsible AI project that aims to create AI systems that are ethically responsible, fair and legal. And within the scope of this project, Halo, there are several companies such as Sonae and Hospital da Luz. This project was born because of people who have amyotrophic lateral sclerosis. People who suffer from this pathology lose their ability to speak. And Unbabel led a project, with the Champalimaud Foundation and the University of Coimbra, in which the algorithm learns the language while the person communicates. When a person loses the ability to communicate, they communicate through brain waves through the machine. The person is asked a question and they respond according to impulses. There is no implanted chip, but there are suction cups that are placed on the body and capture the impulse. And depending on this, the person responds in a certain way. There are also glasses that are placed and the person, through their eyes, can answer the questions asked.

Will a suction cup be able to translate everything in the future?

Yes. Basically, the idea of ​​a project like this is that, in the future, we can talk to ourselves. Our thoughts are stored and we transmit this in a simpler way to a machine.

He was nominated… Did he run or…?

First, I was nominated through people. There is no contest. We are nominated by someone and, through this, they ask for opinions on the work we develop. And we nominate people to evaluate us.

And doesn’t AI come into play there?

[Risos] Not here! It’s a direct question.

And what are the big challenges of AI? And where
is that you enter to avoid
the harm?

The biggest challenge for AI is not to let it dominate us. Great AI models have the ability to learn, they generate content. And the main objective is not to allow it to dominate us. And I think we must have, at a legislative and regulatory level, barriers that allow AI to follow strict ethical standards.

But what are the big challenges? The dangers?

One of the biggest challenges is the manipulation itself. Through images and information that are not true. And I think people don’t know about AI. It is based on statistical probability models. The lower the probability of error, the more correct it is. In the image you draw, in the text you create… The big challenge is to establish limits to the generation of content.

Like this?

Basically, there is specific regulation so that the ethical standards commonly accepted in society are included within AI itself. This has a lot to do with the quality of the data that is collected. To avoid said bias in communication.

But your area doesn’t pass through there?

No. It goes through the center of responsibility we have, but it’s all linked to linguistics.

I didn’t quite understand what the dangers were.

The information provided may be considered valid if it is not valid. You enter a plane and hear that the security door is on the right, but it turns out to be on the left, for example.

Or influence elections.

Exactly. Using AI systems with incorrect information that leads…

So-called fake news.

Yes.

But how was AI used in elections?

It was announced. I don’t know what specifically… Disclosing that an AI system is used is another very important and mandatory aspect under European legislation. Another major challenge is the protection of copyright. AI feeds on information and this information is often protected by copyright. When these AI systems are powered by data, we need to know whether they are protected or not. Another major challenge is defending people’s privacy. Basically, ensuring that the data collected ensures the privacy of the people targeted in the communication made. And in collecting this information and disseminating it. There are mechanisms such as data anonymization. It’s the most effective way to avoid identifying people.

But is this possible to do for those on the dark side of the force?

Yes, you can always do it. This is why it is very important to have effective regulation that limits these biases in the use of AI. Depending on the data we give it, the AI ​​can direct itself to whatever we want.

How can AI collect my data,
for example?

From the outset, we collect data from you and other people.

But fake.

Yes.

How are they introduced into AI?

This data is filtered. There are mechanisms to filter them.

How does the AI ​​receive this information?

We go to Google and ask it to retrieve certain information for a specific purpose.

But how?

It’s in the algorithm. It is mathematically transposed through numbers.

The issue of manipulating elections, creating wars…

Yes! Technology has always been used for two purposes: good and evil. As Kant would say, “Man is a crooked piece of wood”. That’s why he’s corrupt. And as Saint Augustine would say, corruption comes precisely from that: from the corrupted heart. We have this dubious human nature.