Researcher António Branco, coordinator of the first major generative Artificial Intelligence model for the Portuguese language, considers that this technology constitutes a threat to the sovereignty of States.
“Democratic States must accept this as a geostrategic and sovereignty issue of the utmost importance”, said, in an interview with Lusa, the professor at the Department of Information Technology at the Faculty of Sciences of the University of Lisbon.
At issue is the fact that the databases are in the hands of “two or three large companies”, without public access or external supervision, which can condition the entire functioning of organizations and citizens.
“We are witnessing a reverse effect here to what we sought with the Internet”, which is global and does not allow one user to condition the access of others, in a “completely decentralized logic”.
Now, in the case of AI, “we are doing the opposite, we are funneling it to two or three entities that will be receiving the communications, the conversation in natural language from everyone around the world and returning it processed” in the form of useful applications that get people used to using them.
Because, “when we, human beings, get used to a certain level of technological intermediation, we are no longer able to return to the previous level”, he said.
“This is terrifying in terms of sovereignty, it’s terrifying”, highlighted António Branco.
On the one hand, all it takes is for there to be “a technical problem” with one of these suppliers and “we are unable to talk to each other”, because “the data is all passing through all our conversations, important or not important, ideological or not. ideological, confessional or non-confessional”.
The big technology companies “are keeping copies of our data there and as a result we are dependent on this service”, which “can be cut off at any time”.
This already happens today. “Just think that we have a nation at existential risk, which is Ukraine”, whose “war effort depends on a system of satellites that are turned on or off by a person's decision depending on whether that person thinks the battlefront is going in right direction or the wrong direction.”
“If geopolitics is rearranged”, “access can be conditioned”. If “we are on a side of the board that is different from their side of the board, they say 'cut there' and we are left completely in the dark to talk to each other to communicate with each other,” she explained.
Today “we have a risk of subtracting sovereignty in the not very long term”, highlighted António Branco, who defends the emergence of more AI tools, with different databases.
The Portuguese State has two AI-based Chat services, one about divorce and the other about the digital mobile key, which are provided by OpenAI.
“No one knows” what data is collected by OpenAI and that is why the “Portuguese public administration is looking for conditions to free itself from this dependence on a single supplier, but this has costs”, he warned.
“Democratizing use and access is the antidote” and “the more entities that offer these services, the less we will pass on our data alone” through big technology companies.
Furthermore, each country must keep the information it needs within its borders. “This is an issue of the greatest relevance for democratic States”, highlighted the researcher, who gave some examples.
“The United States itself is investing, from the point of view of public money, tens of billions of dollars, despite the 'big techs' being based there”, said António Branco, who also highlighted the United Kingdom in this national effort .
The large companies that manage generative language models, Google or OpenAI, are “opaque” and “it is not known what type of information they generate”, but this problem becomes even more serious in non-democratic states.
“There are three different worlds: there is us here, in the West, there is China and there is Russia”, he recalled.
Comments