The President of the Government, Pedro Sánchez, announced on Sunday at the welcome dinner of the Mobile World Congress (MWC), the construction of a foundational model of artificial intelligence language, trained in Spanish and co-official languages, in open and transparent source, and with the intention of incorporating the Ibero-American countries.
According to Albert Agustinoy, partner of Intellectual Property and Digital Technologies and Media at Cuatrecasas, this announcement responds “more to the aspiration – also expressed by the Ministry of Industry – to establish tools by public administrations that help citizens in carrying out of certain procedures. I do not see Spain going from 0 to 100 in a short time and, from the outset, being used for the granting of benefits, for example. And I don't see large companies or banks embracing this model either.” In this sense, “it would be like when the electronic signature was also promoted by the State administration,” this lawyer recalls.
For Andreas Kaltenbrunner, lead researcher of the AI and Data for Society group at the UOC, “it will not be an easy task due to the amount of resources that competitors such as OpenAI have and it remains to be seen if focusing only on a reduced set of languages does not lose the benefits.” possible synergies,” according to what he said in statements to Science Media Center Spain.
Sánchez also assured that the aspiration is for “Spain to play a relevant role at the European and international level in the development of artificial intelligence.” This is the commitment of other countries. For example, that of France and its president, Emannuel Macron, who this summer assured CNBC that his government will invest “like crazy in training and research” in AI to create two or three “large global players” in this technology. The French approach is different from the Spanish one, since it is based on the administration contributing to the creation of companies that can compete with those in the US.
For Josep Curto, professor of Computer Science, Multimedia and Telecommunications Studies at the UOC, in the Spanish case, since it is “the Government that pushes such creation, it must comply by default with the obligations indicated by the European AI law. On the other hand, copyright, publisher or licensee will surely take into account the rights of the author to exploit the reference sources that, as we well know, have not taken into account some of the most relevant foundational models.”
In this sense, for Kaltenbrunner “another very positive aspect of [Sanchez's] announcement is the focus on using open and transparent source. This will allow us to have greater control over the training data and thus mitigate potential negative aspects.” This is important to avoid mistakes being made, as happened in the Netherlands in 2021, when an algorithm falsely accused 26,000 immigrant families of defrauding the State. Agustinoy agrees, believing that the public push should make it possible to “better parameterize the neural system and avoid cases like the Dutch one.”
Outside the European context, other countries are promoting the creation of their own language models or beginning to think about usage standards for automated decision making. One of them is Canada, whose government is exploring the uses of AI for tasks that involve a low level of risk, as confirmed by the Canada Revenue Agency to Radio Canada.
For its part, India is promoting its own AI solutions to avoid technological colonization. In September 2023, the government launched a bot that works with PM-Kisan, the Indian direct benefit transfer program for farmers, to provide them with financial assistance. This allows them to know if they meet the requirements or know the status of their application just by using their voice. India is a country where more than 600 languages are spoken. On the day of its launch, more than 500,000 users used it.
