“My great-grandparents were Italian, but no one in my family ever taught me the language. Therefore, I apologize if I speak English during this speech“. It starts like this Michael Sellittohead of global affairs in the company artificial intelligence company Anthropic, founded Italian American brothers Dario and Daniela Amodeiformer OpenAI executives, during his hearing at the conference Supervisory commission for documentation activities on the subject of artificial intelligence. The meeting took place (via video link) Tuesday, January 9, 2024 in the chamber of the Foreign Affairs Committee of the Chamber of Deputies.
As explained on the official website, the committee “aims to learn about the progress of artificial intelligence and algorithms and study their possible use in parliamentary documentation” and more generally wants ask about AI-related opportunities and risks. The hearing of the anthropic exponent is part of a cycle of meetings that began in 2023 and is chaired by the Deputy Speaker of the Chamber of Deputies Anna Ascaniová which, as he recalls Political certificatelast October he presented a bill “ensure transparency in the publication and dissemination of content created using artificial intelligence systems“.
For Sellitt, the main goals of Anthropic are “educate, educate politics and public opinion on trends in artificial intelligence with the hope that our work will lead to regulation of this technology“. Technology in full development in which, according to Sellitt, we must continue to invest. “In the next two years the capabilities of current models will increase considerably, which has both positive and negative consequences“.
Among the negative aspects related to the development of AI Sellitto includes power imbalance in the private sector compared to the public sector: “Until 2015, the academic world was largely concerned with the analysis of artificial intelligence. Today, its contribution is almost nil due to the massive entry of the private sector into the field“.
On the role that AI can play in facilitating parliamentary work – the Commission’s ultimate goal – Sellitto explains that Anthropic’s Claude model “can analyze very large textssuch as rewriting legislative processes and automating synthesis“. A useful model for female and male knowledge workers as well with the intention of conducting research. “In short, what we want is to speed up and simplify people’s work“.
The Commission’s questions focused primarily on concerns it has to do with the development of artificial intelligence and not so much its possibilities, which Sellitto seemed more willing to discuss. As for the dangers, the expert said: “This technology can improve almost everything humans can do. So it’s about positive and negative actions“. However, something can be done: “In the fight against the use of artificial intelligence in the field of terrorism, some studies are underway on the ways in which the models can be used to carry out physical or cyber attacks. Thus it is possible to intervene modifying the model itself to prevent such use“.
Not only terrorism: concerns about the use of AI also arise with regard to the important ones elections to be held in 2024 and as far as copyright protection.
As for the election rounds, Ascani specifically refers to the possible use of technology to spread false information and deepfake ahead of June’s European elections and November’s US presidential election. “I hope soor close cooperation between governments, industry and civil society stop the production of this content. We further agree to ask governments to consider watermarking content generated using artificial intelligence models”, says Sellitto. Regarding copyright protection and artificial intelligence, Sellitto concludes by stating “we want to cooperate with the European Union under the AI Act“.