News on foundation models (e.g., ChatGPT)

This news shares recent developments about foundation models and recalls our essay series on the topic with the contribution of cross-sectoral experts: read it here.

1. Germany, France and Italy reached an agreement on future AI regulation

France, Germany, and Italy have agreed on a joint approach to regulating artificial intelligence (AI), aiming to expedite negotiations at the European level. The three countries advocate for "mandatory self-regulation through codes of conduct" for foundational AI models but oppose untested norms. They emphasize that the AI Act should regulate the application of AI, not the technology itself, with the focus on inherent risks in the application of AI systems. Developers of foundational models would need to define model cards providing information about the model's functioning, capabilities, and limits, following best practices. The joint paper suggests the creation of an AI governance body to develop guidelines and monitor model card application. Initially, no sanctions should be imposed, but a system could be established if violations are identified later. Germany's stance is to regulate AI applications, not the technology, to compete globally in AI. The agreement aims to strike a balance between harnessing AI opportunities and limiting risks. This development coincides with global efforts to leverage the economic benefits of AI, as seen in the UK's AI safety summit and Germany's digital summit addressing AI issues. The German and Italian governments will also discuss AI during talks in Berlin.

Does the above threaten to kill the EU’s AI act? Read it here and here.
Is the change of direction of the major European economies the result of extra-EU moves? See here for the UK and here for the US.

2. What is happening to OpenAI (ChatGPT's developer)?

After the co-founder of OpenAI, Sam Altman, was removed as CEO in a tumultuous boardroom coup on Friday, Microsoft (one of OpenAI's biggest stakeholders) has recruited him to enhance its innovation in artificial intelligence. Additionally, the ChatGPT company is set to have its third CEO in as many days. Read it here and here.

Update: Sam Altman Is Back as OpenAI CEO Just Days After Being Fired. The new board will be led by former Salesforce co-CEO Bret Taylor, who also chaired Twitter’s board before its takeover by Elon Musk last year. The other members will be former U.S. Treasury Secretary Larry Summers and Quora CEO Adam D’Angelo. See here.

3. Nello Cristianini, Professor of Artificial Intelligence at the University of Bath and one of the leading experts in the field, gave the following interview on the topic (in Italian) to FBK - Fondazione Bruno Kessler:



  • How machines can be smart (min. 1'00")
  • Can statistical algorithms understand the world? (min. 4'20")
  • Examples of practical applications (min. 6'00")
  • How does GPT work? (min. 8'00")
  • The emerging skills of GPT (min. 11'00")
  • Shortcuts ...and the monkey's paw (min. 15)
  • The risks of AI and its legal regulation (min. 25'00")
  • The importance of the media (min. 31'00")