In anticipation of the Artificial Intelligence Summit to be held in Paris in February 2025, the Parliamentary Office for the Evaluation of Scientific and Technological Choices has produced a report recommending 18 measures to support and control the explosion of AI in our daily lives.
How to encourage the development of artificial intelligence while controlling its effects? How can we regulate artificial intelligence, whose expansion is as worrying as it is seductive and even fascinating? Vast and delicate subjects on which the Parliamentary Office for the Evaluation of Scientific and Technological Choices -OPECST-, seized in July 2023 by the offices of the National Assembly and the Senate. Title “ChatGPT, and after? Assessment and perspectives of artificial intelligence”, The OPECST report was presented this December 4th by three rapporteurs, Alexandre Sabatou, Patrick Chaize and Corinne Narassiguin (*) with 18 recommendations, including 5 to support at the AI Action Summit to be held in Paris on February 10-11, 2025.
Ambitious, this report is the result of significant investigative work with the world's best experts on the subject. "The rapporteurs studied the different AI models and the details of how they work, identified the political, economic, societal, cultural and scientific issues that they raise, compared the French national AI strategy to nearly 20 other strategies, six in the European Union and eleven in the rest of the world, and analyzed around ten of global governance projects "artificial intelligence," explained the rapporteurs when presenting their work to the press.
18 recommendations, including 5 essential ones
Since 2015, artificial intelligence (AI) has seen major advances, particularly with the development of models based on deep learning, such as GPT-3 and GPT-4, and generative tools (images, videos, texts). These innovations have disrupted sectors such as health, transport and even creation, with each time the emergence of advantages and disadvantages. Regulation and ethics are therefore essential in the face of challenges related to transparency, data sovereignty and socio-economic impacts.It is in this context that the rapporteurs have mentioned 18 regulation tracks including 5 are considered a priority and must absolutely, according to the rapporteurs, be supported on the occasion of the AI Action Summitwhich will therefore be held in Paris in February 2025.
Anticipate needs and constraints at a global level
The OPECST report advocates a cross-cutting approach, involving actions at global, European and national levels. The emphasis is on governance, regulation and education, while insisting on the strategic autonomy of France and Europe.. Close coordination between States, parliamentarians, and economic actors is strongly recommended and should make it possible to meet the challenges posed by AI, while maximizing its societal and technological benefits. In summary, promoting the development of AI while maintaining control…
To get to the point, Here are the most significant measures detailed by the rapporteurs during the presentation of the report to the press:
- Prescribing a cross-cutting approach to AI issues:
“AI should not be limited to a risk-based vision,” the report insists. A solemn declaration by participants is expected to broaden the summit’s themes to include the following topics in the discussion:
– Education and culture: promote learning and understanding of AI technologies.
– Digital sovereignty and general interest: guarantee technological autonomy and control of strategic infrastructures.
- Streamlining Global AI Governance :
Propose that the UN oversee global AI governance to avoid the proliferation of fragmented initiatives.
- Establish a comprehensive regulatory framework :
Take inspiration from OECD and EU multidimensional approaches to regulate the entire AI value chain.
- Launch a European AI cooperation programme :
Involve countries sharing a common vision (France, Germany, Italy, Spain) to strengthen collaboration and innovation in this area.
Preparing France for the global challenge of AI
Concerning France in particular, the rapporteurs detailed 13 measures to be implemented to both develop AI and allow all French people, individuals and professionals, to use it without fear:
- Developing an autonomous European sector
Aiming for technological independence, coordinating national and European efforts to build a complete value chain.
- Structuring an effective public policy
Set clear objectives, concrete means and monitoring mechanisms to make the more operational national strategy.
- Strengthening strategic governance at the state level
Involve multiple ministries in centralized coordination to effectively drive initiatives.
- Training the population in AI.
– Schoolchildren, workers and the general public: launch learning programs inspired by models like the one from Finland.
– Continuing education: supporting changes in workl related to generative AI.
- Organize a social dialogue on AI.
Involve businesses and social partners in tripartite discussions to anticipate impacts on employment and reduce resistance.
- Mobilizing the French AI ecosystem.
Inspired by international initiatives (NL AI Coalition), coordinate public and private actors at regional and national levels.
- Restarting trusted AI projects.
Renew initiatives like “Confiance.ai” to ensure a safe and ethical adoption in critical sectors.
- Support public research.
Encourage transdisciplinary research, integrating various paradigms (symbolic AI, connectionist) and innovative technologies (Mixture of Experts, Trees of Thoughts).
- Meeting the challenge of standardization
Strengthening French participation in international standardization committees to defend national and European interests.
- Ensure sovereign control of cultural data.
Create French-language databases to counter the cultural standardization induced by mainly Anglo-Saxon models.
- Reform intellectual property rights.
Adapting copyright to the specificities of generative models, protecting both training data and creators.
- Entrust monitoring of AI policy to OPECST.
Ensure regular evaluation by this body, at the crossroads of the scientific and political worlds.
An ambitious program, a big gap also between the proponents of non-regulation of AI and those who, on the contrary, want to curb its development to better control its effects. Not to mention Donald Trump who declared at the end of November that he will appoint at his side a "Tsar of Artificial Intelligence" who we sense is extremely liberal...
This legitimate list of recommendations drawn up by OPECST will probably require updates, but it already appears to be a serious basis for reflection for the AI Summit next February and an encouragement to quickly get into the race in a constructive manner.
* Alexandre Sabatou, senator of Oise (RN), Patrick Chaize, senator of Ain (Les Républicains), and Corinne Narassiguin, senator of Seine-Saint-Denis. (PS)
