Rregulating artificial intelligence (AI) is no longer an option. It is a necessity. The spectacular success of the ChatGPT conversational robot, launched at the end of 2022 by the OpenAI company, introduced the general public to a sample of the range of possibilities offered by this technology… for better or for worse. One of the main merits of the initiative has been to accelerate awareness of the imperative to set a framework for the development of AI. The question now is to know what form this regulation should take and how it can be applied at the global level.
The initiative of the British Prime Minister, Rishi Sunak, on Thursday June 8, to organize in the coming months a first world summit devoted to AI in the United Kingdom, shows that the subject is no longer only technological, but highly political. . Taking advantage of a meeting with US President Joe Biden, Mr. Sunak set a date so as not to find himself isolated at a time when, each in their own way, the European Union (EU) and the United States are showing their desire to frame AI.
As is often the case in terms of regulation, Brussels has taken the lead. A regulation should be adopted by the European Parliament as soon as Wednesday, June 14, before being the subject of negotiations with the Council of the EU and the European Commission to reach a consensus within a few months. The American approach is more limited at this stage, relying on a corporate responsibility approach.
As the EU and US try to harmonize their positions in the Transatlantic Trade and Technology Council, warnings about the dangers of AI are mounting. After a letter signed in March by researchers and tech personalities calling for a ” break “, a petition, bringing together 350 personalities from the AI sectorcalled, on May 30, to make the subject “a global priority, alongside other large-scale risks such as pandemics or nuclear war”. This initiative is itself the continuation of the offensive launched in particular by OpenAI in favor of the creation of a new regulation “global” inspired of the International Atomic Energy Agency (IAEA), whose mission would be to ensure the safety and non-proliferation of the most advanced and dangerous AI systems.
The EU on the right track
These hypothetical and remote risks should not be ignored. However, the establishment of such a body promises to be long and complicated politically, while China does not seem willing to adopt a multilateral approach on the subject. Above all, this idea should not serve as a distraction from a much more immediate and prosaic job of regulating AI as it exists today and not as it is imagined tomorrow.
From this point of view, the path taken by the EU seems the right one. It is a question of prohibiting certain uses, of subjecting others to obligations of transparency on the data and evaluation of the risks, in particular of discrimination. Europeans are also concerned about the copyright attached to the data used by software and want to impose a label on the content generated by AI in order to identify it.
Our continent will also have to integrate the logic of sovereignty, mastery of technology – in particular thanks to open source – and respect for the diversity of languages. Let’s not wait for the creation of a supposed “IAEA of AI” save humanity from its self-destruction to tackle much more immediate and burning subjects.
The world