“We have to find the right balance between regulation and innovation”

OpenAI CEO Sam Altman at the Elysee Palace in Paris on May 23, 2023.

It’s all about “of balance”After having created controversy by worrying about the consequences for ChatGPT of too strict a framework for artificial intelligence (IA) in Europe, Sam Altman held a much more diplomatic language, Friday, May 26, in Paris: “We have to find the right balance between regulation and innovation”, pleaded the CEO of OpenAI, in front of a conquered audience of tech employees and entrepreneurs, gathered at the Parisian start-up incubator Station F. Just before him, the Minister Delegate for Digital, Jean-Noël Barrot, introduced the arrival of ” Sat ” using the same formula. Sign of the shared desire to be courteous.

The tone marks a certain de-escalation compared to remarks reported, Wednesday, May 24, by several British media. The creator of the chatbot ChatGPT said he had ” several ” points of criticism on the AI ​​Act, the European regulation currently under discussion. “If we can comply, we will. Otherwise, we will stop operating in Europe… There are technical limits to what is possible,” he said, according to the press. A strong reaction from Brussels was not long in coming: “Is this blackmail? »European Commissioner Thierry Breton had tweeted.

“The titles of the articles did not really correspond to what I wanted to say”, ured, Friday, Mr. Altman, pleading a form of misunderstanding. “We will continue to operate in Europe”, he said, adding: “We love Europe. »

Another sensitive point, copyright

Beyond the subtleties of tone, Sam Altman agrees with the spirit of the AI ​​Act: “An authorization regime, with safety standards, is entirely relevant, that suits me very well. » But he recalled that the ” details “ mattered and that there was “still a part of vagueness” in the text.

The most recent version, issued by the European Parliament, thus requires publishers of large multipurpose AI models, such as OpenAI, to describe the data on which this software has been trained, to ensure its quality, as well as to ensure that they do not carry the risk of discriminatory bias or the publication of dangerous content. “These text and image databases are gigantic. If we were asked, for example, to be 100% certain that an element was not there, it would be difficult”he pleaded.

Another very sensitive point: copyright. The Parliament calls for the listing of all protected documents and works used to train AI systems. And Mr. Barrot, like the Minister of the Economy, Bruno Le Maire, is in favor of the media, artists or companies holding the rights being remunerated. “The idea of ​​some form of compensation seems reasonable,” Mr. Altman admitted.

You have 47.91% of this article left to read. The following is for subscribers only.

Source link

Leave a Reply