“One of my biggest fears is that we, this industry, this technology, are causing significant harm to society.” Sam Altman renewed his warnings this Tuesday, May 16, during his hearing with US senators. The boss of OpenAI, a company known for having developed the conversational agent ChatGPT, spoke out in favor of regulating artificial intelligence.
Government intervention is “crucial” in order to “limit the risks associated” with this technology, he said. “It is essential that the most powerful artificial intelligence is developed with democratic values […] which means that the leadership of the United States is decisive.” If nothing indicates for the moment that Congress will adopt rules in this direction, the posture of Sam Altman reassured the senators. Many of them have given credit for recognizing the pitfalls of generative artificial intelligence.
This, of which ChatGPT is a perfect example, consists of creating content – text, image, sound or video – of stunning realism after being trained on huge databases. Senator Richard Blumenthal, who chairs the Senate Subcommittee on Privacy, Technology, and the Law, gave a little demonstration during the hearing. He broadcast an audio extract imitating his voice created from an artificial intelligence. “If you were listening from home, you probably thought it was my voice and my words, when it wasn’t my voice or my words,” he said. He then listed the risks of this technology: deepfakes, disinformation, job losses, theft of intellectual property…
Towards a federal AI agency?
Sam Altman confirmed these abuses, which he described as “serious”. “If this technology goes the wrong way, it can go quite far,” he agreed. But the 38-year-old entrepreneur remains convinced that the benefits can outweigh the risks. “Artificial intelligence has the potential to improve just about every aspect of our lives,” he said, hoping that eventually it can enable humanity to solve “its most important challenges. “, like global warming or cancer.
Provided, of course, that it is regulated. The OpenAI boss mentioned the creation of a federal agency responsible for granting authorizations to organizations that develop AI systems of a certain level. It could also withdraw them if the security standards are not respected. “I know it sounds naive to come up with something like that, and it seems very difficult” to achieve, he said. But “there are precedents,” he added, referring to the International Atomic Energy Agency (AEIA).
The idea of creating a federal agency has aroused some enthusiasm, but it would have to be given sufficient resources, underlined Richard Blumenthal. “And I’m not just talking about dollars, I’m talking about scientific expertise.” Sam Altman, however, noted the risks of too much regulation: “If American industry slows down, China or someone else may progress faster.” He also insisted that any measures should not stifle independent research.
Learn
The seduction operation is successful for the boss of OpenAI, who had swapped his usual sweatshirt for a classic suit. The senators were much less offensive than during the hearings of other tech figures, such as Mark Zuckerberg, creator of Facebook, or Jeff Bezos, creator of Amazon. The American Congress also hopes not to reproduce the mistakes of the past. He had proven unable to regulate social networks, causing the failure of several bills on data privacy. He now wants to draw lessons from this experience for artificial intelligence, hearing Sam Altman being considered a first step.
However, the United States remains far from enacting new laws, unlike Europe. The European Parliament is due to vote next month on sweeping AI regulation legislation. Called “IA Act”, the text aims to prevent the risks of this technology, for example by prohibiting software exploiting vulnerabilities due to age or disability. He also wants to better regulate the use of AI in “high risk” sectors such as transport, education or human resources, in particular for the scoring of exams or the sorting of CVs.