"It is essential that the most powerful AI is developed with democratic values," the young entrepreneur told a US parliamentary committee, "which means that the leadership of the United States is decisive."

ChatGPT's launch in November sparked immense public and business interest in so-called generative artificial intelligence, i.e. one capable of creating content – text, images, sounds or video – after being trained on huge databases.

"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," said Sen. Richard Blumenthal, who chairs the Senate Subcommittee on Privacy, Technology and the Law.

He had just made his opening remarks about the dangers of AI, written with ChatGPT and read by software trained on his real voice.

AI technologies "are no longer science fiction fantasies, they are real and present," he added.

The elected officials debated with the boss of OpenAi and two other experts the need to regulate computer systems that could "literally destroy our lives", in the words of Senator Lindsey Graham.

Federal Agency

Altman recalled that while OpenAI LP, the entity that developed ChatGPT, is a private company, it is controlled by a non-profit organization, which "requires us to work towards the wide distribution of AI benefits and maximize the security of AI-based systems," he added.

He proposed the creation of a new federal agency, responsible for granting permissions to organizations that develop AI systems of a certain level, and which "could withdraw them if they do not comply with security standards".

"Would you be qualified (for such a position)?" asked Senator John Kennedy. "I love my current job," the entrepreneur replied.

Sam Altman has regularly expressed support for the establishment of a regulatory framework for artificial intelligence, preferably at the international level.

"I know it seems naïve to propose something like this, it seems very difficult" to achieve, but "there are precedents," he said, citing the example of the International Atomic Energy Agency (IAEA).

There has been some enthusiasm for creating a federal agency, but it would need to be adequately resourced, Blumenthal said. "And I'm not just talking about dollars, I'm talking about scientific expertise," he said.

Altman noted, however, the risks of too much regulation: "If U.S. industry slows down, China or someone else can move faster."

He also insisted that possible measures should not stifle independent research, and instead focus on dominant companies like his own.

"Bad sense"

For now, the United States is far from following the path of the European Parliament, which is due to vote next month on new legislation to regulate AI.

The U.S. Congress regularly raises the need to better protect data privacy and foster more competition on the Internet. But political divisions have blocked most bills for years.

The rapid deployment of generative AI by OpenAI, Microsoft and Google has revived the issue of technology regulation.

Many are concerned about its potential impact on professionals, from the risk of job losses to intellectual property theft.

Senators reviewed many areas, such as biased algorithms and the spread of increasingly sophisticated misinformation during election periods.

"Artificial intelligence has the potential to improve just about every aspect of our lives," Altman said, hoping that eventually generative AI will enable humanity to solve "its most important challenges," such as global warming or cancer.

"But it also creates serious risks," he acknowledged.

"One of my biggest fears is that we, this industry, this technology, will cause significant damage to society," he said. "If this technology goes in the wrong direction, it can go quite far. (...) And we want to work with the government to prevent that from happening."

© 2023 AFP