Biden administration officials are divided over how aggressively new artificial intelligence tools should be regulated, and the disagreements between the two emerged in Sweden this week.
Some White House and Commerce Department officials support the EU’s proposed strong measures against AI products such as ChatGPT and Dall-E, people familiar with the talks said. Meanwhile, U.S. national security officials and some members of the State Department said aggressive regulation of this nascent technology would put the country at a competitive disadvantage, according to the people, who asked not to be identified because the information has not been made public. I argue that it will put me in a position where
The cacophony has caused the US to respond inconsistently to the EU’s plans to apply additional rules to generative AI at the US-EU Trade Technology Council meeting in Sweden this week. The proposal would allow developers of artificial intelligence tools to document copyrighted material used to train their products and track more closely how that information is used, among many other things. will be forced to comply with the strong regulations of
Commerce Secretary Gina Raimond on Wednesday compared advances in AI technology to social media, noting that in hindsight its development needed more restraint. “The stakes are much higher” in AI, she said during a panel discussion at the TCC conference.
“Just because you can do it doesn’t mean you should,” says Raimondo. “So when it comes to understanding the benefits of AI, I would like to ask all of us to seriously open our eyes to the costs and analyze whether we should adopt AI.”
National Security Council spokesman Adam Hodge said in a statement Tuesday that the Biden administration is not divided and is working across the government to “advance a coherent and comprehensive approach to AI-related risks and opportunities.” said. The United States “has been leading the charge on these issues long before the latest generative AI products,” he said.
How the EU decides to regulate AI is probably more important than the debate in Washington. With Congress unlikely to pass binding rules on AI, how tech giants including Microsoft and Google-owner Alphabet develop foundational models to underpin the next frontier of artificial intelligence. will be decided by the European region for the first time.
main battlefield
Rather than being designed for one specific task, these models rely on training data (often large samples of languages taken from the internet) to determine how they should respond in different situations. to learn This is the technology behind generative AI, which can answer homework questions, design PowerPoints, and create fantastic images from text prompts.
The question for regulators is who is responsible for risks associated with technology, such as dissemination of misinformation and invasion of privacy. The proposed EU regulation would add reporting requirements for companies developing models used in chatbots such as OpenAI.
Michelle Giuda, director of the Crack Technology Diplomacy Institute and former assistant secretary of state for international communications in the Trump administration, said one of the TTC’s fundamental mandates is to foster innovation and stay ahead of the world. to strengthen trust among allies. Chinese progress.
“The context is that AI innovation is not happening in isolation, it’s all happening in the 21st century battle between democracy and authoritarianism,” Giuda said. . “And the main battlefield is technology.”
high risk
Until recently, the U.S. and EU have made loose agreements to regulate applications rather than the technology itself, focusing on high-risk areas such as critical infrastructure and law enforcement.
This approach is embodied in the US non-binding framework for AI systems and the European Commission’s first proposal for AI legislation to regulate this technology. End-use risks were also highlighted at the last board meeting in December.
But with the release of ChatGPT, a broader risk became more apparent. An apparently AI-generated fake image of an explosion near the Pentagon terrorized the US market this month, but the technology has already created corporate winners and losers.
“Europe is important, but this is bigger than Europe,” European Commission Vice-President Marguete Vestager told reporters at the TTC. By working with the United States, she said, “we can all be more comfortable with the fact that generative AI exists in the world and is developing at an alarming rate.” said.
The European Parliament has proposed new rules specifically aimed at underlying models used for generative AI. Commission members agreed earlier this month that more scrutiny of the companies developing these foundational models is needed. Most of these companies, including Microsoft and Google, are based in the United States.
smoldering resentment
It added to already-simmering resentment among tech executives against EU antitrust and content moderation rules that unfairly affect U.S. companies.
The tech industry has criticized the Biden administration for not doing more to defend U.S. companies in the face of what they see as trade discrimination. The EU’s proposed changes could move AI law from a silver lining of cooperation to a new European example of targeting US technology, they warn.
A revised AI law could be voted on by parliament in June, ahead of final negotiations with the 27 EU member states.
“I think our move to also address generative AI is a good move,” said Dragos Tudrash, one of the lead authors of the bill in Congress, after meeting with U.S. officials. Stated.
Some U.S. officials disagreed, warning that restrictions on the foundation model could undermine U.S. competitiveness, according to people involved in the discussions.
OpenAI chief executive Sam Altman has publicized company concerns over regulatory overreach, suggesting the company may withdraw its products from the European market if the rules are too difficult to follow. let me know In a tweet, EU Commissioner Thierry Breton accused Mr. Altmann of an “extortion attempt.”
Altman later said he would try to comply with EU regulations. He is scheduled to meet with Vestager on Wednesday and with Chairman Ursula von der Leyen on Thursday.
European officials consider it inappropriate to bring European democratic processes into multilateral discussions and will discuss the details of the AI law with U.S. officials ahead of the TTC meeting, according to people with direct knowledge of the negotiations. resisting.
The EU is still debating regulation, and some European officials think parliament has gone too far, some people said.
Generative AI will be mentioned in the TTC’s conclusions, according to a draft obtained by Bloomberg. The paper confirms a cross-Atlantic commitment to risk-based approaches, but also highlights the “scale of opportunity and the need to address the associated risks” of generative AI.
