It was supposed to be a week-long presidential trip demonstrating the United States’ commitment to the Asia-Pacific region, but the debt ceiling crisis left by President Joe Biden in Washington forced him to cancel the second half of his itinerary.

The planned summit in Sydney of the leaders of four countries took place: the USA, Australia, India and Japan. Rejected, a meeting of Pacific Island leaders in Port Moresby, the capital of Papua New Guinea, and an announcement by the president of an agreement to give the US military access to the island nation’s ports and airports.

For some, the derailed and shortened tour will only heighten fears that a weakened America, distracted by political divisions at home, may be unable to lead the Indo-Pacific as it confronts an increasingly assertive China.

“The presence matters to all US allies in the region, so yes, canceling the second phase of President Biden’s visit to Asia will cause some frustration and raise some questions,” says Nicholas Széchenyi of the Center for Strategic and International Studies in Washington.

“But this could still be a temporary blip on the radar screen,” he adds, “if the administration sticks to a very clear agenda and the vast and multi-dimensional network it has established across the region.”

When President Joe Biden meets with his G-7 colleagues in Hiroshima, Japan this weekend, he brings up an agenda of burning issues, from increased Western support for Ukraine to international regulation of artificial intelligence.

On Friday, the leaders of the host nation, the United States, Germany, France, Britain, Canada and Italy announced new economic support for Ukraine and another round of sanctions against Russia for its “illegal, unjustified and unprovoked aggression against Ukraine.”

Mr. Biden told G-7 leaders that the US is now supporting the training of Ukrainian pilots on US-made F-16 jets, senior officials told reporters on condition of anonymity. The initiative received support in Europe.

It was also announced on Friday that President Volodymyr Zelensky would attend the closing of the summit on Sunday, another stop on the Ukrainian leader’s fast-paced diplomatic tour, which included a Saudi appeal Friday for support to members of the Arab League.

But in Asia, Mr. Biden, in addition to his short-term political agenda, is also pursuing two key foreign policy directions of his presidency: revitalizing America’s alliances and demonstrating the virtues of democracy as an effective system of governance in an age of progressive authoritarianism.

In February 2019, OpenAI, a little-known artificial intelligence company, announced that its large language model text generator, GPT-2, would not be released to the public “due to our concerns about malicious applications of this technology.” The company said that among the dangers was the possibility of misleading news articles, online impersonation, and automating the production of offensive or fake social media content, as well as spam and phishing content. As a consequence, Open AI suggested that “governments should consider expanding or launching initiatives to more frequently monitor the social impact and diffusion of AI technologies, and to measure progress in the capabilities of such systems.”

This week, four years after the warning, members of the Senate Judiciary Subcommittee on Privacy, Technology, and the Law met for a discussion on AI Oversight: Rules for Artificial Intelligence. As with other tech hearings on the Hill, this took place after a new technology that could revolutionize our social and political lives was already in circulation. Like many Americans, lawmakers became concerned about the pitfalls of the AI ​​model in March when OpenAI released GPT-4, the latest and most advanced version of its text generator. At the same time, the company added it to a chatbot it launched in November that used GPT to answer questions in a conversational manner with a confidence that isn’t always warranted because GPT tends to make things up.

Despite this volatility, in just two months ChatGPT became the fastest growing consumer app in history, reaching one hundred million monthly users by the start of this year. It has over a billion page hits every month. OpenAI has also released DALL-E, an image generator that creates original images from a descriptive word prompt. Like GPT, DALL-E and other text-to-image tools can blur the line between reality and fiction, making us more susceptible to deception. Recently, the Republican Party released the first fully created by AI attack advertising; it shows what appears to be a genuinely dystopian portrayal of the second term of the Biden administration.

Three experts took part in the Senate hearing: Sam Altman, CEO of OpenAI; Christina Montgomery, IBM Director of Privacy and Trust; and Gary Marcus, NYU professor emeritus and AI entrepreneur. But Altman attracted the most attention. He was the head of the company with the most popular tech product that could revolutionize how business is done, how students learn, how art is made, and how humans and machines interact—and what he told senators: “OpenAI believes AI regulation is essential.” “. He aims, he wrote in his prepared testimony, “to help policy makers as they determine how to facilitate regulation that balances security incentives while ensuring that people can access the benefits of technology.”

Senator Dick Durbin of Illinois called the hearing “historic” because he couldn’t recall executives going before lawmakers and “begging” them to regulate their products, but it wasn’t actually the first time a tech CEO had sat in the congressional hearing room and called for more regulation. In particular, in 2018, in the aftermath of the Cambridge Analytica scandal — when Facebook gave a Trump-linked political consulting firm access to the personal information of nearly ninety million users without their knowledge — Facebook CEO Mark Zuckerberg said. some of the same senators that he was open to more government oversight, a position he reiterated the following year, writing in Washington mail“I believe we need a more active role for governments and regulators.” (At the same time, Facebook was paying lobbyists millions of dollars a year to stave off government regulation.)

Like Zuckerberg, Altman began his call for more regulation by explaining the barriers his company is already using, such as training its models to reject certain kinds of “anti-social” requests — like the one I recently asked ChatGPT when I asked it to write the code. for 3D printing Glock. (However, they wrote a script for a 3D printed slingshot. “I would like to emphasize that the creation and use of this device must be done responsibly and legally,” the message says before issuing the code. ) OpenAI usage policies also ban people from using their models to create malware, create child sexual abuse images, plagiarize or produce political campaign material, among other things, although it’s not clear how the company plans to enforce them. “If we find that your product or its use does not comply with these policies, we may ask you to make the necessary changes,” the policy says, effectively assuming that in many cases OpenAI will act after the violation has occurred rather than prevent his. This.

In an opening statement at the hearing, the chairman of the subcommittee, Senator Richard Blumenthal of Connecticut, was merciless. “Artificial intelligence companies should be required to test their systems, disclose known risks, and grant access to independent researchers,” he said. He added: “When AI companies and their clients cause harm, they must be held accountable.” To demonstrate his view of the harm, Blumenthal presented his remarks with a recording of him talking about the need for regulation, but these were words he never uttered. Both “his” voice and “his” statement were fabricated by artificial intelligence. The consequences, especially for the politicians in the room, were frightening.

Figuring out how to assess damages or determine liability can be as tricky as figuring out how to regulate technology that moves so fast that it inadvertently breaks everything in its path. Altman, in his testimony, suggested Congress’s idea. the creation of a new government agency tasked with licensing what he called “powerful” AI models (although it’s not clear how that word would be defined in practice). While this may not sound like a good idea at first glance, it has the potential to be self-serving. As Clem Delang, CEO of AI startup Hugging Face, said, tweeted, “License Requirement for Model Training…. . further concentration of power in the hands of a few. In the case of OpenAI, which was able to develop its large language models without government oversight or other regulatory restrictions, this would allow the company to significantly outperform its competitors and solidify its “first to publish” position while limiting new opportunities. people on the field.

If that were to happen, it would not only give companies like OpenAI and Microsoft (which uses GPT-4 in a number of their products, including the Bing search engine) an economic advantage, but would further undermine the free flow of information and ideas. Gary Marcus, AI professor and entrepreneur, told senators that “there is a real risk of a kind of technocracy coupled with an oligarchy, with a small number of companies influencing people’s beliefs” and “doing this with data we don’t even know” . know about.” He was referring to the fact that OpenAI and other companies have kept the data on which their large language models were trained secret, making it impossible to determine their inherent biases or truly assess their security.

Senator Josh Hawley of Missouri noted that the most immediate danger of LLMs such as ChatGPT lies in their ability to manipulate voters. “This is one of the areas that worries me the most,” Altman told him. “The more general ability of these models to manipulate, to persuade, to provide one-on-one interactive disinformation,” he said, “with elections coming up next year and these models getting better, I think this is a major problem.”

The most viable way to fix this problem would be for OpenAI to take the lead and take their LLMs off the market until they no longer have the ability to manipulate voters, spread misinformation, or otherwise undermine the democratic process. It would indeed be, in the words of Senator Durbin, “historic.” But it was not proposed in the hearing room. Instead, much of the discussion has focused on what kind of regulatory agency, if any, could be created and who should serve in it, and whether such an agency could be made international. It was an exciting exercise in the future that ignored the real danger. Senator Blumenthal told his colleagues, “Now Congress has a choice. We had the same choice when faced with social media. We failed to take advantage of this moment.” With elections approaching and this technology in the game, the predictive power of artificial intelligence is not needed to admit that lawmakers, despite their curiosity and bipartisan civility, missed this moment too. ♦