Enlarge /. Close-up of an unscrewed Predator MQ-9 aircraft.
US representatives Will Hurd and Robin Kelly are on opposite sides of the ever-increasing corridor, but share concerns that the United States will no longer have control over artificial intelligence and could threaten the American economy and the balance of world power.
On Thursday, Hurd (R-Tex.) And Kelly (D-Ill.) Made proposals to prevent the US from falling behind China, especially when it came to using AI on defense and national security. They want to block China's access to AI-specific silicon chips and urge Congress and the federal authorities to provide more resources for the further development and safe use of AI technology.
Although Capitol Hill is increasingly divided, the bipartisan duo claim that a consensus is emerging that China is a serious threat and that support for US technology development is an important tool.
"American leadership and advanced technology have been critical to our success since World War II, and we are racing against the Chinese government," said Hurd. "It is time for Congress to play its part."
Kelly, a member of Congress' Black Caucus, says she has found many Republicans, not just Hurd, the only black Republican in the house who is open to technical cooperation. "I think people in Congress now understand that we have to do more than before," she says.
The Pentagon's national defense strategy, updated in 2018, says AI will be the key to being one step ahead of rivals like China and Russia. Thursday's report includes recommendations on how Congress and the Pentagon should support and use technology in areas such as autonomous military vehicles. It was written in collaboration with the Bipartisan Policy Center and the Georgetown Center for Security and Emerging Technology, in which experts from government, industry and science were consulted.
According to the report, the U.S. should work more closely with its allies on AI development and standards while restricting technology exports such as new computer chips to China to drive machine learning. This hardware has recently made many advances from leading business laboratories like Google possible. The report also urges the federal authorities to provide more money and computing power to support AI development in government, industry, and science. The Pentagon is asked to think about how war courts deal with liability issues when autonomous systems are used in war and to talk more about its commitment to ethical use of AI.
Hurd and Kelly say that military AI is so powerful that America should do some kind of AI diplomacy to prevent dangerous misunderstandings. One of the 25 recommendations in the report is that the U.S. is setting up AI-specific communications with China and Russia so that human-to-human dialogue can mitigate accidental escalation caused by algorithms. The proposal is reminiscent of the Moscow-Washington hotline, which was established in 1963 during the Cold War. "Imagine yourself in a high-stakes problem: What does a Cuban crisis look like with the use of AI?" asks Hurd, who will be withdrawing from Congress at the end of the year.
Cut the hype
Aside from such worst-case scenarios, the report contains more sober ideas that could help reduce the hype surrounding military AI and killer robots. It urges the Pentagon to do more to test the robustness of technologies such as machine learning, which can fail in unpredictable ways in rapidly changing situations like a battlefield. Intelligence agencies and the military should focus AI efforts on back office and non-critical purposes until reliability improves, the report said. This could predict fat new contracts with leading computer companies like Amazon, Microsoft and Google.
Helen Toner, strategy director at Georgetown Center, says that although the Pentagon and intelligence agencies are trying to build reliable and responsible AI systems, "the question is whether they will have the capability or institutional support." Congress funding and supervision would help them get it right, she says.
The paper released on Thursday is the second of four papers on AI strategy published by Hurd and Kelly from the Bipartisan Policy Center. The first, released earlier this month, focused on the workplace. His recommendations included revising education from kindergarten to graduate school to prepare more Americans to work with or at AI. The next two articles deal with AI research and development as well as AI ethics.
Kelly and Hurd have had a common interest in AI since working on hearings from the House Oversight Committee's Information Technology Subcommittee in 2018 on the subject. The couple later wrote a report warning that the U.S. could lose its leading position in AI. Kelly says she wants to make sure that the United States remains the leader in AI, but also that "people in the diverse district I come from have a piece of this cake and that there are no prejudices or concerns about them regarding privacy. "
At the end of the Obama administration, the White House produced detailed documents on how to support the development and use of US AI and address potential drawbacks such as technological unemployment. The Trump administration decided not to build on it, but last year President Trump signed an implementing regulation instructing existing government programs to target AI projects. The US has a less toothy AI strategy than many other nations, including China, that have launched new programs and funding sources. Hurd and Kelly try to change that.
James Lewis, who heads technology policy work at the Center for Strategic and International Studies, welcomes the efforts. The timing is good, he says, because more legislators are interested in technology policy. "They are now realizing that we are in competition with China and have become aware of the fact that technologies like AI, semiconductors and cyber security are important," he says. Last week, the Senate voted 96 to 4 to change the $ 25 billion annual Pentagon budget to support domestic research and manufacturing of new chip technologies.
Lewis supports the limitation of chip exports to China – an idea that could resonate at a congress that shows new interest in technical export controls. He is skeptical that an AI hotline or the invention of special forms of AI diplomacy to prevent autonomous accidents is worthwhile. Events during the Cold War and since then, most recently in areas such as cybersecurity, suggest that China and Russia are not taking such programs seriously, Lewis says.
Hurd and Kelly are currently drafting a Congress resolution that includes their ideas on AI, including national security. After that, they will start working on AI legislation. "I hope some of them are finished at this congress, and others can be adopted and carried out at the next congress," says Hurd.
This story first appeared on wired.com.