An interim report compiled by a U.S. national security panel warns the U.S. government of falling too far behind China and Russia in the AI arms race, while calling for new investments to foster innovation.
Released Wednesday, the November interim report from the U.S. National Security Commission on Artificial Intelligence (NSCAI) advises the U.S. government to get its act together on the development of security- and defence-related AI, lest it fall behind its adversaries, namely China and Russia. Failure to do so would relinquish America’s role as a primary player in AI, while exposing the nation to serious new threats, including a diminishing of U.S. military advantage, unchecked disinformation campaigns, increased cyberattacks, and the erosion of democracy and civil liberties, according to the new report.
“We are concerned that America’s role as the world’s leading innovator is threatened,” wrote commission chairman (and former Google CEO) Eric Schmidt and vice chairman Robert Work in the report’s introduction. “We are concerned that strategic competitors and non-state actors will employ AI to threaten Americans, our allies, and our values.”
The final full report, which will include detailed budget recommendations, won’t be released until next year, but this preliminary version, which will be submitted to the U.S. Secretary of Defence, offered some advice on how the U.S. government should move forward. In summary, the government should invest heavily in AI research and development, increase its use of AI for national security purposes, train, recruit, and maintain AI talent, build upon pre-existing U.S. technologies, and work to foster global cooperation on AI-related matters, according to the report.
To assist with the new report, the NSCAI held a conference Wednesday at the Liaison Washington Hotel in Washington D.C., titled “Strength Through Innovation: The Future of A.I. and U.S. National Security.” The purpose of the conference, which I viewed via livestream, was to discuss the interim report and to kickstart a series of discussions that will lead to the commission’s final report, which will eventually fall into the hands of U.S. Congress.
“We are in a competition,” said Schmidt during his opening remarks. “There’s no question the game is set… and we have to win.” He said the U.S. government “is currently unprepared for the potential of AI,” and that a culture change needs to happen in both the public and private sectors.
In addition to new investments in education, Schmidt said the U.S. needs to expand public and private sponsorship of R&D, work to keep talented researchers inside the U.S., be the first to reach global markets, and develop ancillary technologies like quantum computers and 5G networks. Schmidt said collaborative discussions will also be needed to ensure AI safety, such that AI will do “what we want it to do.” The U.S. would be smart to work with its competitors on this matter, he added.
Christine Fox, an assistant director at Johns Hopkins University Applied Physics Lab, said cultural shifts will be required across many departments, both in the public and private sectors, and that leadership will be key to breaking stubborn bureaucracies resistant to change.
The risks of falling behind in the AI arms race emerged as recurring theme throughout the day.
Lieutenant General John N.T. “Jack” Shanahan, the director of the Joint Artificial Intelligence Centre, talked about the coming shift to “algorithmic warfare” and how “we are going to be shocked by the speed, chaos, and bloodiness” of future combat involving AI. He said humans pitted against machines will have a distinct disadvantage and that it would be incumbent upon the U.S. to avoid this lopsided dynamic on the battlefield. Shanahan commended the authors of the interim report but cautioned that the findings will take some time to implement. “This is a multigenerational problem requiring a multigenerational solution,” he said.
Shanahan heads the Pentagon’s Project Maven—an initiative that seeks to improve drone technology with AI. News that Google will no longer be participating in the program represented a serious setback for the project, but Shanahan said the incident served to expose deeper issues.
He said “employees of these companies see no value in working with the DoD [Department of Defence],” and “we don’t make it easy for them.” Shanahan said the U.S. will not be able to attain the guidelines outlined in the new report without public-private partnerships, which he described as “the very essence of our success as a nation.” And unlike China or Russia, the U.S. government actually takes the time to consider the ethics of militarised AI, he said, in reference to a recently concluded DoD investigation.
Steve Chien, a commissioner of the NSCAI, co-author of the interim report, and research scientist at NASA’s Jet Propulsion Lab, said we’re in the midst of a software revolution and that big advances in hardware are becoming less of an issue. The task at hand, he said, is to create “algorithms of punch and counterpunch.”
Andrew Hallman, the principal executive at the Office of the Director of U.S. National Intelligence, said AI will exert a tremendous influence on American security concerns, including the general speed of operations, identity intelligence (i.e. identifying patterns in the relationships of people and organisations), detecting and defending against “influence operations” (i.e. adversaries who spread false or misleading information or try to influence elections), among other realms. The U.S. will need to “respond to cyber intrusions at machine speed and faster than our adversaries,” said Hallman.
Also speaking at the conference was commission co-chair Robert Work, who said AI can keep Americans secure but only “if we let it.” He said defence and security agencies have to “urgently” accelerate their efforts but warned that the underlying infrastructure at the Department of Defence is “severely” underdeveloped. Work said collaborations should be welcomed, both domestically and internationally, to help solve common problems, including efforts to improve the explainability of AI.
Former secretary of state Henry Kissinger, who spoke at the conference in a kind of fireside chat format, addressed this exact issue—that is, the potential for machine intelligence to operate beyond human comprehension. Kissinger said AI is “bound to change the nature of strategy and warfare” and could also upend the way diplomacy is done. He noted that future engagements involving AI will create a tremendous amount of ambiguity in terms of a country’s ability to understand the nature of a threat and who’s responsible, as an “enemy may not know where the threat came from.”
Kissinger’s fear conjures many different possible scenarios, including nations falsely blaming each other for AI-related attacks, a kind of digital fog-of-war in which no one is even sure what’s happening. This tracks with a recent report finding that AI will increase the risk of nuclear war.
U.S. Senator Chuck Schumer (D-New York) spoke briefly at the conference, saying the U.S. has “not matched the level of commitment” of its adversaries and that “we will rue the day” should the U.S. fall behind China and Russia. Schumer said a discussion draft is currently in development to consider a new branch of the U.S. National Science Foundation. The new agency would fund fundamental research related to AI and other high-tech areas, such as quantum computing and robotics, he said, adding that these grants, amounting to $US100 ($145) billion, would go to universities, companies, and special government agencies.
When the full report is released next year, we’ll see how much money the commission wants the U.S. to spend on artificial intelligence R&D. If this report is any indication, however, it likely won’t be a small amount.