美联合人工智能中心主任强调2025年前做好AI准备

美防务快讯3月10日讯

美国防部人工智能中心主任迈克尔·格伦在参加10日的视频研讨会时认为,人工智能(AI)已不是人们常说的未来事情,AI现在已经到来,要想未来拥有AI,必须从现在做起。格伦称,他正在国防部推行和集成人工智能,目的是加速推进转型,但就像让一艘大型军舰转向一样,挑战巨大。美国人工智能国家安全委员会成员凯瑟琳娜·麦克法兰也参加了会议,两人对美国到2025年前做好人工智能准备必须抓紧的重点工作都发表了看法。麦克法兰指出,该委员会经过调查研究后一致认为,一是美国政府目前尚未有效组织或聚集资源去赢得与对手的竞争,二是到2025年前,美国必须做好人工智能的准备。格伦认为,加强政府与私营企业和学术界的联合在人工智能发展十分重要,也要重视如何确保人工智能应用过程与数据的安全,尤其是嵌入平台的安全。据报道,3月12日,美国众议院武装力量委员会将举行人工智能国家安全委员会成员的听证会。


Get US ‘AI Ready’ By 2025: JAIC’s Lt. Gen. Groen, NSCAI’s McFarland

Approximately half of NSCAI’s final recommendations focus on defense. “We’re already working on two-thirds of them,” Groen said, adding that partnerships with the U.S. domestic private sector and academia are important in pursuing JAIC’s defense-focused AI goals.

By BRAD D. WILLIAMS on March 10, 2021 BreakingDefense

Then-Brig. Gen. Michael Groen works on a Marine Corps intelligence-sharing system.

WASHINGTON: What’s the most significant challenge to getting the Defense Department ready for AI? It’s the idea held by many that “AI is some future thing,” said Lt. Gen. Michael Groen, head of the Joint Artificial Intelligence Center. AI is already here, he said. “If we want AI to be our future, then AI has to be our present.”

Groen currently leads to the DoD’s JAIC, which he characterized as a “do tank” in a town of think tanks. JAIC currently leads implementation and integration of AI across all departments of the DoD. Groen admitted it’s “an enormous challenge” and that it “takes time to turn this big defense ship.” One of JAIC’s goals is to “get the transformation to occur faster, accelerate it.”

Groen’s comments came during a virtual event on Wednesday. Katharina McFarland, a member of the National Security Commission on Artificial Intelligence, was the other panelist.

In a wide-ranging discussion, Groen and McFarland talked about what’s necessary to get the nation “AI ready” by 2025, which was a key finding in the NSCAI’s recently released final report.

Approximately half of NSCAI’s final recommendations focus on defense. “We’re already working on two-thirds of them,” Groen said, adding that partnerships with the U.S. domestic private sector and academia are important in pursuing JAIC’s defense-focused AI goals.

To this last point, Groen talked about the “tech inversion” that has happened within defense. Historically, DoD funded and built much of its own tech under “big programs.” Today, the private sector is leading in the development of AI solutions. While DoD still has a “big programs” approach to some tech, AI is different, Groen noted.

McFarland agreed, noting the “building, testing, and then expanding” traditional approach to DoD tech and contrasting that to AI, which she characterized as “an enabler, a thread that passes through other systems.” The goal should be to “embed AI opportunities space” at DoD, she said.

Groen also talked about the idea of DoD as an enterprise. DoD is “famously oriented” around services, who build their own things, he said. “We have great joint warfighting concepts, but there’s still a culture of individual service development.” However, Groen said he thinks the culture of individual service development is shifting.

Asked about tech challenges specifically at DoD, Groen said, “We have to appreciate the scope of the problem sets.” He noted, across the DoD, some know disciplines and processes, such as artillery, really well. Others know the tech really well. The question is how to bring the two together — process and AI. Groen said DoD has to “get process owners to acknowledge AI is not tomorrow’s technology. It’s today’s technology.” He added, “We need people to think across all domains about tech.”

Two other prominent themes were security and ethics. Groen said he constantly thinks about “AI security of processes, security of our data.” He added, “What keeps me up at night is, how do we secure this? We spend a lot of time working on AI security, especially for our platforms.”

On the ethics front, Groen highlighted the ethical baselines that have been adopted to date by the U.S. and 13 likeminded nations, who want to build AI systems based on ethical considerations. “Our strength is ethical ecosystems,” he said.

McFarland highlighted some findings of NSCAI’s work, which was chartered in the Fiscal Year 2019 National Defense Authorization Act. NSCAI began its work by identifying the AI “playing field,” which includes narrow intelligence and general intelligence. Narrow intelligence is already being developed, but the more difficult task is developing general intelligence, which is broadly transferrable.

McFarland noted that all of NSCAI’s findings were “consensus,” including two judgments: First, the U.S. government is not currently organized or resourced to win this competition against our adversaries. Second, that the nation must be “AI ready” by 2025.

According to the NSCAI, this requires four “pillars” of action: leadership, talent, innovation, and hardware. To this last point, McFarland noted the U.S. is “too dependent” on foreign partners for some manufacturing, such as semiconductors. Underlying every pillar are ethical considerations.

“We need to recognize a whole-of-government approach to this,” McFarland said.

NSCAI members will testify before the U.S. House Armed Services Committee on Friday.

声明:本平台发布部分内容来自公开资料或者网络,版权归原作者所有,转载的目的在于传递信息及用于网络分享,不代表本平台赞同其观点,如涉版权问题,请与我们联系,我们第一时间处理。