The White House and Defense Department unveiled AI strategies. Now what?

By: Megan Lamberth 

Source Link

The White House and Department of Defense made major strides this month in artificial intelligence policy by unveiling two key strategy documents.

On Feb. 11, President Trump signed an executive order enacting the “American AI Initiative,” identifying AI as a priority for government research and development. The Initiative signals a shift within the executive branch, and lawmakers should seize this opening and allocate enough resources to support an enduring national AI strategy. The next day, the DoD released its own AI strategy, positioning its newly created Joint Artificial Intelligence Center (JAIC) at the forefront of its efforts. The DoD has long grappled with the transformative impact of AI and this plan attempts to formalize the Pentagon’s deployment of AI-enabled technologies. While these documents hold promise, the details of their execution and funding are noticeably absent. Without persistent vocal and financial support from the White House, Congress, and the DoD, these strategies will struggle to evolve from vision to reality.

The White House’s initiative wants to establish a whole-of-government strategy that uses the revolutionary power of AI to “sustain and enhance” U.S. economic and national security. It also carves out a road map for directing agencies to prioritize "sustained investment in AI R&D.”

That investment is shaped around five principles:

1. The initiative calls for a reinvigorated partnership between the government, private industry, and academia, recognizing the necessity of this alliance for long-term AI innovation.

2. It directs federal agencies to share more of their “data and models,” giving the AI research community access to large and diverse government datasets.

3. The document addresses the glaring gap of AI training in the American workforce and calls upon existing federal fellowship and service programs to prioritize “AI-related educational and workforce development.”

4. It acknowledges the ongoing public concern over privacy and data security and aims to “foster public trust and confidence in AI technologies.”

5. Finally, the executive order calls for strengthening “industry collaboration with foreign partners and allies,” while protecting against acquisition by “strategic competitors and adversarial nations.”

While the White House’s AI initiative is a step in the right direction, some gaps are evident. First, the executive order does not dictate a new source of federal funding, nor does it indicate how much the federal government should be spending on AI research and development. Instead, it puts the onus on individual agencies to “budget an amount for AI R&D that is appropriate for this prioritization.” Without additional funding, AI agencies will be forced to shift funds, leaving critical research areas underfunded. The Initiative also lacks any specific spending goals, giving the executive branch no metric to gauge if an agency is adequately prioritizing AI research. Overall, scant details are given for the Initiative’s execution. While it identifies those responsible for implementing the guidelines as the “AI R&D agencies,” it fails to provide a blueprint for putting the five principles in practice.

The DoD’s AI strategy is framed on the concept of a “human-centered adoption of AI,” the idea that humans play an essential role in the deployment and use of “thoughtful” and “responsible” uses of artificial intelligence. It calls for the rapid deployment of “resilient, robust, reliable, and secure” AI-enabled technologies to “address key missions” across the DoD. The strategy establishes the JAIC as a central hub, “synchronizing DoD AI activities across all DoD Components.” Like the executive order, the DoD’s strategy focuses on its relationship with the private sector, academia, and U.S. allies, emphasizing the importance of collaboration to “advance AI ethics and safety in the military context.” The conscientious language used throughout the document may reflect the department’s efforts to address public concerns raised by private technology sector employees. Last April, Google employees penned an open letter protesting the company’s involvement in the DoD’s Project Maven, leading the company to not renew the contract. Several months later, Microsoft employees deployed a similar tactic, urging the company to end its bid on the Pentagon’s JEDI project. The DoD strategy promotes using AI to tackle global challenges, such as “disaster relief,” as well as more mundane but nonetheless valuable objectives, like reducing inefficiencies in the DoD workforce, indicating an openness to partner with technology firms on less controversial uses of AI technologies.

Like the White House’s executive order, the DoD’s AI strategy gives little indication of how it intends to pay for its implementation. The JAIC’s 2019 budget was approximately $90 million and an accompanying fact sheet for the strategy explains the DoD “is identifying budget and resource impacts for the Fiscal Year 2020 budget.” The JAIC must have sufficient funding for the strategy to come to fruition. Additionally, while the strategy contains more detail than the White House’s Executive Order, the DoD’s ability to rapidly innovate and deploy AI technologies is uncertain. Established less than a year ago, the JAIC is in its infancy. It is faced with a tremendous task of overcoming the myriad bureaucratic obstacles within the department. The JAIC’s ability to tackle this challenge will depend on its funding and manpower, as well as the continued support at the highest levels of the Pentagon, no matter who is appointed the next Secretary of Defense.

These two AI strategy documents are not perfect, but they are important steps in developing a long-term AI national strategy. Much work remains, and it is too soon to determine the potential longevity and success of these efforts. Apart from the objectives dictated in the executive order, the Initiative is vague, and the White House will need to provide further instruction for its execution. As with any government initiative, sustained funding will be essential for the success of these strategies. Ongoing support from the White House and DoD leadership will also prove critical. Ultimately, the U.S. government must prioritize AI in both word and deed if it wants to keep pace with its strategic competitors. These initiatives are a solid start. Despite the two strategies’ gaps and drawbacks, proponents of safe and ethical AI advancement should be cautiously optimistic about the foundations established this week by the White House and Defense Department.

Megan Lamberth is a research assistant for the technology and national security program at the Center for a New American Security.