Pages

18 February 2019

The American AI Initiative: Bluster or Gangbuster?


President Trump signed an executive order (EO) yesterday outlining the American AI Initiative, a five-part national strategy to maintain U.S. leadership in artificial intelligence (AI). While long overdue, this is a critical first step in developing a concrete and coherent approach to building the technological foundation of the next generation of economic growth. The EO is notably short on specifics, and does not include any actual funding for implementation, but hits the right notes and emphasizes most of the important policies needed to support AI development and U.S. leadership. With a detailed implementation plan and strong leadership, the American AI Initiative could prove to be the most important policy initiative of the Trump administration.

Below we answer a few key questions about the EO and the American AI Initiative.

Q1: What is and is not in the EO?


A1: The EO outlines five key areas of effort as part of the strategy: research and development (R&D), infrastructure, governance, workforce development, and international engagement. In terms of setting priorities and identifying the key lines of effort for policymakers to support AI development, the EO is spot on. In fact, these five areas fit perfectly with our report from last year, “A National Machine Intelligence Strategy for the United States.

The EO is notably short on implementation, however. It includes little information on funding to support the initiative and few details about how departments and agencies can deliver on the ambitious goals of the strategy. As a high-level strategy document, the EO hits most of the right notes, but the real test of the strategy will be whether it translates into action on the ground.

Q2: Are these the right areas of focus?

A2: Perhaps the most important elements of the EO are that it broadens the focus of the AI debate in Washington beyond R&D. While R&D is a necessary and important element of an AI strategy, it is not sufficient to maintain U.S. leadership in AI.

The emphasis on enabling infrastructure and technologies is essential. AI is not a technology, it is the culmination of a series of technological trends; the growth of big data from online platforms and connected devices, the so-called Internet of Things (IoT), the deployment of 5G networks to provide ubiquitous high-bandwidth, low-latency connectivity, the adoption of cloud computing to provide the data storage and computing power necessary to drive machine learning applications, and the development of advanced robotics to allow machines to act on AI-derived insights at machine speed. AI also needs a robust data ecosystem that facilitates access to the vast amounts of data that feeds machine learning algorithms, while also protecting privacy and confidential business information. The EO, for the first time, recognizes the critical role of government in supporting the development of this enabling ecosystem for AI.

The government also has an essential role in managing the impact of AI on people, in particular through governance and workforce development. AI will have a significant impact on jobs, eliminating some existing classes of jobs and creating many new ones, but for many workers the rise of AI will mark a period of transition as they develop the skills and experience necessary to participate in the smart economy. Creating education and training opportunities and supporting workers in transition will be essential to ensuring that AI is a driver of growth and opportunity for Americans.

AI will also have profound social impacts, from changing the way we collect and use data to how resources and opportunities are allocated to different groups within society. Creating governance standards to manage these changes can ensure that they are consistent with American values and are a force for positive change and greater inclusion and opportunity.

The EO also has a section on international engagement, which emphasizes the importance of protecting the United States’ technological advantage and protecting U.S. innovators. However, a clear allusion to Chinese theft of intellectual property, which is an important issue, could become the Achilles heel of the strategy. This could happen if it sets an adversarial and self-interested tone with our international partners, or if it becomes an excuse to build barriers to the exchange of people, investment, and technology that is a key enabler of the U.S. innovation economy.

The EO says little, however, about the critical importance of collaborating with our international allies and partners. One of our greatest strengths as a nation, and one of our greatest advantages over perennial AI competitor China, is the strength of our relationships with innovators and leaders around the world. Leveraging the comparative advantages of our allies – for example, Japan’s strength in robotics, Germany’s use of data to optimize traditional manufacturing processes under Industrie 4.0, or Canada and the United Kingdom’s innovations in machine learning and deep learning—is a huge enabler of competitiveness and growth in the U.S. AI industry.

Taking the lead in international negotiations around AI governance is also overlooked. Managing the transformative forces of AI is a major issue for the international community and international institutions, and the United States has been notably absent from its traditional position of leadership in those discussions.

Finally, the EO addresses the critical role of government in incentivizing responsible adoption of AI across traditional industries and by consumers through “regulatory and non-regulatory approaches.” AI adoption can drive significant efficiency gains and productivity growth, helping businesses to cut costs, expand production, and reach new customers, but for many companies is impeded by cautious regulators and concerns about liability. Consumer mistrust and fear of AI-enabled products and services (often legitimate) is also a hurdle for consumer businesses. Reducing barriers to industry adoption of AI, particularly in regulated industries, and ensuring that security, privacy, and safety practices for AI products and services are robust enough to gain the confidence of consumers is an important part of the EO.

Q3: Will the strategy have a significant impact on U.S. leadership in AI?

A3: Whether the strategy will actually have an impact on the ground remains to be seen. The EO notably does not include funding for these initiatives, and there are few details on how it should be implemented by departments and agencies. A national strategy is only valuable if it leads to coordinated efforts across the government and in partnership with industry, the public, and international stakeholders, so additional details on implementation will be important.

There are three key questions to consider as we look at the potential impact of the EO:

1. Will the strategy translate into resources and attention to this issue across the federal government?

While many of us have highlighted the fact that China has a national AI strategy (they have three, in fact!) as a key area of advantage in the competition for leadership in AI, the power of their approach is that it has driven significant funding and attention to AI development. The EO omits any actual funding and resources to implement the strategy, and White House leaders have put the onus on Congress to appropriate funding to support AI development.

While appropriations are in the hands of Congress, the White House can direct departments and agencies to allocate funding and personnel to these issues without direct congressional action and can ensure that AI receives the attention it needs and deserves from senior leaders across the administration. If they can find $5 billion for a border wall without Congress, surely the Trump administration can find a few billion to build the next engine of U.S. economic growth.

2. Who will take the lead on implementing this strategy at the White House?

Perhaps the biggest question going forward will be figuring out who will be taking the lead on implementing this strategy within the White House. Executing a strategy as complex as this will require significant interagency coordination and strong leadership from the very top, particularly in this personality-driven administration. The value of a national strategy is that it harnesses all the levers of power and tools of policy in a cohesive and complementary series of efforts across the government.

The EO puts responsibility for seeing through the implementation of the strategy in the hands of the National Science and Technology Council, effectively operated by the Office of Science and Technology Policy (OSTP). It remains to be seen whether they have the influence, mandate, and enough presidential attention to manage a complex interagency process and coordinate the efforts of dozens of offices and agencies across multiple departments. Without the direct support and active participation of one of the influential personalities close to the president, the strategy may have little heft on the ground.

3. Can the White House work with Congress, state and local governments, industry, and international partners to implement the strategy?

The importance of partnerships cannot be overstated when it comes to a national AI strategy. Maintaining and building on our leadership in AI will require strong investment and buy-in from the private sector; coordinated policy development at the federal, state, and local levels; coalition building in global forums and institutions; and enabling legislation and funding from Congress. For an administration that is internally divided and plays notoriously badly with others, this may in fact be the greatest hurdle.

On the other hand, the EO is notably strong on substance and aligns well with the recommendations of experts across the government, private sector, academia, and international community. Perhaps if the White House can put aside the politics of division, other stakeholders will support what is, at its heart, a sound strategy.

The EO is a critical first step toward developing unquestioned U.S. leadership in AI. It provides the roadmap and high-level strategic vision that the U.S. government has lacked on these critical issues and should serve as the blueprint for a coordinated national effort to design, build, leverage, and govern the technology of the future. With a combination of strong leadership at the most senior levels, material investment of time, money, and attention in implementation, and the building of deep, broad, and robust coalitions of key stakeholders across the country and around the world, the EO could turn out to be the Trump administration’s most significant policy achievement.

William A. Carter is a fellow and deputy director of the Technology Policy Program at the Center for Strategic and International Studies in Washington, D.C.

No comments:

Post a Comment