Machine Intelligence: We Need a National Strategy

Machine intelligence will be the defining technology of our era—the United States needs a plan to harness it.

On Thursday, July 20, China’s State Council released the New Generation of Artificial Intelligence Development Plan. Numbering nearly 40 pages, the plan lays out China’s aspirations in impressive detail. It introduces massive investment that aims to position China at the forefront of technological achievement by cultivating the governmental, economic, and academic ecosystems to drive breakthroughs in machine intelligence. To achieve these goals, the council aims to harness the data produced by the Internet-connected devices of more than a billion Chinese citizens, a vast web of “intelligent things.”

The plan also details the strategic situation precipitating the need for a bold new vision: “Machine intelligence [is] the strategic technology that will lead in the future; the world’s major developed countries are taking the development of AI [artificial intelligence] as a major strategy… [We] must, looking at the world, take the initiative to plan [and] firmly seize the [technology] in this new stage of international competition.”

The China State Council’s plan evokes a document that marked the beginning of the defining global technological competition of the last century—the space race. In August 1958, 10 months after watching the Soviet Union launch Sputnik 1, President Dwight Eisenhower’s administration released the U.S. Policy on Outer Space. In it, the U.S. National Security Council (NSC) urged massive investment to cultivate the talent and technology base necessary to exceed the Soviet Union’s achievements in space.

The United States is now at the precipice of another defining moment in history.

The NSC included an urgent mandate to act, declaring, “The starkest facts which confront the United States in the immediate and foreseeable future are [that] the USSR has surpassed the U.S. and the Free World in scientific and technological accomplishments in outer space, [and] if it maintains its present superiority…will be able to use [it] as a means of undermining the prestige and leadership of the United States.”

The United States is now at the precipice of another defining moment in history. The world’s greatest powers are entering a technological contest that will parallel or exceed the space race in the magnitude of its economic, geopolitical, and cultural consequences. Maintaining our role as a global superpower requires us to achieve parity, and ideally dominance, in the race to a future powered by intelligent machines. Moreover, we must develop a comprehensive national strategy for maintaining this technological advantage while also advancing our economy, preserving our social norms and values, and protecting our citizens’ dignity, privacy, and equality.

To understand the imperative for an all-inclusive national machine intelligence strategy, we must look at the current state of machine intelligence. The past three years marked a turning point in the development of the technology. Internet-connected devices have produced a glut of machine-readable data on which to teach machine-learning algorithms. Computing power advanced dramatically in the past three years, accelerating the speed with which these algorithms process data. Breakthroughs in research also produced new and more effective algorithms, inspired by the circuitry of the human brain.

Commercial machine intelligence technologies are beginning to proliferate. In August 2016, Microsoft’s machine intelligence software became as accurate at understanding speech as human beings. That same month, journalists reported Tesla’s autopilot software autonomously drove a man experiencing a blood clot much of the way to the hospital. The level of broad global investment and anticipation has few equals in modern history, portending a rapid and wide-reaching transformation. The implications are consequential.

Machine intelligence will fundamentally restructure economies. Routine work will be automated, increasing productivity, but also vastly changing the demand for a wide range of job categories. New categories of work, including jobs we can’t yet imagine, will arise. Workers will increasingly be asked to train, interact with, and trust the judgment of intelligent machines. Experts debate the speed and scale with which these impacts will be realized, but it is clear they will necessitate new approaches to education, skill development, and retraining and placing workers. A country’s preparedness to lead in this new economic landscape will determine its economic trajectory for decades to come.

The societal impacts of machine intelligence will also extend far beyond our economies. Creating intelligent machines necessitates vast troves of data. This information is so valuable that economists are referring to it as “the new oil.” Public and private institutions are driving to gain competitive advantage and achieve scientific breakthroughs using citizen data, which is already triggering heated debate about expectations of privacy. Algorithms may inadvertently acquire the human bias often hidden in the data we produce—and then apply it at a massive scale. Machine intelligence will enable the creation of indiscernible forgeries of news articles, audio recordings, and videos, making it difficult to separate truth from fiction. It will enable hackers to massively scale attacks on computer networks. Governments will be held accountable for their actions, or inactions, taken in the protection of citizen rights to privacy, equity, and justice.

We cannot afford to allow machine intelligence to develop untrammeled and without the thoughtful participation of the government that represents us.

Machine intelligence will also reshape the national security landscape. Some researchers believe it will be as transformative as the advent of aircraft, nuclear weapons, and computers. Entirely new military capabilities will be created while the cost of others will be significantly reduced, potentially arming a broader range of nefarious actors. Governments must rethink their defensive strategies to ensure the continued safety of their citizens and ongoing geopolitical stability.

Given these implications, it is no wonder that China sees this moment as the time to shape the future of this inchoate technology. And they are not alone. Canada announced in March of this year a plan to invest $125 million in a Pan-Canadian Artificial Intelligence Strategy aimed at shaping the trajectory of the technology. In April, the United Kingdom’s Royal Society, the world’s oldest society of academic experts, released a report detailing the likely impacts of machine intelligence on British society and suggesting responses.

The United States is confronted with a mandate to create a comprehensive national machine intelligence strategy that achieves four key policy objectives:

1. Maintains our technological advantage. We must create policy to maintain our globally acknowledged lead in machine intelligence technology development. Our current advantages are not preordained. Investment in basic research at federal agencies will be necessary to enable the long-term scientific breakthroughs that corporations’ balance sheets can’t support. In June 2016, the National Science and Technology Council (NSTC) released a plan for how U.S. agencies might invest in early stage machine intelligence research and development, but its recommendations have not gone into effect. Policymakers should revisit it. While significant investment is being made by venture capital and technology companies, this is often short term and doesn’t lead to research that ultimately yields the breakthroughs that are necessary for global dominance.
2. Advances our economy. We must support the United States’ great academic institutions and public schools in developing a strong, homegrown workforce, both to develop the technology as well as make effective use of it in the workplace. Encouraging the immigration of high-skilled technology experts will further cultivate a dynamic, technologically advanced talent base. Finally, we must put in place mechanisms to enable entirely new models of employment and ongoing education and training, allowing our workforce to advance at a pace matching that of the technology.
3. Preserves our social norms and values. We must develop policy that will balance the need for innovation with the obligation to protect U.S. citizens’ rights to privacy, equity, and justice. Policymakers must explore how to partner with academia and the commercial sector to shape the development of machine intelligence to the benefit of people and society. Regulation may be a component of this effort, but it will not be able to form the full picture.
4. Protects our citizens. We must ensure that we are developing technologies to combat the capabilities our adversaries seek to create and deploy. This must, however, be met with policy aimed at defining acceptable uses of autonomous technologies. We must position ourselves as a global leader in nonproliferation, striving to limit the unchecked spread of potentially threatening technology.

Some of the necessary components for developing our national machine intelligence strategy are already in place. The NSTC’s plan recommending government investment in basic machine intelligence research is a key starting point. Also in 2016, a report from the National Council of Economic Advisers considered the impact of machine intelligence on the U.S. economy and the well-being of U.S. workers. While these represent important areas of research, they are only foundational studies that do not yet begin to form a national strategy. We must begin the next phase of planning for our new technological era. We cannot afford to allow machine intelligence to develop untrammeled and without the thoughtful participation of the government that represents us.

On July 20, 1969, the United States ultimately won the space race when America was the first to reach the moon and placed a plaque that read, “Here men from the planet Earth first set foot upon the moon. July 1969 A.D. We came in peace for all mankind.” With a comprehensive national strategy, the United States can again lead the world toward a bright and peaceful future with intelligent machines. Our action or inaction now will define our place on the world stage for decades to come.

(This article was originally published by the Atlantic. It is reprinted here with permission.)

William A. Carter is a fellow and d eputy director of the Technology Policy Program at the Center for Strategic and International Studies in Washington, D.C. Emma Kinnucan is a machine intelligence and data science strategist at Booz Allen Hamilton. Steven Mills is principal and director of machine intelligence at Booz Allen Hamilton.

Commentary is produced by the Center for Strategic and International Studies (CSIS), a private, tax-exempt institution focusing on international public policy issues. Its research is nonpartisan and nonproprietary. CSIS does not take specific policy positions. Accordingly, all views, positions, and conclusions expressed in this publication should be understood to be solely those of the author(s).

© 2017 by the Center for Strategic and International Studies. All rights reserved.