Why the United States Needs a National Artificial Intelligence Strategy and What It Should Look LikeThis report explains why a national AI strategy is necessary to bolster U.S. competitiveness, strengthen national security, and maximize the societal benefits that the country could derive from AI. It then lays out six overarching goals and 40 specific recommendations for Congress and the administration to support AI development and adoption.WHY THE U.S. NEEDS A NATIONAL AI STRATEGY -- There are three major reasons why the United States needs a national AI strategy: 1) to boost U.S. economic competitiveness; 2) to support U.S. defense capabilities; and 3) to overcome market failures, including the provisioning of public goods, that would otherwise slow AI development and adoption.Center for Data InnovationC4DIID-1b4df920-5e1f-4c8a-a6ef-e5cd5de8317dJoshua NewUnited StatesThe United States is the global leader in developing and using artificial intelligence (AI), but it may not be for long.CompaniesSucceeding in AI requires more than just having leading companies make investments. It requires a healthy ecosystem of AI companies, robust AI inputs—including skills, research, and data—and organizations that are motivated and free to use AI.AI CompaniesOrganizationsFederal GovernmentAnd that requires the federal government to support the development and adoption of AI.CountriesMany other countries, including China, France, and the United Kingdom, are developing significant initiatives to gain global market share in AI.ChinaChina’s State Council issued a development plan for AI in July 2017 with the goal of making China a leader in the field by 2030.28 The document is primarily a statement of intent, but it details some of China’s key objectives in advancing AI and creating a domestic AI industry worth ¥1 trillion (US $147.8 billion). The plan’s goal is for China to be equal to countries leading in AI by 2020. Then over the subsequent five years, China will focus on developing breakthroughs in areas of AI that will be a “a key impetus for economic transformation.” Finally, by 2030, China intends to be the world’s “premier artificial intelligence innovation center.” To support the development plan, China is also preparing a multibillion-dollar investment initiative to promote AI startups, academic research, and ambitious moonshot projects. China, like the United States, has its own natural advantages when it comes to AI. First, the Chinese private sector often moves in lock-step with the government to secure favorable treatment. At the end of 2016, 67.9 percent of the 2.73 million non-state-own companies in China had Communist Party cells within their organizations, and authorities often appoint business leaders with memberships in the National People’s Congress, China’s national legislature. Kenji Kawase, chief business news correspondent at the Nikkei Asian Review, describes that “These positions are largely ceremonial, but membership symbolizes a political recognition by the party and the state, and also provides a certain amount of protection. For the authorities, selecting these business leaders is a way to ensure that private companies cooperate with policy goals, including that of gaining a tighter grip over cyberspace.” Thus, the Chinese government has a de facto authority to direct private-sector investment in AI as it sees fit to advance its goals rather than allow businesses to act pragmatically in their own best interests. This close alignment is also reflected in the actions of Chinese military, financial institutions, and subnational governments, ensuring that in China, “commercial companies, university research laboratories, the military, and the central government routinely work together closely. As a result, the Chinese government has a direct means of guiding AI development priorities and principles.” Second, whether as the result of different values or China’s more authoritarian government, Chinese organizations using AI simply do not have to grapple with the same regulatory and consumer protection considerations that U.S. firms do. For example, China’s plans to deploy ubiquitous facial recognition technology, surveil citizens with drones, monitor messaging apps, and implement a “social credit system” would all likely be met with widespread consumer backlash in the United States and other western democracies, but are being implemented in China without notable domestic resistance (though it should be noted that vocal dissent is likely censored or otherwise discouraged). The Chinese government’s control over private technology firms and its ability to deploy largescale data collection systems with little regard to consumer protection or privacy give it an advantage over the United States because it gives it access to vast troves of consumer and personal data that U.S. firms have difficulty accessing. As this paper will describe, data is a crucial input for AI development, and this gives China a large competitive edge in that regard. Third, China has a long history of stealing intellectual property from foreign firms, particularly U.S. firms, and despite its claims that it respects intellectual property, there is every reason to believe it will continue this practice as international competition regarding AI escalates. Both the Office of the United States Trade Representative and the European Commission published reports in 2018 detailing numerous problems with how China protects the intellectual property of foreign firms. These include forced technology transfers, requiring the disclosure of business information, and theft of trade secrets. Thus, as China seeks to promote its domestic AI industry, it is safe to assume that it will simply steal intellectual property that U.S. firms invested large sums to develop. Thus, when comparing national efforts to support AI, it is crucial for U.S. policymakers to recognize that when it comes to competing with China, this is not an even playing field. This makes a U.S. national AI strategy all the more important to ensure U.S. firms can compete effectively. FranceFrance published a report detailing its national AI strategy titled “For a Meaningful Artificial Intelligence,” in March 2018. This strategy includes investing €1.5 billion ($1.85 billion) over five years to support R&D, promote AI startups, and foster the creation of valuable datasets that can support AI development. Overall, the strategy fails to describe how the French government will promote widespread AI adoption, however it is noteworthy for its emphasis on making data available for AI. The French strategy, drafted by mathematician and Member of Parliament Cédric Villani, calls for legislation to mandate repurposing both public and private-sector data, including personal data, to enable public-interest uses of AI by government or others, depending on the sensitivity of the data. For example, public health services could use data generated by Internet of Things (IoT) devices to help doctors better treat and diagnose patients. Researchers could use data captured by motorway CCTV to train driverless cars. Energy distributors could manage peaks and troughs in demand using data from smart meters. Repurposed data held by private companies could be made publicly available, shared with other companies, or processed securely by the public sector, depending on the extent to which sharing the data presents privacy risks or undermines competition. The report suggests that the government would not require companies to share data publicly when doing so would impact legitimate business interests, nor would it require that any personal data be made public. Instead, Villani argues, if wider data sharing would do unreasonable damage to a company’s commercial interests, it may be appropriate to give only public authorities access to the data. But where the stakes are lower, companies could be required to share the data more widely, to maximize reuse. Villani rightly argues that it is virtually impossible to come up with generalizable rules for how data should be shared that would work across all sectors. Instead, he argues for a sector-specific approach to determining how and when data should be shared. After making the case for state-mandated repurposing of data, the report goes on to highlight four key sectors as priorities: health, transport, the environment, and defense. Since these all have clear implications for the public interest, France can create national laws authorizing extensive repurposing of personal data without violating the General Data Protection Regulation (GDPR), which permits the repurposing of personal data where it serves the public interest. The French strategy is the first clear effort by an EU member state to proactively use this clause in aid of national efforts to bolster AI. However, the strategy primarily focuses on repurposing data in the public interest because the GDPR limits repurposing data for commercial purposes. But many important uses of AI—from banking to agriculture—are commercial. France is making the best of the situation created by the GDPR, but its limited strategy highlights the need for EUlevel reform of data protection law in order to enable more ambitious AI strategies in the member states. Ultimately, unless the EU reforms the GDPR to enable greater collection, use, and sharing of personal data, any European country’s AI strategy will be constrained by the GDPR’s limitations.United KingdomThe United Kingdom has taken several steps to better understand AI and identify ways the government could help secure its benefits. In October 2016, the House of Commons Science and Technology Committee published a report on robotics and AI detailing many of the potential benefits and challenges AI could offer. One of the report’s main conclusions was that the United Kingdom should place a greater focus on improving its education and worker training systems to ensure that the national workforce has the necessary skills to be successful as AI transforms the economy. The report also stressed the need of increased government leadership around robotics and autonomous systems, citing a lack of a government strategy to coordinate policymaking and guide investment. In November 2016, the Government Office for Science published a report detailing the potential implications AI poses for society and government and stressed the need for smart, flexible governance to promote the responsible development of AI. The UK Digital Strategy, published in March 2017, recognizes AI as a key field that can help grow the United Kingdom’s digital economy, and includes £17.3 million (US $22.3 million) in funding for UK universities to develop AI technologies. And in April 2018, the government launched its AI Sector Deal, an extensive industrial strategy that combines and builds on its prior efforts to support AI designed to “boost the U.K.’s global position as a leader in developing AI technologies.” The AI Sector Deal also lays out an interesting approach to increasing the amount of data available for AI called “data trusts,” which are “mechanisms where parties have defined rights and responsibilities with respect to share data.” This is a promising approach as it could encourage businesses, government agencies, and researchers to share sensitive or valuable proprietary data with one another to advance AI research by reducing concerns that this data could be misused. CanadaIn March 2017, Canada launched the Pan-Canadian Artificial Intelligence Strategy, to be led by the Canadian Institute for Advanced Research (CIFAR), a nonprofit research institute that receives government support. Backed by a one-time CAD $125 million (US $98.7 million) government investment, the strategy has four goals: “increase the number of outstanding artificial intelligence researchers and skilled graduates in Canada; establish interconnected nodes of scientific excellence in Canada’s three major centres for artificial intelligence in Edmonton, Montreal, and Toronto; develop global thought leadership on the economic, ethical, policy and legal implications of advances in artificial intelligence; and support a national research community on artificial intelligence.” CIFAR will oversee several programs over the next five years to advance the strategy that focuses on expanding Canada’s human capital, raising Canada’s international profile in the field of AI research, and translating AI research into public and private-sector applications. Canada has also allocated CAD $950 million (US $718 million) to fund the creation of five technology “superclusters” designed to foster collaboration and accelerate growth and job creation around different technology issues. Several of these, including the Scale AI supercluster in Quebec, the Advanced Manufacturing supercluster in Ontario, and the Digital Technology supercluster in British Columbia, have an explicit focus on AI. IndiaIn June 2008, India published a discussion draft of its national AI strategy, which is focused on overcoming barriers limiting the development and adoption of AI at scale, such a lack of technical expertise and restrictions on data access. The strategy includes policy recommendations designed to address these barriers and ensure India can capture AI’s economic and social value, including establishing research centers that focus on advancing AI applications in key sectors such as health care, education, and agriculture and creating annotated “foundational” datasets that could serve as a public resource to spur AI development. India’s strategy also sets the goal of becoming the AI “garage” for the 40 percent of the world whose economies are developing. This means it would create an environment such that if a firm can successfully deploy an AI application in India, it can be confident that it could deploy it in the rest of the developing world.JapanJapan launched its Artificial Intelligence Technology Strategy Council in April 2016 to develop a roadmap for the development and commercialization of AI. It was published in May 2017. The strategy outlines priority areas for research and development (R&D), focusing on the themes of productivity, health, medical care, and mobility. The strategy also encourages collaboration among industry, government, and academia to advance AI research, as well as stresses the need for Japan to develop the necessary human capital to work with AI. Japan also launched its Japan Revitalization Strategy 2017, which details how the government will work to support growth in certain areas of the economy. The 2017 strategy includes a push to promote the development of AI telemedicine as well as the development of self-driving vehicles to help address the shortage of workers in Japan’s logistics sector.South KoreaSouth Korea has made several large commitments to support the development and adoption of AI, starting with a March 2016 announcement of a ₩1 trillion (US $886.7 million) investment in AI R&D over five years, a 55 percent increase in its annual spending on AI. In May 2018, the government announced a ₩2.2 trillion (US $1.95 billion) investment in AI over five years with the goal of establishing six AI graduate schools to train 5,000 AI specialists, advance the development of AI applications in areas including national defense and public safety, and foster AI startups and small businesses.TaiwanTaiwan’s Premier William Lai announced the Taiwan AI Action Plan, which details strategies to grow Taiwan’s AI industry, in January 2018. The four-year plan states the government will allocate between NT$9 billion and NT$10 billion (US $304.4 million and US $338.3 million, respectively) annually to cultivate AI talent, including by recruiting international talent and making it easier for foreign workers to work in Taiwan, developing AI pilot projects, fostering AI startups, and increasing the availability of data, such as by creating open data platforms and developing flexible regulations about data use.U.S. GovernmentWhile the United States has taken some steps to support the technology, it does not have a cohesive national strategy for AI. In 2016, the White House Office of Science and Technology (OSTP) hosted five workshops with academic leaders on different social, ethical, economic, and technological aspects of AI, and in June 2016, solicited public feedback about AI. In October 2016, OSTP published a report titled “Preparing for the Future of Artificial Intelligence” detailing its findings and recommending that the government pursue policies that can help maximize the economic and social benefits of AI. That same month, the Networking and Information Technology Research and Development Subcommittee (NITRD) published its National Artificial Intelligence Research and Development Strategic Plan detailing seven strategies to help guide AI R&D efforts, including “develop effective methods for human-AI collaboration,” “develop shared public datasets and environments for AI training and testing,” and “better understand the national AI R&D workforce needs.”58 Finally, in December 2016, the White House published a report titled “Artificial Intelligence, Automation, and the Economy” reaffirming many of the recommendations from its prior efforts, particularly that the government should ensure the workforce is equipped with the skills to thrive in the transition to an AI-driven economy. However, most of the efforts from the Obama administration were foundational to support more explicit policies later.Trump AdministrationThe Trump administration has taken some additional steps to support AI. In May 2018, the White House hosted a summit titled “Artificial Intelligence for American Industry,” convening leading technology companies to discuss methods for fostering the advancement of AI. Though the summit did not result in the formulation of any new policies or lead to any additional public action, it did establish a Select Committee on Artificial Intelligence under the National Science and Technology Council to advise the White House on AI issues, improve coordination of federal AI R&D efforts, and identify opportunities to leverage federal data and computing resources to support AI R&D. And in September 2018, NITRD issued a request for information to update its AI R&D Strategic Plan. In the absence of a national strategy, NITRD’s AI R&D Strategic Plan amounts to the most substantive and comprehensive effort to maximize the benefits of AI for the United States, however this document does not direct policy, funding, or regulation.CongressCongress has taken several actions to better understand the challenges and opportunities posed by AI and has introduced several pieces of legislation to support the technology. For example, the House Oversight Subcommittee on Information Technology held three hearings about AI, exploring how to improve the government’s use of the technology and how best to address policy questions about ethics and competitiveness in AI. Bipartisan members of the House and Senate have also introduced the FUTURE of Artificial Intelligence Act of 2017 (S. 2217) and the AI in Government Act of 2018 (S. 3502), which would direct the Department of Commerce to create an advisory committee to guide activities to support AI development, and spur the federal government’s adoption of AI, respectively.U.S. competitiveness is bolstered, national security is strengthened, and the societal benefits of AI are maximized_47519bec-1827-11e9-8fa1-8a76d5e8efbcTo lay out goals and recommendations to support AI development and adoption_4751a20e-1827-11e9-8fa1-8a76d5e8efbcAITHE IMPORTANCE OF AI -- As an emerging general-purpose technology, like electricity and information technology, AI has the potential to drive innovation, competitiveness and productivity. According to the consulting firm PwC, AI will increase global GDP by up to 14 percent by 2030 due to its ability to drive productivity gains by automating business processes and augmenting human labor.1 For the United States specifically, consulting firm Accenture estimates that AI could increase labor productivity by 35 percent and increase the annual growth rate of gross value added to the U.S. economy from 2.6 percent to 4.6 percent by 2035 through automation and improving labor and capital management. AI has the potential to generate a range of societal benefits, such as by helping develop new medical treatments, improving public safety, and fighting human trafficking. For example, Facebook has developed computer-vision algorithms that can describe images to blind users, making the Internet more accessible for people with visual impairments. San Francisco-based analytics company Kanjoya has developed machine learning software that can analyze workforce communications and flag signs of implicit gender bias so companies can treat employees more fairly. And conservation technology start-up Conservation Metrics uses a system of acoustic sensors and machine learning to improve conservation efforts for threatened species in California.CompetitivenessThe United States has unique advantages over other countries that have made it an early leader in AI, particularly its large technology sector and its innovation-friendly regulatory environment. For example, in 2016, 66 percent of global investment in AI went to the United States, with Silicon Valley and the San Francisco Bay Area in particular attracting 40 percent of global investment in AI. By contrast, China, which attracted the second largest share of global investment in AI, received just 17 percent. And while the United States has traditionally embraced a regulatory philosophy based on the innovation principle—the idea that the majority of innovations overwhelmingly benefit society, and the government’s role should be to pave the way for widespread innovation while building guardrails, where necessary, to ensure public safety—others have taken a much more precautionary approach, at the expense of innovation. For example, the European Union’s recently enacted General Data Protection Regulation (GDPR) contains a variety of provisions that harm the ability of European firms to take advantage of AI while doing little to protect European consumers. However, there are signs that the United States’ early lead is slipping. In 2017, China secured 48 percent of global investment in AI startups, while the United States received just 38 percent. According to a study by The Economist about national readiness for automation, based on national innovation environments, labor market policies, and education policies, the United States ranked 9th out of 25 countries studied, while South Korea, Germany, Singapore, Japan, and Canada ranked 1st through 5th place, respectively. And since 2015, China began publishing more patents related to AI and deep learning than the United States and has grown that lead considerably, publishing six times as many AI and deep learning patents than the United States in 2017. And while other governments are aggressively increasing their research funding for AI, U.S. government research has remained relatively flat. NSF and the National Science Board noted in February 2018 that if current trends continue, China will surpass the United States in all R&D investments by the end of 2018.National SecurityThe ability of the United States to harness AI for defense purposes will also have crucial implications for national security. The opportunities to use AI for national security are vast, including improving logistics, analyzing surveillance footage and satellite imagery, and improving training exercises. AI can support the use of autonomous and semi-autonomous weapons systems and provide analysis and threat assessments to soldiers on the battlefield. However, the bulk of the benefits AI can offer national security are similar to the benefits it can offer the private sector, such as automating routine processes, improving data analysis, identifying and fighting cyber threats, and processing large amounts of sensor data. Beyond improving existing defense operations, ensuring the United States can use AI effectively for national security will be a necessity to maintain technological superiority over adversaries that are investing heavily in AI for military use. For example, China is developing a range of autonomous aerial, ground, surface, and underwater military vehicles, and Russia is using AI to augment its information warfare activities. To be sure, if the Defense Department needs cutting-edge AI capabilities it can attempt to procure military-specific AI technologies. But while DoD drove IT innovation in the 1950s and even the 60’s, today the center of IT innovation, and AI innovation, is in the private sector. And DoD will be increasingly reliant on partnerships with and purchases from the commercial IT industry for cutting-edge AI capabilities. However, it would be extremely risky to national security to rely on those capabilities from military adversaries.Public GoodsMarket Failures -- There are a number of market failures that will slow AI development and adoption in the absence of supportive policies. First is the fact that effective AI development requires access AI skills and a broad base of technical knowledge. These inputs—skilled workers and public R&D—are public goods. Though the private sector invests in worker training and R&D of its own, it does not capture all of the benefits of this investment. For example, a company could provide years of technical training for a new hire, only to have him or her leave to work for a competitor. Thus, firms often underinvest in these inputs relative to societally optimal levels. Smart policies to increase the availability of these public goods can help correct for this.KnowledgeSkillsRisk MitigationA second failure relates to risk and uncertainty. Because AI is an emerging technology, many potential users, including companies and government agencies, will minimize the benefits it promises and delay adoption until the technology is proven. Economists refer to this challenge as excess inertia or, more commonly, “the penguin effect”—in a group of hungry penguins, no individual penguin is willing to be the first to enter the water to search for food due to the risk of encountering a predator. Yet if no penguin is willing to test the waters, then the whole group risks starvation.ExternalitiesA third failure relates to externalities. The widespread use of AI can generate significant social and economic benefits. However individual firms investing in AI are unable to capture some of these benefits, leading to underinvestment. For example, the ubiquitous use autonomous vehicles could drastically reduce traffic deaths, property damage, and congestion (and the resulting loss of economic activity). Yet no individual company developing autonomous vehicles could expect to capture enough value from these benefits to recoup the costs of developing the technology. Finally, there are collective action problems. One of the key drivers of AI is widespread data availability. However, organizations often have strong incentives not to share data. Even though all parties would be better off if each shared data for mutual benefit, without the proper incentives for participation and organizations willing and able to coordinate these efforts, there will be less data available overall for AI development. Despite these compelling rationales, U.S. policymakers have not yet committed to developing such a strategy. For example, a spokesman for the Senate Commerce Committee stated that because the private sector is already using and investing heavily in AI, “the horse has already left the barn… and I think that any attempts by government to try and intervene could be constraining on the development of this technology.” To be sure, government action to regulate could be constraining. But that doesn’t mean that policies to proactively support AI development and adoption would not be helpful or needed.FacilitationMost nations that have developed AI policies have recognized that active government support can accelerate AI development and adoption. As such, most nations, particularly Southeast Asian nations, are focused primarily on facilitating AI development and adoption. To be sure, an overzealous approach to facilitation, such as China’s approach in particular, can introduce government failures such as supporting the wrong firms or limiting foreign competition. Done correctly however, a facilitation approach combines the best of enterprise-driven market forces with societal support and facilitation. An ideal national AI strategy will incorporate elements of all three approaches, recognizing that the private sector and market forces will play decisive roles in the advancement of AI, modernizing regulations for the AI economy in ways that enable AI innovation, and emphasizing the right government role in facilitating widespread development and adoption of AI.CompetitionASSESSING THE COMPETITION -- Unlike the United States, a number of countries have translated recognition of the importance of AI into significant action. It is important to recognize that this response is unique. During the last major IT revolution of the Internet, virtually no U.S. competitor took it seriously, at least initially, and few developed robust policies to compete globally. Things are different this time. Many nations now understand that AI is a central general-purpose technology of the next era, and they are loath to make the same mistakes they made 30 years ago. To be sure not every country intends to compete with the United States in every aspect of AI development and adoption. Instead, most countries’ strategies focus on capitalizing on their respective comparative advantages by prioritizing their efforts to secure the benefits of AI where it can have the most impact. However, this should not be interpreted as a chance for the United States to rest on its laurels; rather, because as the largest economy globally it needs to be competitive in every aspect of AI. Several leading countries, including China, France, and the United Kingdom, have developed comprehensive national AI strategies and are in the process of implementing them. Additionally, others have made their intentions to support AI clear, though have yet to formally introduce a strategy. For example, German Chancellor Angela Merkel has announced a proposal to make €3 billion (US $3.34 billion) available to Germany’s private sector to support AI R&D through 2025, and Germany is in the process of drafting a broader AI strategy.22 And the European Commission published its Communication on Artificial intelligence in Spring 2018 explaining its plans to boost technical capacity and spur public and private-sector AI adoption, and has outlined its planned activities to support AI, such as by increasing investments and improving research centers, through 2020.23 Many other countries have also signaled that they will be developing policies to support AI, though their plans are less clear.InputsSupport key AI organizational inputs._4751a4fc-1827-11e9-8fa1-8a76d5e8efbc1To effectively develop or use AI, organizations need abundant access to three key resources: high-value data, AI skills, and publicly funded research and development.DataEnsure Data Availability_4751a72c-1827-11e9-8fa1-8a76d5e8efbc1.1AI Training & ValidationDevelop shared pools of training and validation data in key areas of public interest_4751aca4-1827-11e9-8fa1-8a76d5e8efbc1.1.1Federal AgenciesRelevant federal agencies should support the development of shared pools of high quality, application-specific training and validation data in key areas of public interest, such as agriculture, education, health care, public safety and law enforcement, and transportation.Agriculture SectorEducation SectorHealth Care SectorPublic Safety SectorLaw Enforcement SectorTransportation SectorNational Institute of Standards and Technology (NIST)For example, the National Institute of Standards and Technology (NIST) should work with law enforcement agencies, civil society, and other stakeholders to develop shared, representative datasets of faces that can serve as an unbiased resource for organizations developing facial recognition technology. Law Enforcement AgenciesCivil SocietyData TrustsDevelop and pilot data trusts to facilitate data sharing in specific application areas_4751af6a-1827-11e9-8fa1-8a76d5e8efbc1.1.2Federal AgenciesRelevant federal agencies, including the Department of Commerce and the Department of Health and Human Services, should develop and pilot data trusts to facilitate data sharing in specific application areas among academia, businesses, and government agencies. Department of CommerceDepartment of Health and Human ServicesAcademiaBusinessesGovernment AgenciesCongressCongress should also fund and task the Department of Commerce to pursue additional innovative models for increasing the availability of data.Department of CommerceData CouncilsThis should include facilitating the creation of industry-led data councils to identify barriers to data sharing and developing strategies to overcome these barriers. DigitizationAccelerate efforts to digitize all sectors of the economy_4751b1ae-1827-11e9-8fa1-8a76d5e8efbc1.1.3CongressCongress and the administration should accelerate efforts to digitize all sectors of the economy, including health care, education, and municipal issues such as utilities and city management.The AdministrationHealth Care SectorEducation SectorMunicipalitiesCity ManagersFederal AgenciesFederal agencies, such as the Department of Housing and Urban Development (HUD), Department of Health and Human Services (HHS), Department of Transportation (DOT), and Federal Energy Regulatory Commission (FERC), should identify and implement policies that can drive digital transformation in relevant sectors. Department of Housing and Urban Development (HUD)Department of Health and Human Services (HHS)Department of Transportation (DOT)Federal Energy Regulatory Commission (FERC)Private SectorEncourage the private sector to share data for public benefit_4751b514-1827-11e9-8fa1-8a76d5e8efbc1.1.4Private SectorCongressCongress and the administration should encourage the private sector to share data for public benefit.The AdministrationPolicymakersPolicymakers should consider a variety of different approaches to encourage this.FranceFor example, France’s AI strategy proposes requiring the private sector to share certain data sets in select circumstances, when it does not threaten a firm's business and relates to key public interests such as health and safety.United States GovernmentThe United States government should not appropriate data from the private sector, but instead take a more collaborative approach to identifying such datasets and working with the private sector to increase their availability.There are many examples of firms voluntarily making data available for AI research, however this is not the norm.CodificationPass legislation codifying the federal government’s responsibility to publish open data_4751b794-1827-11e9-8fa1-8a76d5e8efbc1.1.5CongressCongress should pass legislation codifying the federal government’s responsibility to publish open data. Since 2013, the federal government has had a policy of treating its data as open and machine readable by default, opening vast troves of government data to the public. This data is a crucial public resource for businesses, academics, and public-sector employees alike, however the government has no legal obligation to continue publishing this data. This means that the government may decide to stop making certain data available at any time, such as with a new administration or agency priorities. Congress should pass legislation to codify open data requirements to ensure that government data remains available as a valuable source of data for AI systems.FundingAllocate additional funding for the federal government’s open data efforts_4751b992-1827-11e9-8fa1-8a76d5e8efbc1.1.5.1CongressCongress should also allocate additional funding for the federal government’s open data efforts to improve the timeliness, quality, and accessibility of its data.Data PovertyEnsure data collection efforts emphasize reducing the "data divide” and combatting data poverty_4751bcee-1827-11e9-8fa1-8a76d5e8efbc1.1.6Federal AgenciesRelevant federal agencies should ensure data collection efforts emphasize reducing the "data divide” and combatting data poverty.Underrepresented CommunitiesFor the public sector, this means supporting and expanding data collection programs that focus on hard-to-reach and underrepresented communities.Federal ProgramsAdditionally, this would mean ensuring that federal programs devoted to closing the digital divide also consider data poverty concerns. NISTMore broadly, an agency such as NIST could develop educational materials about how to improve data collection in the private sector to combat the data divide.Private SectorPrivacyEnsure that any national legislation addressing privacy considers the importance of data for the development and use of AI_4751bfbe-1827-11e9-8fa1-8a76d5e8efbc1.1.7CongressCongress should ensure that any national legislation addressing privacy considers the importance of data for the development and use of AI and does not impose undue restrictions on the collection, sharing, and use of data that come at the direct expense of AI innovation, such as an opt-in requirement for data sharing. R&DConduct AI R&D_4751c1e4-1827-11e9-8fa1-8a76d5e8efbc1.2FundingIncrease R&D funding for AI_4751c568-1827-11e9-8fa1-8a76d5e8efbc1.2.1CongressCongress should substantially increase R&D funding for AI, with an emphasis on basic and applied research. Exact dollar amounts can be debated, but at the very least Congress should provide funding to fully support R&D efforts where the potential supply of high-quality research is greater than the supply of funds.NSFFor example, Congress should appropriate at least an additional $200 million annually to NSF for AI research.ApplicationsSupport R&D for all kinds of AI applications_4751c996-1827-11e9-8fa1-8a76d5e8efbc1.2.2Federal agencies should support R&D for all kinds of AI applications. It is a mistake for federal agencies to support only certain kinds of AI and related technologies. For example, NSF’s National Robotics Initiative only supports research that augments, and does not replace workers. While augmenting human labor is a valuable application of AI, AI that can replace a human worker does more to boost productivity and should at minimum be on equal footing for support of AI that complements workers.Tax CreditIncrease the R&D tax credit_4751cb9e-1827-11e9-8fa1-8a76d5e8efbc1.2.3CongressCongress should increase the R&D tax credit to keep pace with competing countries. A healthy AI ecosystem requires both government and business funding of AI research.CompaniesCompanies will do more AI research in the United States if the R&D tax credit is more generous.OECDHowever, as of 2017, the United States ranked 32nd of OECD nations in terms of R&D tax credit generosity, behind countries such as Canada, China, Germany, and Japan. And in 2018, a number of nations created or expanded their R&D tax incentives.Meanwhile, the 2017 tax legislation passed by Congress in 2017 actually increased the after-tax cost of research spending. As such, Congress should increase the Alternative Simplified Credit from 14 percent to 20 percent.Public SectorAccelerate public-sector adoption of AI, including for national security._4751cf04-1827-11e9-8fa1-8a76d5e8efbc2Public SectorDefense AgenciesOne key area for public-sector adoption is defense. Defense agencies recognize the benefits AI can offer, but they face unique challenges in putting the technology to use to enhance national security.One of the most straightforward and effective steps government can take spur AI progress is to rapidly adopt AI in support of its own missions. Government can help prove the value of deploying AI, as well as provide markets and increase economies of scale for AI firms. To do this, a strategy needs to address challenges related to acquisition, funding, and oversight.TransformationTransform Government With AI_4751d364-1827-11e9-8fa1-8a76d5e8efbc2.1Awareness & CoPsFoster communities of practice and raise awareness about AI within the public sector_4751d5a8-1827-11e9-8fa1-8a76d5e8efbc2.1.1CongressCongress and the administration should support efforts to foster communities of practice and raise awareness about AI within the public sector.The AdministrationEmerging Citizen Technology Office (ECTO)For example, the General Services Administration’s (GSA’s) Emerging Citizen Technology Office (ECTO) coordinates government-wide deployments of AI applications and helps foster relationships between government employees interested in AI and firms working on public-sector applications of AI.General Services Administration (GSA)Government Employeesinterested in AIFirmsworking on public-sector applications of AIVenture CapitalProvide agencies with venture capital funds to pilot AI initiatives_4751d9fe-1827-11e9-8fa1-8a76d5e8efbc2.1.2CongressCongress should provide agencies with venture capital funds to pilot AI initiatives.ProgramsEstablish domain-specific programs to spur AI adoption_4751dd3c-1827-11e9-8fa1-8a76d5e8efbc2.1.3Federal AgenciesFederal agencies should establish domain-specific programs to spur AI adoption.Joint Artificial Intelligence Center (JAIC)For example, the Department of Defense (DoD) recently established the Joint Artificial Intelligence Center (JAIC) to help teams “to swiftly deliver new AI-enabled capabilities and effectively experiment with new operating concepts in support of DoD’s military missions and business functions.”Department of Defense (DoD)HHSOther departments, such as HHS and DOT, should consider developing similar programs.DOTStrategic InitiativeEstablish a strategic initiative devoted to AI_4751df80-1827-11e9-8fa1-8a76d5e8efbc2.1.4The White HouseThe White House should establish a strategic initiative devoted to AI in the CIO Council.CIO CouncilThere is currently an initiative devoted to “data analytics and big data” which would likely cover certain aspects affecting AI adoption, however there should be a more explicit focus on AI.StatesMake it easier for state officials to learn about and procure AI technologies_4751e390-1827-11e9-8fa1-8a76d5e8efbc2.1.5GSAGSA should work with state government CIOs to share best practices for AI implementation and develop shared resources that make it easier for state officials to learn about and procure AI technologies.State Government CIOsNational SecurityPrioritize the use of AI to protect national security_4751e6b0-1827-11e9-8fa1-8a76d5e8efbc2.1.6Defense AgenciesDefense agencies should prioritize the use of AI to support their missions to protect national security.Dual-Use AI TechnologiesCreate a body to accelerate the adoption of dual-use AI technologies by the military_4751e8e0-1827-11e9-8fa1-8a76d5e8efbc2.1.7DoDDoD should create a body with both government and industry stakeholders to accelerate the adoption of dual-use AI technologies by the military.The MilitaryIndustryThis could include publishing performance and safety standards for various key military AI applications so industry could more readily develop those solutions, or creating guidelines for modifying commercial AI applications for military use.Performance & Safety StandardsPublish performance and safety standards for various key military AI applications_4751ec6e-1827-11e9-8fa1-8a76d5e8efbc2.1.7.1IndustryGuidelinesCreate guidelines for modifying commercial AI applications for military use_4751ef48-1827-11e9-8fa1-8a76d5e8efbc2.1.7.2Commercial AI ApplicationsAcquisition Task ForceEstablish a cross-agency task force to identify opportunities to simplify the acquisition process for AI_4751f16e-1827-11e9-8fa1-8a76d5e8efbc2.1.8DoDDoD should establish a cross-agency task force to identify opportunities to simplify the acquisition process for AI.Acquisition MechanismsPursue and expand the use of alternative acquisition mechanisms_4751f51a-1827-11e9-8fa1-8a76d5e8efbc2.1.9DoDDoD should pursue and expand the use of alternative acquisition mechanisms as a workaround for cumbersome procurement policies.For example, the 2016 National Defense Authorization Act granted DoD the permanent Other Transaction Authority (OTA), which allowed DoD to circumvent the traditional acquisition process in certain circumstances.RelationshipsFoster better relationships between the defense community and the U.S. technology industry_4751f95c-1827-11e9-8fa1-8a76d5e8efbc2.1.10DoDDefense CommunityU.S. Technology IndustryDoD should foster better relationships between the defense community and the U.S. technology industry, such as by expanding industry outreach efforts like DoD’s Defense Innovation Unit Experimental (DIUx) designed to make it faster for DoD to take advantage of emerging commercial technologies. There should be specific emphasis on creating greater incentives for technology firms to work with DoD.DIUxExpand DoD’s Defense Innovation Unit Experimental (DIUx)_4751fba0-1827-11e9-8fa1-8a76d5e8efbc2.1.10.1Defense Innovation Unit Experimental (DIUx)IncentivesCreate greater incentives for technology firms to work with DoD._4751ff6a-1827-11e9-8fa1-8a76d5e8efbc2.1.10.2Technology FirmsProgram ElementEstablish a Program Element (PE) for AI_475202b2-1827-11e9-8fa1-8a76d5e8efbc2.1.11DoDDoD should establish a new Program Element (PE) specifically for AI to increase the visibility of AI appropriations.Spending PriorityPrioritize the development and adoption of AI in defense spending_475204ec-1827-11e9-8fa1-8a76d5e8efbc2.1.12CongressCongress should prioritize the development and adoption of AI in defense spending.This could entail either focusing greater attention on AI projects as compared to less-important work or increasing overall spending.National SecuritySupport productive conversations about the appropriate way to oversee the use of AI for national security_475208f2-1827-11e9-8fa1-8a76d5e8efbc2.1.13CongressCongress and the administration should support productive conversations about the appropriate way to oversee the use of AI for national security. The AdministrationThis will include rejecting bans on lethal autonomous weapons (LAWS) and differentiating between concerns about LAWS specifically and broader concerns about different military activities, which are often the underlying concern in these discussions.LAWSRecognize that the benefits of AI to national security are too important to let concerns about LAWS oversight or other defense activities involving AI limit national security AI support and adoption_47520bc2-1827-11e9-8fa1-8a76d5e8efbc2.1.14CongressCongress and the administration should recognize that the benefits of AI to national security are too important to let concerns about LAWS oversight or other defense activities involving AI limit national security AI support and adoption.The AdministrationThis is particularly important because foreign adversaries will pay little heed to such oversight concerns and gain a competitive advantage in certain areas of AI.IndustrySpur AI development and adoption in industry, including through sector-specific AI strategies. _47520e10-1827-11e9-8fa1-8a76d5e8efbc3IndustryThe federal government has significant influence and involvement in sectors such as health care, transportation, and education through funding, procurement, and regulation. Federal agencies should be charged with developing sector-specific AI strategies to shape their policies affecting these industries in ways that support AI transformation. Additionally, many nations look to AI as an important industry for future competitiveness. They are putting in place a host of development policies designed to grow their domestic AI industries. The United States needs to do the same.Sector StrategiesCreate strategies for supporting AI adoption in relevant sectors_475211ee-1827-11e9-8fa1-8a76d5e8efbc3.1Federal AgenciesFederal agencies should work with industry to create strategies for supporting AI adoption in relevant sectors of the economy.Industry SectorsThese strategies should provide guidance about how best to leverage AI to advance agency missions, as well as identify opportunities to encourage AI adoption in relevant industries, such as by proactively providing guidance on policy questions, ensuring that procurement supports AI, ensuring regulations do not limit AI usage, and creating incentives for firms to invest in AI. These strategies should be updated regularly as agencies become more familiar with the technology and as AI matures, creating new challenges and opportunities to address.Agency MissionsProvide guidance about how best to leverage AI to advance agency missions_475214dc-1827-11e9-8fa1-8a76d5e8efbc3.1.1Industry OpportunitiesIdentify opportunities to encourage AI adoption in relevant industries_4752172a-1827-11e9-8fa1-8a76d5e8efbc3.1.2IndustriesPolicy GuidanceProactively provide guidance on policy questions_47521b08-1827-11e9-8fa1-8a76d5e8efbc3.1.2.1ProcurementEnsure that procurement supports AI_47521dec-1827-11e9-8fa1-8a76d5e8efbc3.1.2.2RegulationsEnsure regulations do not limit AI usage_47522238-1827-11e9-8fa1-8a76d5e8efbc3.1.2.3IncentivesCreate incentives for firms to invest in AI_47522634-1827-11e9-8fa1-8a76d5e8efbc3.1.2.4FirmsUpdatesUpdate the strategies regularly as agencies become more familiar with the technology and as AI matures_4752299a-1827-11e9-8fa1-8a76d5e8efbc3.1.3Sector OrganizationsEstablish organizations designed to advance the development of innovative AI applications in various sectors_47522bf2-1827-11e9-8fa1-8a76d5e8efbc3.2Department of CommerceThe Department of Commerce should establish organizations designed to advance the development of innovative AI applications in various sectors.Manufacturing USAFor example, Manufacturing USA, overseen by federal agencies including the Department of Commerce and the Department of Energy, is a network of research institutes focused on fostering innovation and collaboration in the manufacturing sector.Department of EnergyAdvanced Robotics for ManufacturingAmong them is the Advanced Robotics for Manufacturing, a public-private partnership in Pittsburgh which focuses on AI and automation.InstitutesUsing this model, agencies should support similar institutes that include industry, academia, and government agency resources to advance AI in other sectors such as city management and precision medicine.IndustryAcademiaGovernment AgenciesCity ManagersPrecision Medicine PractitionersState GovernmentsEnable state governments to foster AI industry development_47522fd0-1827-11e9-8fa1-8a76d5e8efbc3.3CongressCongress should direct the Economic Development Administration to enable state governments to foster AI industry development.Economic Development AdministrationState GovernmentsCongress should appropriate funds for the Economic Development Administration to create a state economic development competition in which states would compete for funds to establish their own state development plans and policies for supporting AI development, especially through new startups. Trade PoliciesSupport digital free trade policies._475232c8-1827-11e9-8fa1-8a76d5e8efbc4Ensuring Trade Policy Supports AI -- Data is at the core of AI, and many nations are enacting policies that restrict cross-border data flows. The U.S. government needs to accelerate its efforts to establish free trade in data and fight other protectionist efforts that inhibit AI, such as source code disclosure requirements.Cross-Border DataAdvocate for cross-border data flow protections_4752352a-1827-11e9-8fa1-8a76d5e8efbc4.1United States Trade Representative (USTR)The United States Trade Representative (USTR) should continue to advocate for cross-border data flow protections in all future trade negotiations.Source CodeFight source code disclosure requirements_475239d0-1827-11e9-8fa1-8a76d5e8efbc4.2USTRUSTR should continue to fight source code disclosure requirements other nations may enact to unfairly disadvantage U.S. firms or exploit their intellectual property. U.S. FirmsRegulationFoster innovation-friendly regulation._47523d22-1827-11e9-8fa1-8a76d5e8efbc5Ensuring Any Regulation of AI Is Innovation Friendly -- If poorly implemented, AI can produce undesirable outcomes. In response, some have called for strong regulations on AI, including through tougher enforcement of antitrust and regulation of algorithms. Most proposals thus far would harm AI innovation and use, often without providing meaningful protections. Policymakers should instead pursue a more innovation-friendly framework built around the principle of “algorithmic accountability,” in which the operators of algorithms are held accountable for explicit and serious harms. Additionally, antitrust regulators should resist viewing the possession of large amounts of data as a threat to competition.AccountabilityEncourage adherence to the principle of algorithmic accountability_47523fe8-1827-11e9-8fa1-8a76d5e8efbc5.1RegulatorsRegulators should encourage adherence to the principle of algorithmic accountability.RiskMinimize risk_47524470-1827-11e9-8fa1-8a76d5e8efbc5.1.1PolicymakersImportantly, policymakers need to recognize that the goal of algorithmic accountability is not to achieve perfect, error-free algorithms, but to minimize risk—just as vehicle safety standards do not require cars to be 100 percent safe, but as reasonably safe as can be expected. Recognition & OversightRecognize the framework for algorithmic accountability and integrate it into oversight_4752497a-1827-11e9-8fa1-8a76d5e8efbc5.1.2RegulatorsThe most important step is for regulators to formally recognize this framework for algorithmic accountability and integrate it into their oversight. This applies to both domain-specific and consumer-protection regulators.MandatesReject blanket mandates for algorithms_47524c54-1827-11e9-8fa1-8a76d5e8efbc5.2CongressCongress should reject blanket mandates for algorithms, such as algorithmic transparency requirements, or the creation of new regulatory bodies focused only on regulating algorithms.Technical ExpertiseIncrease the technical expertise of regulators and policymakers_47526716-1827-11e9-8fa1-8a76d5e8efbc5.3CongressCongress and the administration should support increasing the technical expertise of regulators and policymakers.RegulatorsRegulators should foster relationships with communities of developers, academics, civil society groups, and private-sector organizations invested in algorithmic decisionmaking to stay abreast of technical developments and concerns about algorithmic harms that could influence how algorithmic accountability is achieved or enforced.PolicymakersAlgorithmic Decision MakersAI DevelopersAcademicsCivil Society GroupsPrivate-Sector OrganizationsThis requires ensuring regulators have the resources to hire staff with the necessary technical expertise to scrutinize algorithms.Large DatasetsAvoid considering the act of collecting or possessing large amounts of data as potentially anticompetitive behavior_47526b62-1827-11e9-8fa1-8a76d5e8efbc5.4CongressCongress and the administration should caution regulators against viewing the mere act of collecting or possessing large amounts of data (which is necessary for certain uses of AI) as potentially anticompetitive behavior.The AdministrationPrivacyReject overly stringent rules in privacy legislation_47527008-1827-11e9-8fa1-8a76d5e8efbc5.5CongressCongress should reject overly stringent rules in privacy legislation, such as broad opt-in requirement, purpose specification, data erasure, and data minimization, as well as other policies modeled on the EU’s General Data Protection regulation.EUData InvestmentEncourage companies to invest in collecting data_47527580-1827-11e9-8fa1-8a76d5e8efbc5.6CongressCongress and the administration should emphasize that data is a crucial business input for the development of AI, and that companies should be encouraged to invest in collecting data, not punished for it.The AdministrationCompanies Collection & AvailabilityMake it easier to collect data and ensure data is readily available_47527940-1827-11e9-8fa1-8a76d5e8efbc5.6.1PolicymakersIf policymakers are concerned that startups and small businesses cannot access the data necessary to develop AI and compete with larger firms, they should focus on making it easier to collect data and ensure data is readily available, as described earlier in this report, rather than penalize a company that has succeeded in collecting and making data available.StartupsSmall BusinessesWorkforce TransitionsProvide workers with better tools to manage AI-driven workforce transitions._47527c4c-1827-11e9-8fa1-8a76d5e8efbc6WorkersAI-enabled automation will increase productivity and per-capita incomes but also will likely modestly increase the rate of worker displacement, which can lead to support for policies that restrict how firms can use AI. To help workers more effectively make transitions, policymakers need to modernize workforce training and worker dislocation policies and programs.TalentDevelop AI Talent_4752814c-1827-11e9-8fa1-8a76d5e8efbc6.1InvestmentInvest in AI talent_4752852a-1827-11e9-8fa1-8a76d5e8efbc6.1.1CongressCongress should invest in cultivating AI talent. Countries such as Canada and the United Kingdom have launched initiatives to do just that, and the United States should adapt these approaches. CanadaFor example, Canada’s AI strategy funds the creation of AI research institutes, programs to attract and retain AI talent in Canadian universities.United KingdomSimilarly, the United Kingdom’s AI Sector Deal describes how the government will fund at least 1,000 AI PhD students by 2025.National Science Foundation (NSF)Congress should fund and direct the National Science Foundation (NSF) to create a competitive AI fellowship program for at least 1,000 computer science students annually.Computer Science StudentsResearchAuthorize and fund a program to provide competitive awards for academic AI researchers_475287be-1827-11e9-8fa1-8a76d5e8efbc6.1.2CongressCongress should fund and authorize a program at the National Science Foundation to provide competitive awards for up to 1,000 academic AI researchers for a period of five years. Awards should be conditional on remaining in academia for five years.National Science FoundationUniversitiesThough individual businesses may benefit from attracting the best AI talent away from universities, the overall AI innovation ecosystem in the United States suffers.AI ResearchersThese awards would incentivize more AI researchers to stay in academia and help U.S. universities meet the demand for AI skills.H-1B VisasEnable more foreign AI talent to work in the United States_47528c32-1827-11e9-8fa1-8a76d5e8efbc6.1.3CongressCongress should enable more foreign AI talent to work in the United States by increasing the cap on H-1B visas to ensure U.S. firms can hire as much AI talent as they need. To the extent global AI talent works in the United States, they are not working for competitors in other nations.U.S. FirmsComputer Science CoursesAddress barriers that limit the number of students able to take computer science courses_47528f84-1827-11e9-8fa1-8a76d5e8efbc6.1.4Federal AgenciesFederal agencies should address barriers that limit the number of students able to take computer science courses at the university level.NSFNSF should provide grants to colleges and universities that have increased or are implementing programs to increase enrollment and retention in computer science. The federal government should also require increased transparency as a prerequisite for receiving NSF awards.CollegesUniversitiesSchoolsFor instance, schools should be required to monitor and disclose the number of computer science applicants, prospective majors, and their retention rates in computer science subjects. Ideally this would be done for all STEM disciplines.Training & AdjustmentImplement comprehensive reforms to the nation’s workforce training and adjustment policies_47529222-1827-11e9-8fa1-8a76d5e8efbc6.2CongressCongress and the administration should implement comprehensive reforms to the nation’s workforce training and adjustment policies, as the Information Technology and Innovation Foundation (ITIF) outlined in its February 2018 report, “How to Reform Worker-Training and Adjustment Policies for an Era of Technological Change.”The AdministrationInformation Technology and Innovation Foundation (ITIF)Providing Workers with Better Tools to Manage AI-driven Workforce Transitions2018-12-042019-01-14http://www2.datainnovation.org/2018-national-ai-strategy.pdfOwenAmburOwen.Ambur@verizon.net