Rise of the Machines Artificial Intelligence and its Growing Impact on U S Policy Chairman Will Hurd and Ranking Member Robin Kelly Subcommittee on Information Technology Committee on Oversight and Government Reform U S House of Representatives September 2018 EXECUTIVE SUMMARY Beginning in February of 2018 the Subcommittee on Information Technology of the House Committee on Oversight and Government Reform held a series of hearings on artificial intelligence AI In connection with those hearings majority and minority staff met jointly with experts from academia industry and government and reviewed multiple reports from leading AI experts Through these efforts several points became evident First AI is an immature technology its abilities in many areas are still relatively new Second the workforce is affected by AI whether that effect is positive negative or neutral remains to be seen Third AI requires massive amounts of data which may invade privacy or perpetuate bias even when using data for good purposes Finally AI has the potential to disrupt every sector of society in both anticipated and unanticipated ways In light of that potential for disruption it’s critical that the federal government address the different challenges posed by AI including its current and future applications The following paper presents lessons learned from the Subcommittee’s oversight and hearings on AI and sets forth recommendations for moving forward Underlying these recommendations is the recognition the United States cannot maintain its global leadership in AI absent political leadership from Congress and the Executive Branch Therefore the Subcommittee recommends increased engagement on AI by Congress and the Administration Chief among the Subcommittee’s recommendations is for the federal government to increase federal spending on research and development to maintain American leadership with respect to AI In response to concerns about AI’s potential economic impact federal state and local agencies are encouraged to engage more with stakeholders on the development of effective strategies for improving the education training and reskilling of American workers to be more competitive in an AI-driven economy The Subcommittee further recommends the federal government lead by example by investing more in education and training programs that would allow for its current and future workforce to gain necessary AI skills In response to concerns about privacy the Subcommittee recommends federal agencies review federal privacy laws and regulations to determine how they may already apply to AI technologies within their jurisdiction and where necessary update existing regulations to account for the addition of AI To account for potential biases in AI systems federal state and local agencies that use AI systems to make consequential decisions about people should ensure that the algorithms that support these systems are accountable and inspectable Finally any regulatory approach to AI should consider whether the risks to public safety or consumers already fall within any existing regulatory frameworks and if so whether those existing frameworks can adequately address the risks Where a risk falls outside an existing regulatory framework an approach should consider whether modifications or additions are needed to better account for the addition of AI 1 As AI technology continues to advance its progress has the potential to dramatically reshape the nation’s economic growth and welfare It is critical the federal government build upon and increase its capacity to understand develop and manage the risks associated with this technology’s increased use 2 INTRODUCTION Defining AI This paper defines AI as computational technology that works and reacts in humanlike ways AI generally falls into two categories “narrow AI” and “general AI ” Narrow AI addresses or solves specific tasks “such as playing strategic games language translation self-driving vehicles and image recognition ” 1 General AI on the other hand can accomplish more than one task and can move between these tasks based on reasoning 2 Witnesses who testified before the Subcommittee suggested that while narrow AI is commonly utilized today “more general systems that can work across multiple tasks” are underdeveloped at this time 3 The examples of AI that are referred to in this paper concern the field of narrow AI Brief Background on AI The idea of AI first emerged in 1950 with “Computing Machinery and Intelligence ” Alan Turing’s seminal paper addressing the question of whether machines can think 4 Turing’s paper also set forth a test for answering this question and highlighted the issue of whether machines could be developed to learn from experiences similar to the way people do 5 The term “artificial intelligence” was later coined in 1956 by John McCarthy after holding the first academic conference on the topic 6 The last ten years have seen the most significant developments in AI largely due to advancements in computing power and increased access to data Current Uses of AI While AI is most closely associated with Silicon Valley and Hollywood various industries have already deployed the technology For example AI is now used in connection 1 EXECUTIVE OFFICE OF THE PRESIDENT NATIONAL SCIENCE AND TECHNOLOGY COUNCIL COMMITTEE ON TECHNOLOGY Preparing for the Future of Artificial Intelligence Oct 2016 online at https obamawhitehouse archives gov sites default files whitehouse_files microsites ostp NSTC preparing_for_the_ future_of_ai pdf see also GOVERNMENT ACCOUNTABILITY OFFICE Artificial Intelligence Emerging Opportunities Challenges and Implications GAO-18-142S March 2018 online at https www gao gov assets 700 690910 pdf 2 Id 3 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Oren Etzioni CEO Allen Institute for Artificial Intelligence 4 Alan M Turing Computing Machinery and Intelligence 49 Mind 433-460 1950 online at https www csee umbc edu courses 471 papers turing pdf last accessed Sept 14 2018 5 Id 6 Gil Press Artificial Intelligence Defined as a New Research Discipline This Week in Tech History FORBES Aug 28 2016 online at https www forbes com sites gilpress 2016 08 28 artificial-intelligence-defined-as-a-newresearch-discipline-this-week-in-tech-history #6913216e6dd1 3 with mapping applications or “apps” on mobile phones 7 tax preparation 8 song writing 9 and digital advertising 10 It is also being used in video games and movies to create special effects 11 More recently the Food and Drug Administration approved an AI algorithm that aids radiologists in detecting wrist fractures 12 The State of Ohio uses robotics in the Bureau of Criminal Investigation laboratories to help reduce the turnaround time on untested rape kits 13 The application of AI facilitated the state testing 14 000 previously untested rape kits and identifying 300 serial rapists linked to 1 100 crimes 14 While today’s AI products and applications are largely limited to solving specific discrete tasks because of its current and potential benefits AI has generated broad interest across the economy and as a result will likely be a key driver of future economic growth and progress IMPORTANCE OF U S LEADERSHIP IN AI The United States has traditionally led the world in the development and application of AI-driven technologies 15 This is due in part to the government’s prior commitment to investing heavily in research and development R D that has in turn helped support AI’s growth and development In 2015 for example the United States led the world in total gross domestic R D expenditures spending $497 billion 16 The Subcommittee’s hearings however highlighted the fact that the United States’ leadership in AI is no longer guaranteed During the hearings several of the witnesses testified 7 Nick Statt AI is Google’s secret weapon for remaking its oldest and most popular apps THE VERGE May 10 2018 online at https www theverge com 2018 5 10 17340004 google-ai-maps-news-secret-weapon-remakingold-apps-products-io-2018 8 Adelyn Zhou EY Deloitte and PwC Embrace Artificial Intelligence For Tax and Accounting FORBES Nov 14 2017 online at https www forbes com sites adelynzhou 2017 11 14 ey-deloitte-and-pwc-embrace-artificialintelligence-for-tax-and-accounting #2ad76d253498 9 Matt Jancer More Artists are Writing Songs in the Key of AI WIRED May 17 2018 online at https www wired com story music-written-by-artificial-intelligence 10 Tom Simonite Google and Microsoft Can Use AI to Extract Many More Ad Dollars from Our Clicks FORBES Aug 31 2017 online at https www wired com story big-tech-can-use-ai-to-extract-many-more-ad-dollars-fromour-clicks 11 Cade Metz Lights Camera Artificial Action Start-Up is Taking A I to the Movies NEW YORK TIMES March 26 2018 online at https www nytimes com 2018 03 26 technology artificial-intelligence-hollywood html 12 FOOD AND DRUG ADMINISTRATION FDA permits marketing of artificial intelligence algorithm for aiding providers in detecting wrist fractures May 24 2018 online at https www fda gov NewsEvents Newsroom PressAnnouncements ucm608833 htm 13 OHIO OFFICE OF THE ATTORNEY GENERAL Law Enforcement Bureau of Criminal Investigations Laboratory Division online at https www ohioattorneygeneral gov Law-Enforcement Bureau-of-CriminalInvestigation Laboratory-Division last visited August 22 2018 14 Anna Boiko-Weyrauch Her Rapist was Convicted because of a Rape Kit So Why are So Many Untested KUOW August 15 2018 online at http kuow drupal publicbroadcasting net post her-rapist-was-convictedbecause-rape-kit-so-why-are-so-many-kits-untested 15 McKinsey Global Institute Artificial Intelligence Implications for China April 2017 online at https www mckinsey com media McKinsey Featured%20Insights China Artificial%20intelligence%20Implicati ons%20for%20China MGI-Artificial-intelligence-implications-for-China ashx 16 NATIONAL SCIENCE BOARD 2018 SCIENCE ENGINEERING INDICATORS 2018 online at www nsf gov statistics 2018 nsb20181 assets nsb20181 pdf 4 the United States needs to increase its R D spending to remain competitive in the field of AI One witness stated “ w hile other governments are aggressively raising their research funding U S government research has been relatively flat ” 17 Another witness expressly warned “ c urrent federal funding levels are not keeping pace with the rest of the industrialized world ” 18 Notably China’s commitment to funding R D has been growing sharply up 200 percent from 2000 to 2015 19 On February 7 2018 the National Science Board’s Board and the National Science Foundation’s NSF Director who jointly head NSF said in a statement that if current trends continue the Board expects “China to pass the United States in R D investments” by the end of 2018 20 Recent progress was made when the Defense Advanced Research Projects Agency DARPA announced the creation of the Artificial Intelligence Exploration program “AI Next ” to bolster the United States’ leadership in AI DARPA plans to invest more than $2 billion into this program and other existing programs The program focuses research on “‘third wave’ AI theory and application that will make it possible for machines to contextually adapt to changing situations ” 21 17 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ian Buck Vice President and General Manager Tesla Data Center Business NVIDIA 18 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Amir Khosrowshahi Vice President and Chief Technology Officer Artificial Intelligence Group Intel Corp 19 NATIONAL SCIENCE BOARD 2018 SCIENCE ENGINEERING INDICATORS online at www nsf gov statistics 2018 nsb20181 assets nsb20181 pdf 20 NATIONAL SCIENCE FOUNDATION NATIONAL SCIENCE BOARD STATEMENT ON GLOBAL RESEARCH AND DEVELOPMENT R D INVESTMENTS NSB-2018-9 Feb 7 2018 online at www nsf gov nsb news news_summ jsp cntn_id 244465 last accessed September 17 2018 21 Defense Advanced Research Projects Agency DARPA Announces $2 Billion Campaign to Develop Next Wave of AI Technologies Sept 7 2018 online at https www darpa mil news-events 2018-09-07 5 The chart on the previous page shows China’s rapidly growing investment in AI Particularly concerning is the prospect of an authoritarian country such as Russia or China overtaking the United States in AI As the Subcommittee’s hearings showed AI is likely to have a significant impact in cybersecurity and American competitiveness in AI will be critical to ensuring the United States does not lose any decisive cybersecurity advantage to other nationstates 22 The loss of American leadership in AI could also pose a risk to ensuring any potential use of AI in weapons systems by nation-states comports with international humanitarian laws In general authoritarian regimes like Russia and China have not been focused on the ethical implications of AI in warfare and will likely not have guidelines against more bellicose uses of AI such as in autonomous weapons systems 23 The United States’ competitiveness in AI is also critical to its economic security as AI is poised to be a key driver of economic growth AI applications promise to make industry more efficient—cutting down costs limiting the use of natural resources and improving the use of finite resources such as “the increasingly crowded electromagnetic spectrum ” 24 For example the Government Services Administration has a robotic processing automation RPA pilot that automates portions of the Multiple Award Schedules new offer review process Presently contract officers must go through a tedious administrative process reading through dozens of pages of documentation across multiple IT systems to ensure a vendor’s new offer is consistent with information already in government databases RPA software offers the capability to perform these tasks so the contract officers can spend more time engaging with customers 25 In an effort to keep the United States at the forefront of AI developments and advancements the Obama Administration released three reports in 2016 that assessed the state of AI The reports focused on the public policy questions AI raises for the country and proposed a series of recommendations 26 The current Administration should prioritize building on these 22 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ben Buchanan Postdoctoral Fellow Belfer Center Cyber Security Project Science Technology and Public Policy Program Harvard University Dr Buchanan currently serves as an Assistant Teaching Professor at Georgetown University’s School of Foreign Service 23 CENTER FOR A NEW AMERICAN SECURITY Autonomous Weapons online at https www cnas org research technology-and-national-security autonomous-weapons accessed June 19 2018 24 Spectrum Collaboration Challenge What is the Spectrum Collaboration Challenge DEFENSE ADVANCED RESEARCH PROJECTS AGENCY available at https spectrumcollaborationchallenge com about accessed Aug 15 2018 25 Game Changers Artificial Intelligence Part II Artificial Intelligence and the Federal Government Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Mr Keith Nakasone Deputy Assistant Commissioner Category Acquisition Management U S General Services Administration 26 EXECUTIVE OFFICE OF THE PRESIDENT NATIONAL SCIENCE AND TECHNOLOGY COUNCIL COMMITTEE ON TECHNOLOGY Preparing for the Future of Artificial Intelligence Oct 2016 online at https obamawhitehouse archives gov sites default files whitehouse_files microsites ostp NSTC preparing_for_the_ 6 efforts and ensure federal agencies are implementing these recommendations Recent efforts by the Trump Administration recognize AI’s growing importance In particular the Administration has highlighted AI in the 2017 National Security Strategy 2018 Summary of the National Defense Strategy the Office of Management and Budget OMB and Office of Science and Technology Policy’s OSTP FY2019 Administration Research and Development Budget Priorities and the FY2019 Budget Request OSTP also held a summit on the topic in May 2018 and established the National Science and Technology Council Select Committee on AI 27 As the United States considers the impact of AI in both foreign and economic affairs we must keep in mind that American values—the right to privacy free speech the rule of law and respect for intellectual property—give America many intangible advantages over other nations As Subcommittee Chairman Hurd said in June “the United States boasts a creative risk-taking culture that is inextricably linked to its free enterprise system ” 28 Gary Shapiro testified before the Subcommittee “ the United States’ competitive strength is innovation ” 29 Throughout the hearings witnesses repeatedly acknowledged these advantages are important components driving American leadership in AI’s development and deployment We cannot presume that these values will automatically secure our leadership but we should be aware of them and regard them as strengths AI’S CHALLENGES The Subcommittee’s hearings showed AI faces a number of different challenges In this paper we will address the four challenges discussed during the hearings—workforce privacy bias and malicious use of AI Workforce One of the central concerns raised during the hearings is that AI advancement in the short-term could lead to the loss of jobs due to AI-driven automation 30 A December 2017 report future_of_ai pdf EXECUTIVE OFFICE OF THE PRESIDENT NATIONAL SCIENCE AND TECHNOLOGY COUNCIL COMMITTEE ON TECHNOLOGY NETWORKING AND INFORMATION TECHNOLOGY RESEARCH AND DEVELOPMENT SUBCOMMITTEE The National Artificial Intelligence Research and Development Strategic Plan Oct 2016 online at https obamawhitehouse archives gov sites default files whitehouse_files microsites ostp NSTC national_ai_rd_stra tegic_plan pdf EXECUTIVE OFFICE OF THE PRESIDENT Artificial Intelligence Automation and the Economy Dec 2016 online at https obamawhitehouse archives gov sites whitehouse gov files documents Artificial-IntelligenceAutomation-Economy PDF 27 EXECUTIVE OFFICE OF THE PRESIDENT OFFICE OF SCIENCE AND TECHNOLOGY POLICY SUMMARY OF THE 2018 WHITE HOUSE SUMMIT ON ARTIFICIAL INTELLIGENCE FOR AMERICAN INDUSTRY 2018 online at https www whitehouse gov wp-content uploads 2018 05 Summary-Report-of-White-House-AI-Summit pdf 28 Will Hurd Washington Needs to Adopt AI soon or We’ll Lose Millions FORTUNE June 12 2018 online at http fortune com 2018 06 12 rep-will-hurd-artificial-intelligence 29 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Mr Gary Shapiro CEO of the Consumer Technology Association 30 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 7 from the McKinsey Global Institute reported that as a result of AI-driven automation “up to 1 3 of the workforce in the United States and Germany may need to find work in new occupations ” 31 Another study released by Oxford University in 2013 found the impact on U S workers by AI technologies may even be higher According to the Oxford study “about 47 percent of total U S employment is at risk ” 32 These studies indicate the negative impact AI may have on jobs which has the potential to increase wealth inequality in the United States Some of the hearing witnesses and other studies however show AI has the capacity to improve and increase jobs 33 For example the Organisation for Economic Co-operation and Development OECD released a study in March 2018 finding that about fourteen percent of jobs in OECD countries have a high risk of automation 34 Additionally a Deloitte study of 140 years of census data in the United Kingdom examined the effect of new technologies on jobs and found in spite of new technologies more jobs were ultimately created than eliminated 35 The common thread from all of these studies is that our economic policies must take into account the uncertain future of work faced by Americans as AI takes hold and the need for increased investments in education and worker retraining As such federal state and local agencies should be encouraged to engage more with educators employers employees unions and other stakeholders on the development of effective strategies for improving the education training and reskilling of American workers to be more competitive in an AI-driven economy The federal government should also lead by example by investing more in education and training programs that would allow for its current and future workforce to gain the necessary AI skills Privacy statement of Ben Buchanan Postdoctoral Fellow Belfer Center Cyber Security Project Science Technology and Public Policy Program Harvard University 31 McKinsey Global Institute Jobs Lost Jobs Gained Workforce Transitions In a Time of Automation Dec 2017 online at www mckinsey com media McKinsey Global%20Themes Future%20of%20Organizations What%20the%20futur e%20of%20work%20will%20mean%20for%20jobs%20skills%20and%20wages MGI-Jobs-Lost-Jobs-GainedReport-December-6-2017 ashx 32 CARL BENEDIKT FREY MICHAEL A OSBORNE The Future of Employment How Susceptible are Jobs to Computerization Comment Uni Oxford Sept 17 2013 online at www oxfordmartin ox ac uk downloads academic The_Future_of_Employment pdf 33 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Mr Gary Shapiro CEO of the Consumer Technology Association 34 Organisation for Economic Co-operation and Development Putting faces to the jobs at risk of automation POLICY BRIEF ON THE FUTURE OF WORK March 2018 online at https www oecd org employment Automationpolicy-brief-2018 pdf 35 IAN STEWART DEBAPRATIM DE ALEX COLE Technology and People The Great Job-Creating Machine Deloitte August 2015 online at https www2 deloitte com uk en pages finance articles technology-andpeople html last accessed September 14 2018 8 AI technologies rely on computer algorithms that use data to determine how they will respond to new inputs 36 Since AI requires vast amounts of data witnesses in the Subcommittee’s hearings frequently cited individuals’ privacy as a potential challenge For example Dr Ben Buchanan an AI expert testified about the various privacy risks consumers face when their personal data is used in AI systems 37 According to Dr Buchanan “ t here is the risk of breaches by hackers of misuse by those who collect it or access it and of secondary use—in which data collected for one purpose is later re-appropriated for another ” 38 These fears were realized when one of the nation’s largest credit reporting agencies Equifax Inc announced in 2017 that hackers had successfully penetrated its systems and gained access to the personal data of approximately 145 5 million Americans 39 The data compromised included sensitive information Equifax had collected on consumers ranging from “Social Security numbers birth dates addresses and in some instances driver’s license numbers ” 40 Other examples where privacy concerns are implicated is the use of smart speakers that often feature voice-activated digital assistants For instance if a smart speaker is always listening in order to respond to a command what is done with the hours of data it overhears Law enforcement has already begun requesting voice recordings from technology companies for criminal cases This raises the question of how devices that are always listening could potentially be used as evidence against consumers 41 To address the different privacy challenges raised by AI-driven technologies one witness Dr Buchanan testified before the Subcommittee that companies need to adopt more stringent safeguards in the design and development of their AI systems 42 However other 36 Tom Simonite The Wired Guide to Artificial Intelligence WIRED Feb 1 2018 online at www wired com story guide-artificial-intelligence 37 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ben Buchanan Postdoctoral Fellow Belfer Center Cyber Security Project Science Technology and Public Policy Program Harvard University 38 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ben Buchanan Postdoctoral Fellow Belfer Center Cyber Security Project Science Technology and Public Policy Program Harvard University 39 Equifax Equifax Announces Cybersecurity Firm Has Concluded Forensic Investigation of Cybersecurity Incident Potentially Impacted U S Consumers Increased by 2 5 Million Oct 2 2017 online at https investor equifax com news-and-events news 2017 10-02-2017-213238821 Equifax Equifax Announces Cybersecurity Incident Involving Consumer Information Sept 7 2017 online at https investor equifax com news-and-events news 2017 09-07-2017-213000628 40 Equifax Equifax Announces Cybersecurity Incident Involving Consumer Information Sept 7 2017 online at https investor equifax com news-and-events news 2017 09-07-2017-213000628 41 Jeff John Roberts Police Ask Amazon’s Echo to Help Solve a Murder FORTUNE Dec 27 2016 online at http fortune com 2016 12 27 amazon-echo-murder 42 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ben Buchanan Postdoctoral Fellow Belfer Center Cyber Security Project Science Technology and Public Policy Program Harvard University 9 witnesses such as Dr Oren Etzioni CEO of the Allen Institute for Artificial Intelligence 43 and Gary Shapiro President of the Consumer Technology Association 44 argued that rather than trying to regulate all AI-related privacy issues under one umbrella regulations should be tailored to individual AI applications Some AI products and applications may already be subject to federal privacy laws such as the Health Insurance Portability and Accountability Act Children’s Online Privacy Protection Act Gramm-Leach-Bliley Act or fall under the jurisdiction of the Federal Trade Commission the primary federal privacy regulator 45 The growing collection and use of personal data in AI systems and applications raises legitimate concerns about privacy As such federal agencies should review federal privacy laws regulations and judicial decisions to determine how they may already apply to AI products within their jurisdiction and—where necessary—update existing regulations to account for the addition of AI Biases The increasing reliance on AI to make consequential decisions about individuals has also heightened concerns about the technology’s accuracy particularly when used by governments As the Electronic Privacy Information Center EPIC explained in a statement to the Subcommittee “ w hen the government uses AI to make decisions about people it raises fundamental questions about accountability due process and fairness ” 46 At its core AI is reliant upon data If the data itself is incomplete biased or skewed in some other fashion the AI system is at risk of being inaccurate 47 As AI systems rely upon larger and larger quantities of data the risk increases that the data sets may knowingly or unknowingly 43 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Oren Etzioni CEO Allen Institute for Artificial Intelligence 44 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Mr Gary Shapiro CEO of the Consumer Technology Association 45 See Health Insurance Portability and Accountability Act of 1996 Pub L No 104-191 110 Stat 1936 1996 codified at 42 U S C § 300gg and 29 U S C § 1181 et seq and 42 U S C 1320d et seq Children’s Online Privacy Protection Act of 1998 Pub L No 105-277 112 Stat 2681-728 2000 codified at 15 U S C §§ 6501– 6506 Federal Services Modernization Act of 1999 Gramm-Leach-Bliley Act Pub L No 106-102 113 Stat 1338 1999 codified at 15 U S C §§6801-6809 §§ 6821-6827 and at 12 U S C §1831u et seq 46 Letter from Electronic Privacy Information Center to Rep William Hurd Chairman Subcommittee on Information Technology House Oversight and Government Reform Committee and Rep Robin Kelly Ranking Member Subcommittee on Information Technology House Oversight and Government Reform Committee April 19 2018 online at https www epic org testimony congress EPIC-HCOGR-AI-Apr2018 pdf 47 Tom Simonite The Wired Guide to Artificial Intelligence Wired Feb 1 2018 online at www wired com story guide-artificial-intelligence National Science and Technology Council Preparing for the Future of Artificial Intelligence Oct 2016 online at obamawhitehouse archives gov sites default files whitehouse_files microsites ostp NSTC preparing_for_the_future_ of_ai pdf 10 contain biases There are legitimate concerns that if an AI system is trained on biased data the AI system will produce biased results 48 For example in 2016 ProPublica began to investigate the use of computerized “risk prediction” tools by some judges in criminal sentencing and bail hearings The investigation found the algorithm the systems relied upon was racially biased and inaccurate 49 In total African-Americans were “almost twice as likely as whites to be labeled a higher risk ” despite the fact those African-Americans who had been labeled high risk were subsequently found to not go on to commit another crime 50 As AI technology is increasingly deployed into industries such as finance law and medicine these biases could be reinforced by the technology and harm populations During the Subcommittee’s hearings witnesses and other stakeholders also made clear one of the most effective ways in which bias or potential biases in AI systems can be addressed is by increasing transparency in the use of these systems As Dr Charles Isbell explained in his testimony An AI system should be inspectable The kind of data the algorithm uses to build its model should be available The decisions that such algorithms make should be inspectable In other words as we deploy these algorithms each algorithm should be able to explain its output 51 The solutions to addressing bias in AI are also cross-cutting For example Accenture recently introduced an “AI fairness tool ” which uses AI to examine how data influences variables such as age gender and race in a model 52 Civil society groups such as the Partnership on AI and the AI Now Institute are also researching and engaging in discussions around bias and AI 53 In short addressing biases and potential biases in AI systems will necessitate improvements in transparency when those systems are used to make consequential decisions about individuals Federal state and local agencies that use AI-type systems to make consequential decisions about people should ensure the algorithms supporting these systems are accountable 48 Forget Killer Robots – Bias is the Real AI Danger MIT TECHNOLOGY REVIEW Oct 3 2017 online at https www technologyreview com s 608986 forget-killer-robotsbias-is-the-real-ai-danger 49 Julia Angwin et al Machine Bias PROPUBLICA May 23 2016 online at www propublica org article machinebias-risk-assessments-in-criminal-sentencing 50 Id 51 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Dr Charles Isbell Senior Associate Dean College of Computing Georgia Institute of Technology 52 Natasha Lomas Accenture Wants to Beat Unfair AI with a Professional Toolkit TECHCRUNCH June 9 2018 online at https techcrunch com 2018 06 09 accenture-wants-to-beat-unfair-ai-with-a-professional-toolkit 53 AI NOW INSTITUTE https ainowinstitute org accessed on Aug 23 2018 PARTNERSHIP ON AI https partnershiponai org accessed on Aug 23 2018 11 and inspectable In addition federal state and local governments are encouraged to more actively engage with academic institutions non-profit organizations and the private sector in discussions on how to identify bias in the use of AI systems how best to eliminate bias through technology and how to account for bias Malicious Use of AI The Subcommittee’s hearings also highlighted the need to prepare for and protect against the malicious use of AI Earlier this year OpenAI a non-profit AI research company that testified at one of the hearings co-authored a report finding that unless adequate defenses are developed AI progress will result in cyberattacks that are “more effective more finely targeted more difficult to attribute and more likely to exploit vulnerabilities in AI systems ” 54 The report’s findings are consistent with those from a 2017 survey conducted by the cybersecurity firm Cylance According to the Cylance survey “62 percent of information security experts believe artificial intelligence will be used for cyberattacks in the coming year ” 55 AI’s computing power enables it to increase the severity of cyberattacks exponentially Three areas of malicious AI merit particular attention political security physical security and digital security 56 The common theme between these three spheres is AI expands the level of risk For example cyberattacks can affect more devices people and companies in a single moment undermining digital security In fact the United States has begun to see cyberattacks against it using AI 57 Consider the Russian disinformation campaigns of the past few years With AI “fake news” can be exponentially more convincing with the potential to create videos of people making statements that they never made—known as “deepfakes”—as featured on Radiolab’s Breaking News episode last summer 58 AI using data can pinpoint those who are most susceptible to disinformation with little human effort This capability has the potential to greatly expand the number of people impacted in any disinformation campaign employed by hostile nation-states to disrupt another state’s political system Physical security can also be compromised by AI systems For example a team of researchers from American universities showed how autonomous vehicles could be tricked by hackers into misinterpreting stop signs with a few simple stickers that would otherwise appear innocuous 59 54 Miles Brundage et al The Malicious Use of Artificial Intelligence Forecasting Prevention and Mitigation 2018 online at https www eff org files 2018 02 20 malicious_ai_report_final pdf 55 Cylance Black Hat Attendees See AI as Double-Edged Sword THREATMATRIX Aug 1 2017 online at threatmatrix cylance com en_us home black-hat-attendees-see-ai-as-double-edged-sword html 56 Miles Brundage et al The Malicious Use of Artificial Intelligence Forecasting Prevention and Mitigation 2018 available at https www eff org files 2018 02 20 malicious_ai_report_final pdf 57 The Morning Download First AI-Powered Cyberattacks are Detected WALL STREET JOURNAL Nov 16 2017 online at https blogs wsj com cio 2017 11 16 the-morning-download-first-ai-powered-cyberattacks-are-detected 58 RadioLab Breaking News WNYC STUDIOS July 27 2017 online at https www wnycstudios org story breaking-news 59 David Z Morris Researchers Show How Simple Stickers Could Trick Self-Driving Cars FORTUNE Sep 2 2017 online at http fortune com 2017 09 02 researchers-show-how-simple-stickers-could-trick-self-driving-cars 12 As we consider all the positive ways AI will be used in the future the government must also consider the ways it could be used to harm individuals and society and prepare for how to mitigate these harms How should the government respond to AI The U S government has traditionally taken a hands-off approach to emerging technologies During the 1990s the government’s preferred policy when it came to the internet was to favor private action over public regulation During that period although Congress and state legislatures passed several laws governing commerce content and competition they generally focused on crafting legal superstructures for the emerging web not micromanaging its development examples include the Internet Tax Freedom Act 60 and the Telecommunications Act of 1996 61 Concurrently the Clinton Administration adopted a policy of salutary neglect President Clinton ordered his administration to “do nothing that undermines the capacity of emerging technologies to lift the lives of ordinary Americans ” 62 It is the Subcommittee’s recommendation that the federal government should approach any potential regulation of AI with the above history in mind The government should begin by first assessing whether the risks to public safety or consumers already fall within existing regulatory frameworks and if so consideration should be made as to whether those existing frameworks can adequately address the risks If those risks fall outside the existing regulatory framework an approach should consider carefully whether modifications or additions are needed to better account for the addition of AI At minimum a widely agreed upon standard for measuring the safety and security of AI products and applications should precede any new regulations A common taxonomy also would help facilitate clarity and enable accurate accounting of skills and uses of AI The National Institute of Standards and Technology NIST is situated to be a key player in developing standards Similar private sector efforts exist from the Institute of Electrical and Electronics Engineers’ Global Initiative on Ethics of Autonomous and Intelligent Systems 63 The AI Index which is a part of Stanford’s “One Hundred Year Study on AI ” collects data about AI in order to track and measure its progress which will be critical in the standards development process to provide historical context 64 The federal government should look to support public academic and private sector efforts in the development of standards for measuring the safety and security of AI products and applications 60 Internet Tax Freedom Act of 1998 Pub L No 105-277 112 Stat 2681–719 amending 15 U S C § 151 1998 Telecommunications Act of 1996 Pub L No 104-104 110 Stat 56 codified as § 47 U S C 151 et seq 1996 62 Bill Clinton President and Al Gore Vice President Remarks by the President at and the Vice President at Electronic Commerce Event Nov 30 1998 online at https govinfo library unt edu npr library speeches rmkselec html last accessed September 17 2018 63 IEEE Standards Association The IEEE Global Initiative on Ethics and Autonomous and Intelligent Systems online at https standards ieee org develop indconn ec autonomous_systems html accessed on July 17 2018 64 Artificial Intelligence Index Our Mission online at http aiindex org #mission accessed on July 17 2018 61 13 There are also tangible ways for the federal government to facilitate greater AI innovation absent regulation For one the government holds thousands of different data sets that could be useful to researchers studying AI and to agencies in accomplishing their missions Many of these data sets are a public good and should be accessible to the public As such the Senate should work to pass the OPEN Government Data Act OGDA which the House passed unanimously last year 65 OGDA would allow for all non-sensitive government data to be made freely available and accessible to the public 66 During the Subcommittee’s hearings several witnesses testified AI’s development would be enhanced by more open data policies 67 Provided the Senate passes OGDA the President should sign it into law and the Administration should implement it as quickly as possible The Subcommittee also sees great value in encouraging innovation through national competitions There should be a Grand Challenge similar to DARPA’s Grand Challenges using data to solve a problem The benefit of DARPA’s Grand Challenges is their ability to foster innovative collaborative research among teams seeking to overcome seemingly unattainable goals Take for example DARPA’s Self-Driving Car Challenge which offered $1 million to the first team to autonomously navigate a desert course from California to Nevada In the first year of the Challenge no team completed the course In fact the farthest any vehicle went was 7 5 miles Yet eighteen months later 5 out of the 195 competing teams completed the 132-mile course with the winner having crossed the finish line in a little under seven hours 68 DARPA’s Grand Challenges provide strong incentives for innovation and as seen with its Self-Driving Cars Challenge can effectuate quick technological advancement Such competitions have spurred creativity research and collaboration leading to some of the most groundbreaking inventions in recent history Another recommendation universally supported by witnesses at the hearings was a need for more financial support for R D The Subcommittee is encouraged by the research already occurring at universities and not-for-profits as well as commercial R D The aggregate work of these bodies is why the United States is a leader in AI However to maintain American leadership there is a need for increased funding for R D at agencies like the National Science Foundation National Institutes of Health Defense Advanced Research Project Agency Intelligence Advanced Research Project Agency National Institute of Standards and Technology Department of Homeland Security and National Aeronautics and Space Administration As such the Subcommittee recommends the federal 65 Foundations for Evidence-Based Policymaking Act H R 4174 §201-202 115th Cong 1st Sess 2017 Open Public Electronic and Necessary OPEN Government Data Act H R 1770 S 760 115th Cong 1st Sess 2017 67 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Ian Buck Vice President and General Manager Tesla Data Center Business NVIDIA Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 statement of Amir Khosrowshahi Vice President and Chief Technology Officer Artificial Intelligence Group Intel Corp 68 Defense Advanced Research Projects Agency The DARPA Grand Challenge Ten Years Later March 13 2014 online at https www darpa mil news-events 2014-03-13 last accessed on September 17 2018 66 14 government provide for a steady increase in federal R D spending An additional benefit of increased funding is being able to support more graduate students which could serve to expand the future workforce in AI CONCLUSION AI has implications for every sector of industry and each Congressional committee should examine the role AI will play within their jurisdictions To date there have been six hearings held on AI three by this Subcommittee two by the Senate Committee on Commerce Science and Transportation in 2016 and 2017 and one by the House Science Space and Technology Committee in June 2018 69 More oversight action by Congress is needed There is also a pressing need for conscious direct and spirited leadership from the Trump Administration The 2016 reports put out by the Obama Administration’s National Science and Technology Council and the recent actions of the Trump Administration are steps in the right direction However given the actions taken by other countries—especially China— Congress and the Administration will need to increase the time attention and level of resources the federal government devotes to AI research and development as well as push for agencies to further build their capacities for adapting to advanced technologies The government has an essential role to play in securing American leadership in AI Fulfilling this role will require balancing the creative energy of innovative Americans whose knowledge and entrepreneurial spirit have driven the development of this technology with regulatory frameworks that protect consumers To ensure the appropriate balance is met it is vital Congress and the Executive Branch continue to educate themselves about AI increase the expenditures of R D funds help set the agenda for public debate and where appropriate define the role of AI in the future of this nation As our hearings have shown these steps are necessary for the United States to remain at the forefront of AI advancement 69 Game Changers Artificial Intelligence Part 1 Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 Game Changers Artificial Intelligence Part II Artificial Intelligence and the Federal Government Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 Game Changers Artificial Intelligence Part II1 Artificial Intelligence and Public Policy Hearing Before the Subcomm on Information Technology of the H Comm on Oversight and Government Reform 115th Cong 2018 The Dawn of Artificial Intelligence Hearing Before the Subcomm on Space Science and Competitiveness of the S Comm on Commerce Science Transportation 114th Cong 2016 Digital Decision-Making The Building Blocks of Machine Learning and Artificial Intelligence Hearing Before the Subcomm on Communications Technology Innovation and the Internet of the S Comm on Commerce Science Transportation 114th Cong 2017 Artificial Intelligence – With Great Power Comes Great Responsibility Hearing Before the Subcomm on Research and Technology and Subcomm on Energy the H Comm on Science Space and Technology Committee 115th Cong 2018 15