Queen s Global Markets. The Future of Artificial Intelligence

Transcription

1 Queen s Global Markets A PREMIER UNDERGRADUATE THINK-TANK The Future of Artificial Intelligence An Analysis of Its Economic and Social Impact Ethan Vera, Katie Russell, Endi Xherro, Kyle Bernat-Riddle, Nabil Rafat

2 Agenda What We Will be Discussing Today 1 What is Artificial Intelligence? 2 Current and Future Applications 3 Economic Impact 4 Security Considerations 5 Government Involvement 2

3 What is Artificial Intelligence? Artificial Intelligence (AI) refers to a set of computer science techniques that enable systems to perform tasks normally requiring human intelligence, such as visual interpretation, speech recognition, decision making and language translation. - The Economist Intelligence Unit Artificial Intelligence Taxonomy Subsectors of Artificial Intelligence Machine Learning and Deep Learning are branches of AI based on algorithms and complex data analysis, which enables computers to learn and adapt differently The repetitive analysis of large data sets allows computers to recognise patterns which are then extrapolated to perform a wide range of activities Today s use of AI is narrow it is used to perform one specific task, such as speech and facial recognition Sources: Bloomberg, Garter, Harvard Business Review, MIT Tech. Review, Softbank, The Economist Intelligence Unit 3

4 Current and Future Applications The 21 st Century s Most Revolutionary Technology is Poised to Change the World Current Applications Today s use of AI is narrow it is used to perform one specific task, such as speech and facial recognition AI is also used for mass data analysis on consumer behaviour, the most famous being Facebook and Netflix s algorithm that recommends content to its users Areas of Disruption Projects Under Development Autonomous Logistics (Automobiles, Trucking, Pilot-less Aircraft) Medical Treatment Recommendations & Treatment Design Plans Autonomous Investment Recommendations & Robotic Process Automation Natural Language Processing Security Monitoring & Analysis Journalism Sources: BBC, The Economist, United Nations Sources: Bloomberg, MIT, Softbank, The Economist 4

5 Economic Impact Short-Term Pain Leads to Long-Term Gain The Risks Economists are becoming increasingly concerned about job polarization The stagnation of growth in median wages has been viewed as being potentially connected to increasing application of AI Many economic pundits have seen AI as a means of reducing entry-level jobs Transition in Type of Work Within the US The Reward AI has pioneered dramatic increases in high paying computing jobs Subtopic Replaces tasks and positions that humans are inefficient or not technically apt at completing Significant evidence that AI will improve the quality of life and enhance life expectancy Dramatic decreases in information asymmetry Correlation of Wages & Potential Automation Xyz - Xyz Sources: BBC, The Economist, United Nations Sources: The Economist, MIT, McKinsey & Co. 5

6 Economic Impact Tech Advancement Does Not Mean Unemployment US Unemployment Rate (January, 1948 to August, 2017) Xyz - Xyz GM Assembly Robotics Intro IBM PC Launch Dotcom Boom Historical Context Growth in High-Tech Employment Historically, automation has created more jobs than it has destroyed Automating a task increases the demand for complimentary ones It is easier to identify the jobs that will be automated, however more difficult to determine where new jobs will be created US Average Alabama North Carolina Ohio Texas Washington Kansas Michigan South Carolina Delware Sources: The Economist, MIT, U.S. Federal Reserve FRED System 6

7 Economic Impact Labour Market Disruption The Occupation Impact Low-Skill Labour Market Disruption Telemarketers Accountants & Auditors Retail Salespeople Technical Writers Real Estate Agents Typists Machinists Commercial Pilots Economists Health Technologists Firefighters Chemical Engineers Athletic Trainers Doctors Recreational Therapist When focusing on whole occupations rather than specific single-job tasks, it was found that 47% of jobs in the United States were at a high risk of being replaced by AI empowered automation However, other studies have shown that only 9% of jobs in the US are fully automatable There is consensus between both estimation methods that low-paid and low-skill work is threatened the most Probability of Automation to Median Wage Jobs with Highly Automatable Skills (%)x Less than High School Trade Associates Bachelors Graduate Less than 20 Dollars 20 to 40 Dollars More than 40 Dollars High School Degree or School Degree Degree Degree Equivalent Certificate Source: Frey & Osborne, Arntz, Gregory & Zierahn, Executive Office of the President 7

8 Economic Impact Inequality of Productivity Distribution Top 1% Income Share in the U.S. 30% 25% 20% 15% 10% 5% 0% Wealth Concentration and Inequality Creative Destruction AI will cause a wide variety of jobs, primarily low-skill/paid, to be destroyed, and Subtopic simultaneously create new jobs to support the technological change The new jobs will almost exclusively be high-skill jobs, and the number of new jobs created will be far less than those destroyed Labour with Automatable Activities The productivity gains from AI will directly benefit those who own the means of automation, i.e. corporate shareholders Meanwhile, workers that are replaced suffer depressed wages, leading to further income inequality If the disruption occurs rapidly enough, consumption will fall and economic stagnation or decline could follow China Europe Big 5 Japan 235 India 395 United States Remaining Countries Sources: BBC, The Economist, United Nations Sources: Piketty and Saez, Martin Ford, Bruegel, McKinsey Global Institute, Executive Office of the President 8

9 Security Considerations The Zuckerberg Perspective The Vision A utopian future with Genetics, Nanotechnology and Robotics (GNR) revolutionizing everything - Changing our genes to avoid disease; aiding in drug discovery; robots making tasks easier and less repetitive The Case for AI The truth is no online database will replace your daily newspaper, no CD-ROM can take the place of a competent teacher and no computer network will change the way government works Clifford Stoll The dangers are well understood - Time to debate the issues and take wise actions to deal with them effectively Increased unemployment and greater wealth inequality are debatable given that the Industrial and Digital Revolutions decreased rather than increased unemployment Possible catastrophes, - those that threaten to destroy the environment or end human civilization - are meant to simply attract attention Dealing with the Dangers Sources: MIT, The Economist, University of Nicosia Study 9

10 Security Considerations Safety Concerns of Artificial Intelligence Dangers of AI In a recent Oxford survey of 350 AI experts, the probability of an extremely bad outcome, such as human extinction as a result of AI, was estimated at 5% Musk: reason we needed to colonize Mars so that we ll have a bolt-hole if AI goes rogue and turns on humanity Hawkings: "The development of full AI could spell the end of the human race" Importance of Proactive Regulation Digital currencies, drones, autonomous vehicles, peer-to-peer lending and renewables are just some areas of tech that have largely ignored regulatory guidelines Musk envisions a government agency forming first and seeking to gain insight into AI and its use before regulating Artificial Intelligence Development Timeline 1997: IBM's Deep Blue beats the chess grandmaster 2017: AlphaGo beats the world s best human player 2026: Writing highschool essays 2031: Working in retail 2049: Writing a bestseller 2053: Working as surgeons 2070: Judgement Day 2011: Beats champions in Jeopardy Sources: BBC, The Economist, United Nations 2024: Better than humans at translating languages 2027: Driving trucks 2057: Beat humans at every task Source: Oxford & Yale Study, Gartner, TechCrunch 10

11 Government Involvement Bounding the Unknown Reasons for Government Involvement AI will inevitably change citizens lives, affecting their privacy, labour and security Key issues that will require government attention: - Who is at fault in the event AI causes societal harm - Job automation and the further concentration of wealth - How to regulate AI technology while promoting domestic innovation Job Automation (% at Potential High Risk) 56% 46% 44% 37% 32% Transportation and Storage Manufacturing Wholesale and Retail Trade Administrative and Support Services Financial and Insurance Issues with Regulating Artificial Intelligence Regulators lack an understanding of AI - A proposed solution is creating a Federal Robotics Commission comprised of industry experts who would recommend policy to federal agencies (e.g. SEC, FDA) AI cannot be specifically defined and the impact on both firms and consumers is largely unknown The robots being developed are seemingly independent, thus their legal status is ambiguous Sources: Brookings Institute, Harvard Business Review, PwC 11

12 Government Involvement Regulation v. Innovation Proactive Regulation Governments would regulate AI s development and implementation in advance of its use by firms and consumers The intention is to prevent societal harm before it occurs Possibility of discouraging innovation, putting the domestic economy at a great disadvantage Reactive Regulation Regulations would be created after incidence of societal harm Subtopic This is the most common regulatory process There are concerns AI could cause such extreme societal harm that subsequent regulation would be ineffective in aiding vulnerable and previously affected citizens World Digital Competiveness Top Five Ranking Countries 3 USA 2 Sweden 5 Denmark 4 Finland 1 Singapore Sources: BBC, The Economist, United Nations Sources: Brookings Institute, IMD World Competitiveness Center 12