Reflections on Artificial Intelligence for Humanity (Lecture Notes in Artificial Intelligence)

We already observe the positive effects of AI in almost every field, and foresee its potential to help address our sustainable development goals and the urgent challenges for the preservation of the environment. We also perceive that the risks related to the safety, security, confidentiality, and fairness of AI systems, the threats to free will of possibly manipulative systems, as well as the impact of AI on the economy, employment, human rights, equality, diversity, inclusion, and social cohesion need to be better assessed. The development and use of AI must be guided by principles of social cohesion, environmental sustainability, resource sharing, and inclusion. It has to integrate human rights, and social, cultural, and ethical values of democracy. It requires continued education and training as well as continual assessment of its effects through social deliberation. The “Reflections on AI for Humanity” proposed in this book develop the following issues and sketch approaches for addressing them: How can we ensure the security requirements of critical applications and the safety and confidentiality of data communication and processing? What techniques and regulations for the validation, certification, and audit of AI tools are needed to develop confidence in AI? How can we identify and overcome biases in algorithms? How do we design systems that respect essential human values, ensuring moral equality and inclusion? What kinds of governance mechanisms are needed for personal data, metadata, and aggregated data at various levels? What are the effects of AI and automation on the transformation and social division of labor? What are the impacts on economic structures? What proactive and accommodation measures will be required? How will people benefit from decision support systems and personal digital assistants without the risk of manipulation? How do we design transparent and intelligible procedures and ensure that their functions reflect our values and criteria? How can we anticipate failure and restore human control over an AI system when it operates outside its intended scope? How can we devote a substantial part of our research and development resources to the major challenges of our time such as climate, environment, health, and education? Preface 6 Organization 9 Contents 11 Reflections on AI for Humanity: Introduction 13 1 Context of the Book 13 2 What Is AI Today 14 3 AI Risks and Challenges 15 4 Worldwide Initiatives on the Societal Impact of AI 16 5 Outline of the Book 19 6 What's Next: An Opening for GPAI 22 References 23 Trustworthy AI 25 1 The Necessity of Trustworthy AI 25 2 The Meaning of Trust Regarding Machines 27 2.1 Technical Trust 27 2.2 Governance 29 3 The Difficulty of Understanding 29 3.1 Complexity of AI Systems 30 3.2 Human Understanding 32 4 Explainability – Opening the Black Box 33 4.1 Approaches 35 4.2 Open Challenges 36 5 Verification 37 5.1 Issues 37 5.2 Approaches 38 5.3 Challenges 39 6 Human Rights and AI 39 6.1 Why Should Human Rights Provide the Foundational Ethical Standards for Trustworthy AI? 40 6.2 How to Ensure that Trustworthy AI Offers Effective Human Rights Protection? 41 6.3 Open Challenges 43 7 Beneficial AI 43 7.1 AI in the Standard Model 44 7.2 AI in the New Model: Assistance Games 45 7.3 Research Agenda and Open Questions 46 8 Conclusion: The Way Ahead 47 References 48 Democratising the Digital Revolution: The Role of Data Governance 52 Abstract 52 1 Introduction 52 2 Data for Intelligence: The Role of Data Governance in Creating AI that Benefits Humanity 53 3 The Role of Law and Governance in the Digital Environment 54 3.1 Understanding the Lessons from Recent History 54 3.2 Current Legal Structures and Data Rights 55 3.3 The Changing Technology Environment 57 3.4 Bridging the Gaps: A Democratic Model for Data Governance? 58 4 Commons, Cooperatives, and Counter-Power 58 4.1 Mutualisation as a Tool to Counter Power Asymmetries 58 4.2 The Emergence of Data Trusts as a Governance Tool 59 5 Optimising for Democracy? A Data Governance System that Benefits Humanity 60 References 62 Artificial Intelligence and the Future of Work 65 1 Introduction 65 2 Is This Time Different? 67 2.1 Impact on the Manufacturing Sector 68 2.2 Impact on the Service Sector 68 2.3 Information as Key Driver 69 2.4 Expected or Explosive? 69 2.5 “Telemigration” in Perspective 70 3 Artificial Intelligence and Job Quality 71 3.1 Artificial-Artificial Intelligence 72 3.2 Algorithmic Management and Job Quality 73 3.3 Platform and Beyond 74 4 Policy Implication 74 4.1 Testing New Approaches 75 4.2 International Organizations’ Perspectives 75 5 Conclusion 77 Acknowledgements 78 References 78 Reflections on Decision-Making and Artificial Intelligence 80 1 Introduction 80 2 Machine Learning Predictions for AI Decisions 80 3 Case Studies 81 3.1 AI Shared Decision-Making in Health Care 81 3.2 AI Decision-Making as a Social System 82 3.3 Fair and Accountable AI Decisions 83 3.4 Designing for Responsible Decision-Making 84 3.5 Educating for Human Agency 85 4 Conclusion 86 Acknowledgements 86 References 86 AI & Human Values 88 Abstract 88 1 Introduction 88 2 Concepts 90 2.1 Privacy 90 2.2 Bias 91 2.3 Fairness and Discrimination 92 2.4 Nudge 92 3 Inequalities 93 3.1 Bias 93 3.2 Fairness and Discrimination 94 3.3 Nudge 95 3.4 Feedback Loops 96 4 Putting Human Values at the Core of AI 97 References 99 Next Big Challenges in Core AI Technology 102 Abstract 102 1 The Need to Address Scientific and Technological Challenges for an AI for Humanity 102 2 Endowing Deep Neural Networks to Show and Explain Behavior and Decision Making 106 2.1 AI Landscape and Architecture Search 107 2.2 Interpreting Deep Neural Networks 109 2.3 Explainable AI 112 3 The Challenge of Trustworthy AI 114 4 Addressing the AI Talent Bottleneck by Automating Artificial Intelligence 116 4.1 Causes and Consequences of the Current Boom in AI 117 4.2 The Biggest Risk Associated with AI 119 4.3 Automating Artificial Intelligence 120 4.4 The Way of the Future 122 References 124 AI for Humanity: The Global Challenges 128 1 What AI Can Do for Us 128 2 Global Challenge: Health 129 3 Global Challenge: Education 131 4 Global Challenge: Earth 133 5 Global Challenge: Science 134 6 Conclusion: Technical and Philosophical Challenges for AI 135 References 137 AI and Constitutionalism: The Challenges Ahead 139 Abstract 139 1 Introduction 139 2 Big Data, Privacy and the Limits of Informed Consent 140 2.1 ICT and AI: A Consciously Misinformed Consent 141 2.2 The Overestimation of Informed Consent in AI-Driven Medical Research 141 3 The Risk of Monopoly Power 143 4 Political Profiling and the Bubble Democracy 143 5 AI and Equality 144 5.1 Workforce and Job Market 145 5.2 Justice 146 5.3 Health and Medicine 148 6 The Central Issue: Evaluation and Decision 150 6.1 The Technical Arguments 151 6.2 The Anthropological Arguments 152 7 A List of New ‘human’ Rights 153 8 Concluding Remarks 156 References 156 Analyzing the Contribution of Ethical Charters to Building the Future of Artificial Intelligence Governance 162 1 Introduction 162 2 The Ethical Charter Landscape: The First Component of AI Governance Development 163 2.1 The Good, the Bad and the Ugly 163 2.2 The Formalization of Ethics: A Social Regulation Tool 166 3 Building on Ethical Charters to Expand the AI Normative Landscape 169 3.1 The Deployment of AI Regulations 170 3.2 Seizing Opportunities for an International Scaling-Up of AI Normativity 174 4 Conclusion 177 References 178 What Does “Ethical by Design” Mean? 183 1 Introduction 183 2 The “by Design” Family 184 2.1 Safety by Design 185 2.2 “Privacy by Design”: A Legal Approach to the Protection of Personal Information 186 2.3 The Comprehensive Framework of “Responsible Innovation” 188 3 State of the Art: Top-Down and Bottom-up Approaches 190 3.1 A Deontological Approach: A Hippocratic Oath for AI? 191 3.2 Applied Ethics for AI 192 3.3 Moral Machines? 193 3.4 Fairness by Design 194 4 Guidelines for an Ethics by Design 195 4.1 Ethics in the Process 195 4.2 Experiential and Experimental 196 4.3 A Processual Care Ethics with AI 197 References 200 AI for Digital Humanities and Computational Social Sciences 203 Abstract 203 1 AI as an Object of Research for Social and Human Sciences 203 1.1 AI and the History of Science 203 1.2 AI and Its Imagination 204 2 AI Methods and Tools for Social and Human Sciences 205 2.1 Text, Language and Data Analysis 205 2.2 Network Analysis 208 2.3 IA in Art History 209 3 AI in a Social Research Practice and Organizaztion 210 References 212 Augmented Human and Human-Machine Co-evolution: Efficiency and Ethics 215 Abstract 215 1 Introduction 216 2 Short Definition of Terms 217 2.1 Augmented Human (Physical/Cognitive/Virtual) 218 2.2 Human-Machine Co-evolution 218 2.3 Core Principles for Ethical AI 220 3 Facets of Human Machine Co-creation, Co-learning and Co-adaption 220 3.1 Surviving in Man-Made Environments: The Case for Language and Vision 221 3.2 Robots Learning from Humans: Past, Current and Future to Purposive Learning 223 3.3 Empowering Multimodal Affective Behavior Analysis by Interactive Machine Learning 226 3.4 Symbiotic Interaction to Socialware – Social and Semantic Interactions of Augmented Human and Ambient Intelligence 227 3.5 Socially Aware AI - Maintaining the Human at the Center of AI Design 228 4 Best Practices in Education 230 4.1 IntelliChalk – Teaching Mathematics with a Data Wall 230 4.2 Lumilo – AI for Personalized Learning: Students, Teachers and AI Systems Augmenting Each Other’s Abilities 231 4.3 Wordometer, CoaLA and LeAE – Experiential Supplements: Sharing Human Experiences for Co-learning 233 5 Conclusion 234 References 236 Democratizing AI for Humanity: A Common Goal 240 Abstract 240 1 Background 240 1.1 The Need to Democratize AI 242 1.2 Problem Space 243 2 The AI Commons 244 2.1 History of the AI Commons 244 2.2 The AI Commons Model 245 A Framework for Global Cooperation on Artificial Intelligence and Its Governance 249 Abstract 249 1 Introduction 249 2 The Need for Global Cooperation on AI 251 2.1 Why AI Necessitates Governance 251 2.2 Why AI’s Governance Requires a Global Approach 252 2.3 Challenges to Overcome 254 3 Areas for Global Cooperation on AI and Its Governance 256 3.1 The Horizontal Dimension of AI Governance 257 3.2 The Vertical Dimension to AI Governance 263 4 Organizing Global Cooperation on AI and Its Governance 266 4.1 Balancing the Need for Swift Action, a Holistic Approach and Attention to Context-Specificity 267 4.2 Clarifying the Rules of Engagement 267 4.3 Building on Existing Cooperation Structures 268 4.4 Developing a Network of Networks 268 4.5 Maintaining Openness to Differentiated Cooperation 270 4.6 Securing an Inclusive and Transparent Way of Working, Mindful of Power Imbalances 271 4.7 Establishing a Feedback Loop and Preparing for the Future 272 5 Conclusions 273 Acknowledgement 274 References 274 We already observe the positive effects of AI in almost every field, and foresee its potential to help addressing our sustainable development goals and the urgent challenges for the preservation of the environment. We also perceive the risks related to the safety, security, confidentiality, and fairness of AI systems, the threats to free will of possibly manipulative systems, as well as the impacts of AI on the economy, employment, human rights, equality, diversity, inclusion, and social cohesion need to be better assessed. The development and use of AI must be guided by principles of social cohesion, environmental sustainability, resource sharing, and inclusion. It has to integrate human rights, and social, cultural, and ethical values of democracy. It requires continued education and training as well as continual assessment of effects through social deliberation. The "Reflections on AI for Humanity" proposed in this book develop the following issues and sketch approaches for addressing them: How can we ensure the security requirements of critical applications and the safety and confidentiality of data communication and processing? What techniques and regulations for the validation, certification, and audit of AI tools are needed to develop confidence in AI? How can we identify and overcome biases in algorithms? How do we design systems that respect essential human values, ensuring moral equality and inclusion? What kinds of governance mechanisms are needed for personal data, metadata, and aggregated data at various levels? What are the effects of AI and automation on the transformation and social division of labor? What are the impacts on economic structures? What proactive and accommodation measures will be required? How will people benefit from the decision support systems and personal digital assistants without the risk of manipulation? How do we design transparent and intelligible procedures and ensure that their functions reflect our values and criteria? How can we anticipate failure and restore human control over an AI system when it operates outside its intended scope? How can we devote a substantial part of our research and development resources to the major challenges of our time such as climate, environment, health, and education?