Note: Intro -- Preface -- Organization -- Contents -- Self-Modification of Policy and Utility Function in Rational Agents -- 1 Introduction -- 2 Preliminaries -- 3 Self Modification Models -- 4 Agents -- 5 Results -- 6 Conclusions -- References -- Avoiding Wireheading with Value Reinforcement Learning -- 1 Introduction -- 2 Setup -- 3 Agent Belief Distributions -- 3.1 Consistency of B and C -- 3.2 Non-Assumptions -- 4 Agent Definitions -- 5 Avoiding Wireheading -- 6 Discussion and Conclusions -- References -- Death and Suicide in Universal Artificial Intelligence -- 1 Introduction -- 2 Preliminaries -- 3 Definitions of Death -- 4 Known Environments: AI -- 5 Unknown Environments: AIXI and AI -- 6 Conclusion -- References -- Ultimate Intelligence Part II: Physical Complexity and Limits of Inductive Inference Systems -- 1 Introduction -- 2 Notation and Background -- 3 Physical Limits to Universal Induction -- 3.1 Logical Depth and Conceptual Jump Size -- 3.2 A Graphical Analysis of Intelligent Computation -- 3.3 Physical Limits, Incremental Learning, and Digital Physics -- References -- Open-Ended Intelligence -- 1 Introduction -- 2 What Is Intelligence? Definition and Critique -- 3 The Theory of Individuation -- 3.1 Assemblages -- 3.2 A New Conceptual Approach to Intelligence -- 4 Intelligence, Cognition, Sense-Making -- 5 A Framework for Open-Ended Intelligence -- 5.1 Structure -- 5.2 The Unfoldment of Individuation -- 5.3 Compatibility, Complexity and OEI -- 5.4 Coordination -- 6 Conclusion -- References -- The AGI Containment Problem -- 1 Introduction -- 2 Motivation -- 2.1 Testing and Experimentation in Safe AGI Development -- 2.2 Emergent Goals of Test AGIs -- 3 Requirements for an AGI Container -- 3.1 Human Factors and Information Hazards -- 4 Defense in Depth -- 5 Light, Medium and Heavy Containment -- 6 Existing Mechanisms. , 7 Topics for Future Work -- 8 Conclusion -- References -- Imitation Learning as Cause-Effect Reasoning -- 1 Introduction -- 2 Demonstrating Hard-Drive Maintenance -- 3 Imitation Learning with Causal Inference -- 3.1 Learning Skills by Explaining Demonstrations -- 3.2 Imitation and Generalization -- 4 Theoretical and Empirical Results -- 5 Conclusion -- References -- Some Theorems on Incremental Compression -- 1 Introduction -- 2 Preliminaries -- 3 An Example -- 4 Definitions -- 5 Properties of a Single Compression Step -- 6 Orthogonal Feature Bases -- 7 Efficiency of Incremental Compression -- 8 Discussion -- A Proofs -- References -- Rethinking Sigma's Graphical Architecture: An Extension to Neural Networks -- Abstract -- 1 Introduction -- 2 How Did This Come About? -- 3 To What Extent Does It Occur? -- 3.1 Directed Links -- 3.2 Closed-World Semantics -- 3.3 Universal Variables -- 3.4 Filter Nodes -- 3.5 Transform Nodes -- 4 What Are Its Implications (Including to Neural Networks)? -- 5 Conclusion -- Acknowledgments -- References -- Real-Time GA-Based Probabilistic Programming in Application to Robot Control -- Abstract -- 1 Introduction -- 2 Lightweight Implementation of GA-Based Optimization Queries in Probabilistic Programming -- 3 Planning as Probabilistic Programming -- 4 Simultaneous Plan Optimization and Execution -- 5 Conclusion -- Acknowledgements -- References -- About Understanding -- 1 Introduction -- 2 Related Work -- 3 Towards a Theory of Pragmatic Understanding -- 4 Meaning -- 5 A System that Acquires Understanding and Meaning -- 6 Conclusions -- References -- Why Artificial Intelligence Needs a Task Theory -- 1 Introduction -- 2 What We Might Want from a Task Theory -- 3 Requirements for a Task Theory -- 4 What a Task Theory Might Look Like -- 5 Conclusions -- References -- Growing Recursive Self-Improvers -- 1 Introduction. , 2 Scope and Delineation -- 3 Essential Ingredients of expai -- 4 Recursive Self-Improvement -- 5 Towards a Test Theory -- References -- Different Conceptions of Learning: Function Approximation vs. Self-Organization -- 1 Learning: Different Conceptions -- 2 Learning in NARS -- 3 Comparison and Discussion -- 4 Conclusions -- References -- The Emotional Mechanisms in NARS -- 1 Intelligence and Emotion -- 2 Desirability of Events -- 3 Feelings of the System -- 4 Emotion in Concepts -- 5 Effects of Emotion -- 6 Comparison to Other Approaches -- 7 Comparison to Human Emotions -- 8 Conclusions -- References -- The OpenNARS Implementation of the Non-Axiomatic Reasoning System -- 1 Introduction -- 2 Memory -- 3 Logic Module -- 4 Temporal Inference Control -- 5 Projection and Eternalization -- 6 Anticipation -- 7 Evidence Tracking -- 8 Processing of New and Derived Tasks -- 9 Attentional Control -- 10 Conclusions -- References -- Integrating Symbolic and Sub-symbolic Reasoning -- 1 Introduction -- 2 System Components -- 2.1 Status Signals -- 2.2 Long-Term Memory -- 2.3 Activity -- 2.4 Attention -- 2.5 Working Memory -- 2.6 Decision -- 3 Update Functions -- 3.1 Activity Update -- 3.2 Status Update -- 3.3 Attention Update -- 3.4 WM Update -- 3.5 Decision Update -- 3.6 LTM Update -- 4 Reasoning Mechanisms -- 4.1 Sub-symbolic Reasoning -- 4.2 Symbolic Reasoning -- 5 Prototype Implementation -- 6 Conclusion -- References -- Integrating Axiomatic and Analogical Reasoning -- 1 Introduction -- 2 Mathematical Model -- 2.1 Basic Concepts -- 2.2 Domains -- 2.3 Axiomatic Reasoning -- 2.4 Analogical Reasoning -- 3 System Description -- 4 System Evaluation -- 4.1 Rutherford's Analogy -- 4.2 Natural Language Analogy -- 5 Conclusions -- References -- Embracing Inference as Action: A Step Towards Human-Level Reasoning -- 1 CEC and CECAC -- 1.1 Boxes -- 1.2 Evaluated Codelets. , 2 Future Work -- References -- Asymptotic Logical Uncertainty and the Benford Test -- 1 Introduction -- 2 Related Work -- 3 The Benford Test -- 4 Irreducible Patterns -- 5 A Learning Algorithm -- 6 Passing the Generalized Benford Test -- 7 Final Remarks -- References -- Towards a Computational Framework for Function-Driven Concept Invention -- 1 Introduction -- 2 Concept Representation -- 3 Computing Blends -- 3.1 Concept Combination -- 3.2 Selecting Concepts -- 3.3 Computational Results -- 4 Related Work -- 5 Conclusions and Future Work -- References -- System Induction Games and Cognitive Modeling as an AGI Methodology -- 1 Introduction -- 2 Related Work -- 3 General Observations on Human SIG-Playing Behavior -- 4 A Model of Early Decision-Making on SIGs -- 4.1 Data -- 4.2 Model -- 4.3 Discussion -- 5 Conclusion -- References -- Integrating Model-Based Prediction and Facial Expressions in the Perception of Emotion -- 1 Introduction -- 2 Method -- 2.1 Expressing Individual Difference in Bayesian Inference -- 2.2 Appraisal Theory and Theory of Mind -- 2.3 Display Rules -- 2.4 Calculation -- 3 Simulation -- 3.1 Context and Model -- 3.2 Display Rules -- 3.3 Experiment -- 4 Simulation Results -- 5 Discussion and Future Work -- References -- A Few Notes on Multiple Theories and Conceptual Jump Size -- 1 Understanding and Learning -- 2 Algorithmic Probability and the Suite of Theories -- 3 Using Bayes' Rule -- 4 Incomputability -- 5 Metamorphoses of a Theory -- 6 Lsearch -- 7 Conceptual Jump Size and Descriptions -- 8 Can the Search Be Practical? -- 9 Agents -- 10 Fun with Unconscious Jumps -- 11 On the Back Porch Just Beyond the Universe -- References -- Generalized Temporal Induction with Temporal Concepts in a Non-axiomatic Reasoning System -- Abstract -- 1 Introduction -- 2 Temporal Concurrency -- 3 Implementation -- 4 Discussion -- 5 Conclusion. , References -- Introspective Agents: Confidence Measures for General Value Functions -- References -- Automatic Sampler Discovery via Probabilistic Programming and Approximate Bayesian Computation -- 1 Introduction -- 2 Related Work -- 3 Approach -- 3.1 Grammar and Production Rules -- 3.2 Probabilities for Production Rules -- 4 Experiments -- 4.1 Samples from Sampled Probabilistic Programs -- 4.2 Learning Sampler Code for Common One-Dimensional Distributions -- 4.3 Evaluating Our Approach Versus Evolutionary Algorithms -- 4.4 Generalising Arbitrary Data Distributions -- 5 Discussion -- References -- How Much Computation and Distributedness is Needed in Sequence Learning Tasks? -- 1 Introduction -- 2 Cellular Automata in Reservoir Computing: ReCA -- 2.1 Encoding Stage -- 2.2 Cellular Automata Reservoir Stage -- 2.3 Read-Out Stage -- 3 Covariance and Stack Representations -- 4 Experiments -- 5 Results and Discussion -- 6 Conclusion -- References -- Analysis of Algorithms and Partial Algorithms -- 1 Introduction: Shortcomings of Traditional Analysis of Algorithms -- 2 Expected-Reward Analysis of Algorithms -- 2.1 Definition -- 2.2 Theory and Practice -- 3 Self-improving AI -- 4 Future Work -- References -- Estimating Cartesian Compression via Deep Learning -- 1 Introduction -- 2 Related Works -- 3 Methods -- 3.1 Theoretical Background -- 3.2 Problem Formulation -- 3.3 Numerical Experiment -- 4 Results -- 5 Discussion -- 6 Conclusions -- References -- A Methodology for the Assessment of AI Consciousness -- Abstract -- 1 Methodology -- 2 Instructions -- 3 Ability to Reason and Use Logic -- 4 Situational Awareness -- 5 Natural Language Ability -- 6 Goals, Opinions, and Emotions -- 7 Experiencing Existence -- 8 Growth and Learning -- 9 Self Knowledge -- 10 Self Control -- 11 Knowledge About Humans -- 12 Knowledge About the Current Conversationalist. , 13 Curiosity and Imitation.