Center for Analysis and Design of Intelligent Agents

Make sure to read the required number of papers (first number inside the brackets - the second one is the “minimum recommended”) and make sure to not fall behind on readings (I assign you only a few papers per day for a good reason - so you can get through them in time for the discussion session on that day). Note: We will interweave content from prior sessions in the following ones, so if you fall behind two or more days in a row, you will be significantly challenged to keep up (there are subtleties in the content that is really key to understanding the content and passing the course - you may feel like you're following along the discussion, but there will likely be important things you're missing).

Guidelines for how to read in this course (seriously! - do not skip).

What is intelligence?
How do experts talk about it?
What has been uncovered?
What uniquely separates the phenomenon of intelligence from other similar phenomena in the world?

• The Future of AI Research: Ten Defeasible 'Axioms of Intelligence' by K.R.Thórisson and H. Minsky
• Discretionarily Constrained Adaptation Under Insufficient Knowledge & Resources by K.R.Thórisson
• Cognitive Architecture Requirements for Achieving AGI by J.E. Laird et al.
• On Defining Artificial Intelligence by P. Wang
• What is GMI? by K. R. Thórisson.
• Cognitive Architecture Requirements for Achieving AGI by J.E. Laird et al.
• Introduction: Aspects of Artificial General Intelligence by P. Wang & B. Goerzel (first 3 sections)
• Mapping the Landscape of Human-Level Artificial General Intelligence by Adams et al.
• Four Basic Questions about Artificial Intelligence by P. Wang.
• Three Fundamental Misconceptions of Artificial Intelligence by P. Wang.
• Artificial General Intelligence on Wikipedia.
• The very laws of physics imply that artificial intelligence must be possible. What's holding us up? by D. Deutch.
• Can artificial intelligence become sentient, or smarter than we are - and then what? Techtopia @ Deutsche Welle

What is reasoning for?
What kinds of processes does reasoning consist of?
How can the sub-processes of reasoning be coordinated at runtime?

• The Logic of Learning by P. Wang
• Behavioral Self-Programming by Reasoning by P. Wang
• Cumulative Learning With Causal Relational Models by K.R.Thórisson & A.Talbot
• Critical Reasoning by M. Talbot
• Bayesianism and Causality, or, Why I Am Only A Half-Bayesian by J. Pearl
• Predictive Heuristics for Decision-Making in Real-World Environments by H.P.Helgason et al.

What is learning?
Do different kinds of learning exist?
What are the component processes of learning?
How can these processes be unified in a single coherent system?
Is “machine learning” comparable to human (kinds of) learning?

• Cumulative Learning by K.R. Thórisson et al.
• What is GMI? by K. R. Thórisson.
• Seed-Programmed Autonomous General Learning (sections 1 and 2) by K.R. Thórisson
• A Brief Intro to Reinforcement Learning by Kevin Murphy.
  • Alternative to Murphy: Reinforcement Learning in the Encyclopedia of Cognitive Science by P. Dayan and C. Watkins.
• Comparison of Machine Learners on an ABA Experiment Format of the Cart-Pole Task by Eberding et al.
• Reinforcement Learning: An Introduction by Rich Sutton and Andrew Barto (1998), First chapter. This is “the” introductory text on RL.
• The Emergence of Artificial Intelligence: Learning to Learn by P. Bock.
• The Logic of Learning by P. Wang.

How is autonomy defined?
What are the levels of autonomy?
What are the minimum requirements for different autonomy levels?
How can autonomy be achieved in an artificial system?
Is learning necessary for autonomy?

• Creating a Child Machine: Reflections on Turing’s Proposal by R. Reddy
• Cognitive Architectures & Autonomy: A Comparative Review by K.R. Thórisson & H.P. Helgason.
• Fridges, Elephants, and the Meaning of Autonomy and Intelligence by R. Sanz et al.
• Principles of Integrated Cognitive Control by R. Sanz et al.
• Towards a Programming Paradigm for Control Systems With High Levels of Existential Autonomy by E. Nivel & K.R.Thórisson.
• A Primer For Conant & Ashby's “Good-Regulator Theorem” by Daniel L. Scholten.
• Every Good Regulator of a System Must be a Model of that System by Conant & Ashby.
• Hybrid Systems on Wikipedia.
• A Framework for Autonomy Levels for Unmanned Systems (ALFUS) by Huang et al.
• Self-awareness in Real-Time Cognitive Control Architectures by R. Sanz, I. López & C. Hernández
• Cybernetics and Second-Order Cybernetics by Heylighen & Joslyn.
• Robert Rosen’s Anticipatory Systems Theory: The Art and Science of Thinking Ahead by J. Rosen.
• Time, Clocks, and the Ordering of Events in a Distributed System by L. Lamport
• Fridges, Elephants, and the Meaning of Autonomy and Intelligence by R. Sanz et al.
• The Emergence of Artificial Intelligence: Learning to Learn by P. Bock

Are symbols and words the same thing?
Is the relation between words and symbols bijective?
Can anything be a model of anything?
How are symbols related to models?
How are models of causal relations made?

• Minds, Brains & Programs by John Searle
• Subsymbolic Computation and the Chinese Room by D. Chalmers
• The 'Cybernetic Cut': Progressing from Description to Prescription in Systems Theory by D. L. Abel
• Syntactic Autonomy - Or Why There is no Autonomy Without Symbols and how Self-Organizing Systems Might Evolve Them by L. M. Rocha

Where does meaning come from?
Who and what is meaning for?
Are there different kinds of meaning?
How does reasoning fit into the concept of meaning?
Is meaning necessary for understanding?

• About Understanding by K. R. Thorisson et al.
• The ‘Explanation Hypothesis’ in General Self-Supervised Learning by K.R. Thórisson
• The Physical Symbol System Hypothesis: Status & Prospects by N. J. Nilsson.
• Understanding & Common Sense: Two Sides of the Same Coin? by K. R. Thórisson & D. Kremelberg
• Understanding is a Process by Blaha et al.
• Cell psychology: An evolutionary approach to the symbol-matter problem by H. Pattee
• Evaluating Understanding by J.Bieger & K.R.Thórisson
• Do Machines Understand? A Short Review of Understanding & Common Sense in Artificial Intelligence by K.R. Thórisson & D. Kremelberg

What role does a cognitive architecture play in intelligence?
How is cognitive architecture different from software architecture?
How does reasoning, goals, understanding and meaning come into a cognitive architecture?
How does reasoning fit into the concept of meaning?
Is meaning necessary for understanding?

• Architectural Requirements for Human-like Agents Both Natural and Artificial by A. Sloman
• A New Constructivist AI: From Manual Construction to Self-Constructive Systems by K.R. Thórisson
• Architectural Mismatch or Why it’s hard to build systems out of existing parts by Garlan, D., R. Allen and J. Ockerbloom.
  • Also available here
• Introduction to Software Architecture by Garlan & Shaw.
• Growing Recursive Self-Improvers by B. Steunebrink et al.
• Bounded Seed-AGI by E. Nivel et al.
• Self-Programming: Operationalizing Autonomy by E. Nivel & K.R. Thórisson.
• Approaches & Assumptions of Self-Programming in Achieving Artificial General Intelligence by K.R. Thórisson et al.

• On Defining Artificial Intelligence, Journal of Artificial General Intelligence 10(2):1-37 by P. Wang.
• An integrated theory of mind by Anderson et al.
• (ACT-R) An Integrated Theory of the Mind by Anderson, J. R.; Bothell, D.; Byrne, M.D.; Douglass, S.; Lebiere, C. & Qin, Y.
• Does the Future of AGI Lie in Cognitive Synergy? by B. Goertzel
• Universal Intelligence: A Definition of Machine Intelligence by S. Legg and M. Hutter.
• Essentials of General Intelligence: The Path Towards AGI by P. Voss.
• Constructivist AI Methodology by K.R. Thórisson
• Can there be a science of complex systems? by H. A. Simon
• Cybernetics and Second-Order Cybernetics by Heylighen & Joslyn

The Subsumption Architecture is definitely GOFAI-style architecture: With baked-in hand-coded goals and control structures, these systems are notoriously difficult to build for autonomous adaptation of any kind. But they are fun to build, robust and easy to debug.

• Subsumption Architecture on Wikipedia.
  • (supplementary, optional) How to Build Complete Creatures Rather than Isolated Cognitive Simulators by Rodney Brooks.
  • (supplementary, optional) A Robust Layered Control System for a Mobile Robot by R. Brooks.
  • (supplementary, optional) Whatever happened to the subsumption architecture? by Simon Birrell.

This is the definitive reasoning architecture, under development since the 1990s.

• From NARS to a thinking machine by P. Wang
• What is NARS by K. R. Thórisson
• Introduction to NARS by P. Wang
• NARS intro paper by P. Wang
• NARS demos by P. Wang

The Reykjavik University architecture that has been shown to learn very complex tasks by observation.

• Achieving Artificial General Intelligence Through Peewee Granularity by Thórisson, K. R. & Nivel, E.
• Self-Programming: Operationalizing Autonomy by Nivel, E. & K. R. Thórisson
• Bounded Seed-AGI by E. Nivel et al.
• Autonomous Acquisition of Natural Communication by K. R. Thórisson et al.
• Anytime Bounded Rationality by E. Nivel et al.
• Bounded Recursive Self-Improvement by E. Nivel et al.

One of the earliest examples of implemented self-guided learning systems.

• Made-Up Minds: A Constructivist Approach to Artificial Intelligence by G. Drescher
  • Read chapters 1 & 2 for understanding Drescher's (and Piaget's) psychological motivation; Ch. 3 & 4 for an overview of the computational framework.

Sigma doesn't learn, but it's a great tool for learning about generality and autonomy.

• The Sigma Cognitive Architecture and System by P. S. Rosenbloom

Originally based on NARS (see above), the latest incarnation of the OpenCOG is called .

• OPENCOG PRIME: A COGNITIVE SYNERGY BASED ARCHITECTURE FOR ARTIFICIAL GENERAL INTELLIGENCE by Ben Goertzel
• Getting Started with Open Cog by Ben Goertzel

• Franklin, S. (2007). (LIDA) A Foundational Architecture for Artificial General Intelligence. Advances in Artificial General Intelligence: Concepts, Architectures and Algorithms. IOS Press, Amsterdam, The Netherlands, The Netherlands, pp. 36-54. PDF
• Anderson, J.R. & Schunn, C.D. (2000). Implications of the ACT-R learning theory: No magic bullets. Advances in instructional psychology. 5:1-34. Lawrence Erlbaum | PDF
• Laird, J.E.; Newell, A. & Rosenbloom, P.S. (1987). SOAR: An architecture for general intelligence. Artificial Intelligence, Volume 33, Issue 1, Pages 1-64 PDF
• Snaider, J; McCall, R. & Franklin, S. (2011). The LIDA framework as a general tool for AGI. Artificial General Intelligence, Lecture Notes in Computer Science. 2011. Volume 6830/2011. pp. 133-142 PDF
• J. Schmidhuber (2016). Learning how to Learn Learning Algorithms: Recursive Self-Improvement. - slides

• On Attention Mechanisms for AGI Architectures: A Design Proposal by Helgason et al.; accompanying video can be found here.
• A model of attention that takes global scene factors into account by Torralba.
• Towards a General Attention Mechanism for Embedded Intelligent Systems by H. P. Helgason et al.
• Attention Capabilities for AI Systems by H. P. Helgason & K. R. Thórisson.

• Selected Self-Organization and the Semiotics of Evolutionary Systems by L. M. Rocha
• Weak Emergence by M. A. Bedau
• Strong and Weak Emergence by D. Chalmers
• THE SCIENCE OF SELF-ORGANIZATION AND ADAPTIVITY by Heylighen

• What's it Like to be a Bat? by T. Nagel
• Consciousness and its Place in Nature by D. J. Chalmers
• The Future of Consciousness - TEDx lecture by R. Hameroff
• Consciousness is a Mathematical Pattern - TEDx lecture by M. Tegmark
• Consciousness and the Brain - TEDx lecture by J. Searle

• Leave no dark corner by Matthew Carney.
• THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION? by Frey and Osborne

• Reinforcement Learning: An Introduction by Rich Sutton and Andrew Barto (1998) is the introductory text on RL.
• Machine Super Intelligence is Shane Legg's 2008 PhD thesis. While it is not on reinforcement learning, it does connect the concepts of RL to AI, what he calls “universal AI”.
• Reinforcement Learning is a full Udacity course on RL from Georgia Tech.
• A Short Course on Reinforcement Learning by S. Singh at MLSS'11 introduces RL and discusses some important shortcomings and proposed first steps to solving them.
• Advanced Topics: RL by D. Silver is a more in-depth, modern RL course from one of the people who worked on Google DeepMind's Atari playing system that received a lot of (media) attention.
• Model-Based Reinforcement Learning is a tutorial given by Michael Littman at NeurIPS'09 about model-based RL, which is a lot less common than model-free RL, but not less interesting.
• Resource-Bounded Machines are Motivated to be Effective, Efficient & Curious by B. Steunebrink

• A gentle introduction to neural networks by J. Le - gives a good overview of the different approaches
• Form, function, and the giant gulf between drawing a picture and understanding the world by G. Marcus

• Kognit: Intelligent Cognitive Enhancement Technology by Cognitive Models and Mixed Reality for Dementia Patients by D. Sonntag and accompanying video
• Comparison table of cognitive architectures, courtesy of BICA Society / A. Samsonovich
• Why are so many smart people such idiots about philosophy? by O. Goldhill
• Cognitive Map Architecture: Facilitation of Human-Robot Interaction in Humanoid Robots by V. Ng-Thow-Hing et al. (2009).
• Toward a unified artificial intelligence by Wang, P.
• Judea Pearl lecture on Causation on YouTube
• Creativity in a Nutshell by M. Boden.
• The Role of Cognitive Systems Engineering in the Systems Engineering Design Process by Militello, L.G., Dominguez, C.O., Lintern, G. & Klein, G. (2010). In Systems Engineering, Vol 13(3), pp. 261-273.
• A survey on transfer learning by Pan, S. J. & Yang, Q. (2011). IEEE Transactions on Knowledge and Data Engineering, 22(10), pp. 1345–1359.
• Systems, models and self-awareness: Towards architectural models of consciousness by Sanz, R., Hernandez, C., Gomez, J., Bermejo-Alonso, J., Rodriguez, M., Hernando, A. & Sanchez, G. (2009). International Journal Of Machine Consciousness, 1(2), pp.255–279.
• Requirements for Machine Lifelong Learning by Silver, D.L. & Poirier, R. (2007). IWINAC, LNCS (4527), pp.313-319.
• Physicalism, Emergence and Downward Causation by Campbell and Bickhard
• Man as Machine - J. Bach's AGI intro lecture at AGI Summer School 2009
• Mastering the game of Go with deep neural networks and tree search by Silver, Huang et al. (DeepMind)
• Why Philosophers Should Care About Computational Complexity by S. Aaronsson
• A Vision for Computer Science - the System Perspective by J. Sifakis

• Journal of Artificial General Intelligence
• Introduction to Artificial Intelligence - A Modern Approach by S. Russell & P. Norvig
• Chapter 1: An Introduction to Philosophy of Science by Malcolm Forster

• Algorithms for Reinforcement Learning by Csaba Szepesvári (2010) is a much more recent, shorter book that discusses the strengths and weaknesses of various RL algorithms. See also: Rich Sutton's FAQ.
• Deep learning on Wikipedia (Sections: Intro, Overview, and Neural Networks).

Note: DO NOT SKIP READING THE BELOW TEXT

As you read papers from each of the following categories I want you ask yourself a few questions:

• For each paper in each category X, ask yourself:
  • What is X?
  • How does the human mind do X?
  • Do current computers do X?
  • …and …
  • Do we need (to replicate or capture) what the human mind does to achieve X to create a machine that rivals the ability of humans to do X?

If you can answer them satisfactorily when you're done reading you're good! Even if you can't you'll be fine if you: Write down the discrepancies and bring them to class in the form of questions. There is no such thing as a 'stupid question' when you're learning something new!

2025©K.R.Thórisson