Overall course focus: (hmm…)

1. Reinforcement learning.
2. Concept formation and program synthesis.
3. Adaptive and probabilistic logics.
4. Mental development theory.

Notes:

1. Universal Artificial Intelligence: universal induction, exhaustive program search and reinforcement learning algorithms (TeXmacs source)
   • A glimpse on Warren Smith’s idea of IQ test (source)
2. Techniques of Reinforcement Learning (TeXmacs source)
   1. Attach:RL_Ch6_Evolutionary_modular.pdf
   2. Attach:RL_Ch7_Hierarchical_RL.pdf
3. General Game Playing (TeXmacs source)
4. Knowledge Representation and Language (TeXmacs source)
   • Multi-layered Semantic Networks (replaces a subchapter above)
   • Ontological Semantics
   • Automated Language Acquisition
5. Adaptive Logics for Reasoning Systems
   1. Adaptive (or Defeasible) Logics and OSCAR (TODO: complete the notes about OSCAR)
   2. Frequency and/or Uncertainty Logics: Non Axiomatic Logic, Probabilistic Logic Networks (to come)
   • I’ve presented OSCAR and comments on John Pollock’s theory supported by his slides and article figures, I also introduced “propositional” NARS
6. Estimation of Distribution Algorithms and Genetic Programming (TeXmacs source)
7. Inductive (Logic) Programming
8. Spreading Activation: memory retrieval, distributed reasoning, action selection, probabilities (work in progress) approaches based on spreading activation mechanism or strong biological inspirations
9. The Representation and Acquisition of Concepts Δ (to come later)
   • based mostly on The Representation and Acquisition of Concepts. Psychology 747, Section 4097 - Fall 2003 [1], but also “Conceptual Spaces. Geometry of thought” and more
10. Probabilistic Modeling and Probabilistic Logics
    1. OLD: Markov Logic Networks
    2. NEW: Propositional Probabilistic Graphical Models (TeXmacs source)
    3. to come: Relational Probabilistic Models and Logics
11. Cognitive Architectures
    • Agents, Cognitive Architectures and Artificial General Intelligence Projects (OpenOffice presentation source)
    • A Quick and Thorough Introduction to ACT-R 6 (odp source)
    • OpenCog Prime

I’ve moved KR before logics to introduce representation means of representation-specific reasoning systems (NARS and PLN) there.

Considered:

1. Values and Others: Grounding Agents in Game Semantics Δ (to come)
   • We build semantics in both representational and logic aspects based on the notion of agent specific rewards/motivations.

Reviews: Artificial General Intelligence: A Gentle Introduction [2] by Pei Wang

Major online reading:

• Reinforcement Learning: An Introduction [3] Richard S. Sutton and Andrew G. Barto
• Proceedings of the AGIRI Workshop 2006 [4] (Editors: Ben Goertzel, Pei Wang), IOS Press
• AGI-08. Proceedings of the First Conference on Artificial General Intelligence [5], (Editors: Pei Wang, Ben Goertzel, Stan Franklin), IOS Press

More online reading:

• Thinking about Acting [6] by John Pollock
• A Working Hypothesis for General Intelligence [7], Eric B. Baum, October 2006
• Mathematical definition of “intelligence” (and consequences) [8], Warren D. Smith, June 2006

Major offline reading (available to me):

• “Artificial General Intelligence”, Ben Goertzel, Cassion Pennachin (editors), 2007, Cognitive Technologies series at Springer
• “Rigid Flexibility. The Logic of Intelligence”, Pei Wang, 2006, Applied Logic series at Springer
• “Universal Artificial Intelligence. Sequential Decisions based on Algorithmic Probability”, Marcus Hutter, 2005, Texts in Theoretical Computer Science series at Springer
• “Knowledge Representation and the Semantics of Natural Language”, Hermann Helbig, 2006, Cognitive Technologies series at Springer
• “The Cambridge Handbook of Thinking and Reasoning”, Keith Holyoak, Robert Morrison (editors), Cambridge University Press, 2005

Places:

• Artificial General Intelligence Research Institute [9]
• OpenCog [10]
• Association for Uncertainty in Artificial Intelligence [11]
• Adaptive Logics Home Page [12]
• Reinforcement Learning and Artificial Intelligence [13] “RLAI research is research directed toward the long-standing goals of AI (understanding the mind, reproducing human abilities) and is based on reinforcement learning ideas (learning from and while interacting with the world).”

Architectures / projects:

• Novamente [14]
• LIDA [15]
• NARS [16]
• PolyScheme [17]
• SNePS [18]
• LEVELS OF ORGANIZATION IN GENERAL INTELLIGENCE [19] Eliezer S. Yudkowsky
• AdaptiveAI [20]
• Project Joshua Blue [22] (Bootstrapping semantics in an autonomic computing system [21])
• Ai [23]
• SAIL (Self-organizing Autonomous Incremental Learner) [24]
• SHRUTI [25] “From Simple Associations to Systematic Reasoning”
• Overview of the Texai project [26] (Steve Reed)
• SOAR [27]
• ACT-R [28]
• OSCAR

Some video lectures (currently not well selected):

• IBM Research’s Almaden Institute Conference on Cognitive Computing [29]
• Model-based Bayesian RL [30] and related lectures from ICML-07 Bayesian RL Tutorial [31] (site [32])

Other links:

• NIPS 2005 workshop. Towards human-level AI? [33] (with slides)
• Essentials of General Intelligence: The direct path to AGI [34] by Peter Voss
• AGIRI [35] AGI Research Institute wiki
• Marcus Hutter [36] / Towards a Universal Theory of Artificial Intelligence based on Algorithmic Probability and Sequential Decisions [37]
• Juergen Schmidhuber [38] / The New AI: General & Sound & Relevant for Physics [39]
• Site by Włodzisław Duch [40]
• Confabulation Theory [41]
• Logica Universalis [42]
• The Emotion Machine [43]

Attic

Outdated plan:

1. Information, distributions, programs, intelligence.
   1. Shannon information and Kolmogorov information, measures of complexity.
   2. Decision and control theory topics. Markov decision processes, reinforcement learning (Q-learning, SARSA etc.).
      1. SAIL and Dav: robots that learn “from scratch”.
   3. “General algorithmic intelligence” AIXI.
   4. Self improving programs: “Goedel Machine”. “Verificationist” program synthesis.
   5. Graphical probability models.
      1. Bayesian networks.
      2. Hierarchical Temporal Memory from Numenta.
      3. Introduction to “estimation of distribution” algorithms.
   6. “Optimizationist” competent program synthesis: algorithm MOSES.
   7. Representing and learning concepts. PAC-learnability.
      1. Learning grammars.
      2. Higher order and recursive structure representation induction.
2. Logic in a dynamic world.
   1. Adaptive logics overview (circumscription, defeasible argumentation, belief revision, etc.) Intensional and term logics.
      1. From semantic networks to logic: system SNePS.
   2. Reasoning about probability and uncertainty.
      1. Game semantics for logics.
      2. Recursive probability models.
      3. Probabilistic term logic. “Two-dimensional” truth values: system NARS.
      4. “Probabilistic Logic Networks” in Novamente.
   3. Inductive probabilistic logic programming vel probabilistic logic learning.
3. Cognitive loop (in search for the “main()” of the artificial mind).
   1. Inference system as an agent: goals and activations. (SNePS, NARS)
   2. Cognitive loop in LIDA.
   3. Concept formation and modeling of self. (Novamente)
4. Theory of mental development.
   1. Piagetan psychology.
   2. Mental development of an AGI.

Copyright © 2005–2006 the Main wiki and its authors