T-720-ATAI, ADVANCED TOPICS IN ARTIFICIAL GENERAL INTELLIGENCE, FALL 2012

Instructor: Kristinn R. Thórisson
Teaching Assistant: Deon Garrett
8 ECTS Units, full Master's-level course
Days: Tuesdays & Fridays
Time: 12:20 - 14:00
Classroom: M112

The course focuses on recent advances in holistic intelligence architectures, often called artificial general intelligence (AGI). At the core are questions originally asked by the founders of the field, such as Turing, Minsky and others, in the very early days of the field: What is intelligence? and How can we implement intelligence in a machine? Issues related to intelligence architecture, system autonomy, realtime attention, anytime planning, and model-based knowledge representation, with concepts from systems theory, autopoiesis, constructivist AI, and cybernetics providing key foundational concepts. Historical background and relevant topics from constructionist AI (good old-fashioned AI) provide a backdrop to the challenge building of more autonomous and self-contained intelligent systems; relevance of AGI to autonomous robotics and systems operating in the real-world will be discussed.

• Programming experience necessary, LISP, Prolog or Haskel a plus
• A prior introductory class in one or more of the following is recommended: Artificial intelligence, Simulation techniques, cognitive science

After taking the course, diligently attended the classes, thoroughly read reading and homework assignments,

• students should be able to:
  • identify key challenging research questions related to AGI;
  • list the methodological difficulties and known solutions to building AGI systems;
  • apply presently-known techniques and methodologies for building complex AI systems;
  • explain some key components of AGI architectures and how these relate to the creation of truly intelligent machines of the future;
• furhermore, students should have a good idea of:
  • how AGI differs from “narrow AI”;
  • what requirements exist for building complete minds;
  • what methodologies are available for building complete minds;
  • how software architecture plays a central role in AGI (and AI);
  • some AGI projects in progress;
• and finally, students should have some experience with:
  • one programming environment targeting AGI;
  • the phenomenon of emergence, and why it is relevant to AGI.

• Readings and Study Material page - readings and material organized by topic.
• To see readings in temporal order, see schedule below on this page.
• Readings for the class will be published incrementally, as needed.

Note: This assignment outline is indicative only; until September 31st some details of these assignments, and their percentage of total grade, may be changed.

• Students should hand in their assignments (using MySchool) on time; if there will be any unnecessary delay in handing in the assignment then students may assume a lower grade as a result.

• Regular Discussions of reading material will be held, for 30-40 minutes every Friday.
• For discussions we will use the forum on proboards.
• After most Tuesday classes the instructor will post a link on the course online forum to a paper or article on an interesting aspect of Artificial General Intelligence (which may or may not be directly related to the topic being covered that week). You need to read this paper or article by Thursday at midnight and post, under the same forum thread, 2 questions you have about the contents of the reading. This post has to arrive by 23:59 on Thursday night to count towards your paper discussion grade. The questions can point out concepts that you have difficulty understanding, but preferably they should be questions that provoke discussion from the material. In the discussion section of the following Tuesday class, chosen questions from those submitted will be discussed by the group as a whole and you are expected to participate.
• Participation is mandatory
• Grading: 20% of final grade
  • each topic counts equally towards the grade
  • 2 discussions may be omitted with no effect on grade
  • in-person attendance at 2 discussion classes may be omitted with no effect on grade

• Four small programming projects in the NARS environment (see examples).
• Grading: 25% of final grade.
• Programming Projects Resources Page

• One short essay.
• Grading: 15% of final grade.
• Short essay page

• Draft Due: Nov 6 (Mandatory, but not part of grade)
• Final Version Due: Before midnight Nov 30

• Final Exam will count 40% of final grade (includes 2 bonus points, each counting 0,25 out of 10,0).
  • A grade of 4,75 or higher is required to pass the course.
  • Handing in the assignments is not a requirement to get permission to take the final exam.
  • All material covered may be subject of the Final Exam.
  • Assume to be able to draw architectural diagrams, write pseudo-code, and write (short) essays.
  • No helping material is allowed.
  • Example Final Exam Questions

• Programming assignments
  • Prog Assignm 1: Due before midnight Nov 27
  • Prog Assignm 2: Due before midnight Dec 3
• Short essay:
  • Choice of topic: by noon, Tuesday September 18
  • Draft Due: Nov 6 (Mandatory, but not part of grade)
  • Final Version Due: Before midnight Nov 30

Aug Tue 22: Introduction to T-720-ATAI
Aug Fri 24: What is artificial general intelligence? | Readings:

• Artificial General Intelligence Research Institute
• Journal of Artificial General Intelligence home page.
• Does the Future of AGI Lie in Cognitive Synergy? by B. Goertzel
• Four Basic Questions about Artificial Intelligence by P. Wang

Aug Tue 28: What is intelligence? | Readings:

• The g factor on Wikipedia.
• Multiple theory of intelligence on Wikipedia.
• Strong AI on Wikipedia.

Aug Fri 31: What is intelligence? | Remember to post 2 questions to the forum | Readings:

• Cognition on Wikipedia.
• Situated cognition on Wikipedia.
• Animal cognition on Wikipedia.
• A Computational Foundation for the Study of Cognition by D. Chalmers

Sep Tue 04: Requirements: How do we build an intelligent machine? | Readings:

• Requirements for deliberative systems by A. Sloman – key sections: from 8 onwards.
• Cognitive Architecture Requirements for Achieving AGI by J.E. Laird et al.
• Architectural Requirements for Human-like Agents Both Natural and Artificial by A. Sloman

Sep Fri 07: Methodology: How do we build an intelligent machine? | Readings:

• Introduction to Software Architecture by Garlan & Shaw
• Second-Order Cybernetics by Heylighen, F. & C. Joslyn

Sep Tue 11: What is constructionist AI? | Readings:

• Constructionist Design Methodology paper by K.R. Thórisson
• A Mind Model for Multimodal Communicative Creatures and Humanoids by K.R. Thórisson
• Beliefs, Desires, Intentions - model for agents on Wikipedia
• Subsumption architecture by R. Brooks.

Sep Fri 14: Why is architecture critical in AI? | Readings:

• An integrated theory of mind by Anderson et al.
• AGI architecture review by Duch et al.

Sep Tue 18: How is architecture important in AI?

• Architectural Mismatch by Garlan, Alan & Ockerbloom
• Achieving Artificial General Intelligence Through Peewee Granularity By K.R. Thórisson & E. Nivel.
• Self-Programming: Operationalizing Autonomy by E. Nivel & K.R. Thórisson.

Sep Fri 21: What architectures have been built? | Readings:

• JAGI paper on Autonomy by K.R. Thórisson & H.P. Helgason

Sep Tue 25: NC
Sep Fri 28: NC

Oct Tue 02: What is NARS? | Readings:

• Online intro to NARS by P. Wang
• NARS intro paper by P. Wang

Oct Fri 05: Intro to Prolog and NARS | Instructor: Deon Garrett | Readings:

• Unification and deduction in Prolog; Prolog can only make deductions it can prove correct
• Prolog & NARS materials

Oct Tue 09: Intro to NARS | Instructor: Deon Garrett | Readings:

• NARS demos by P. Wang
• NAL-1, and major extensions to NAL-1: the new copula like equivalence, sets, and higher-order representation
• Intro to Progr. assignments
• Programming Assignment 1

Oct Fri 12: What is constructivist AI? | Remember to post 3 questions to the forum | Readings:

• A New Constructivist AI by K.R. Thórisson

Oct Tue 16: What is constructivist AI? - part 2 | Readings:

• A New Constructivist AI by K.R. Thórisson

Oct Fri 19: What is emergence and how is it relevant to AGI? | Readings:

• Emergence explained on Wikipedia
• Four principles of multi-cellular systems on evolutionofcomputing.org

Oct Tue 23: More on constructivist AI | Readings:

• (No new readings)

Oct Fri 26: AGI Programming Assignments session | Instructor: Deon Garrett

Oct Tue 30: Attention | Instructor: Helgi Páll Helgason | Readings:

• Overview of human attention on Wikipedia.
• Broadbent's filter model of attention on Wikipedia.
• Computational models of primate visual attention on Scholarpedia.
• Attention slides by Helgi Páll Helgason

Nov Fri 02: Flexible resource management: Attention for AGIs | Instructor: Helgi Páll Helgason | Readings:

• A three-layer model of selective attention by Mancas et al.
• A model of attention that takes global scene factors into account by Torralba.
• On Attention Mechanisms for AGI Architectures : A Design Proposal by H.P. Helgason, E. Nivel & K.R. Thórisson
• Attention part II slides by Helgi Páll Helgason

Nov Tue 06: Autonomy | Instructor: Helgi Páll Helgason | Readings:

• JAGI paper on Autonomy by K.R. Thórisson & H.P. Helgason
• Self Programming: Operationalizing Autonomy by Nivel & Thórisson
• A Framework for Autonomy Levels For Unmanned Systems (ALFUS) by Huang et al
• Fridges, Elephants and the Meaning of Autonomy and Intelligence by Ricardo Sanz et al
• Autonomy slides by Helgi Páll Helgason

Nov Tue 09: No Class

Nov Fri 13: Topics in AGI: What is creativity? | Readings:

• The very laws of physics imply that artificial intelligence must be possible. What's holding us up? by David Deutch
• Evolution of Creativity: A Summary Report by Thorisson & Thórisson

Nov Fri 16: Situatedness, embodiment, and embededness | Readings:

• When is a cognitive system embodied? by Riegler
• Social situatedness by Lindblom and Ziemke
• Don't forget to post three questions to the pboard before 11 am Friday Nov 16.

Nov Tue 20: Topics in constructivist AI architectures: Self-representation | Readings:

• Remote Agent: to boldly go where no AI system has gone before by Muscoletta et al.

Nov Fri 23: AGI Programming Assignments session | Instructor: Deon Garrett

• The second half of the programming assignment will be presented
• NARS refresher + Q&A: Bring some questions of your own!

Nov Tue 27: How do we measure and evaluate AGIs? | Readings:

• Psychometric Artificial General Intelligence: The Piaget-MacGuyver Room by Bringsjord, S. & Licato, J. (2012). In Theoretical Foundations of Artificial General Intelligence, edited by P. Wang and B. Goertzel (Atlantis Press).
• Don't forget to post three questions to the pboard before 10 am Friday Nov 27.

Nov Fri 30: What is the AERA system? & Course Review | Readings:

• Videos of S1/AERA on the HUMANOBS.org website wiki.
• Take a close look at the course notes

Dec Wed 05: Final Exam, 14:00-16:00, in V201

EOF