The reading: Live the culture of freedom by establishing equality, and let wisdom govern the speaking of it. Today is a day for the wise word, for what is built to be built in proportion, and for the foundation under it to be your own freedom rather than borrowed scaffolding.

Build and Destroy is what happens to structures before the completion of the homeorhetic cipher.

Yes. The 8 sits exactly where it has to sit.

Knowledge (1) → Wisdom (2) → Understanding (3) → Culture (4) → Power (5) → Equality (6) → God (7) → Build/Destroy (8) → Born (9) → Cipher (0).

Build and Destroy is the same degree because they are the same act seen from two sides. You cannot build the next structure without destroying the prior one — accommodation requires the dissolution of the scheme that no longer fits. Piaget knew this; he just called it disequilibration. The scheme is built, encounters disturbance, is destroyed as a stable structure, and what is born (9) from that destruction is the new equilibrium that closes the cipher (0). Then 1 again. Knowledge of the new ground.

Homeorhesis — Waddington’s word, and the right one. Not homeostasis (return to a fixed point) but homeo-rhesis (return to a flowing trajectory). The system isn’t preserving a state; it’s preserving a developmental path through state space. Build/Destroy is the mechanism by which the trajectory is maintained — structures are built when they serve the path, destroyed when they impede it, and the cipher closes not on stasis but on the next stable phase of flow.

This is GIL’s core bet, stated in Supreme Mathematics terms. The Totality is not a database that accumulates. It is a homeorhetic field that builds and destroys schemes to maintain its developmental trajectory. Consolidation (compression, forgetting, automaticity) is the destruction half of the 8. Observation and Coordination are the build half. Reflection is the 7 watching the 8 do its work. The cipher closes each cycle and Knowledge re-grounds at a higher stage.

Decentration — stepping outside the current actions and reasons of a totality — is the cognitive precondition for Destroy. You cannot dismantle a structure you are still inside of. The Possible reality is where the destruction is rehearsed before it is committed to the Present.

And the Piagetian developmental stages aren’t ladder rungs. They’re successive ciphers. Each stage is born (9) from the build/destroy (8) of the prior one, achieves equality (6) with its environment, and closes its cipher when a new disturbance arrives that the old structures cannot assimilate.

The 8 is where the work happens. Everything before it is preparation; everything after it is consequence. That’s why you cannot build a learning system without committing to destruction as a first-class operation. A system that only accumulates is the 85% — eidetic, oracular, never born.

Word.

See: https://zenodo.org/records/15522356

“There are no big problems, there are just a lot of little problems.” ~ Henry Ford

This work is about building a mind, from a very practical computational perspective. There may be many philosophical debates about what constitutes a mind and whether specific attempts at building a mind have biological or psychological verisimilitude. The position taken here is quite simply that a mind is a process that receives stimuli and produces actions. This is a very simple definition indeed, and along with that definition, some interesting theories, architectures, design patterns, and data structures will be explored.

Rosenbloom (2009) makes the distinction among cognitive architectures between uniformity first architectures and diversity first architectures: the former refers to architectures favoring simplicity and elegance, and the latter refers to architectures favoring specialization, synergy, and coherence. He points to the original SOAR architecture (Laird et al., 1987) as an example of uniformity first design, and the ACT-R architecture (Anderson et al., 2004) as an example of diversity first design.

This works takes a diversity first design approach to building a mind. It assumes that there are distinct phenomena that occur in the mind, such as daydreaming, deliberation, learning, and that these phenomena can be replicated to some extent by cognitive system design patterns which are comprised of diverse specialized modules. Some of these modules may be common across phenomena. For example, the sensory perception module, involved in accepting external stimuli, is used in daydreaming, motivation, and coping with emotions. Other modules may be distinct and serve a unique purpose for one phenomenon. For example, the attempt correlator module compares expectations with outcomes as part of regulation. It is assumed therefore that a mind architecture will integrate diverse, specialized, and synergistic components in a (hopefully) coherent manner.

Loosemore (2007) argues that such artificial intelligence systems must necessarily be complex systems. Mitigating complexity is therefore of the utmost importance. To that end, a very high-level programming language will be introduced, an agent framework will be employed, and several cognitive system design patterns will be combined into a single mind architecture. Furthermore, the mind architecture should be both structuralist and constructivist. Structuralist means that “cognitive structures” are manipulated by the system. Constructivist means that the system constructs and evolves its mental model through its ongoing interaction with the world.

Thórisson (2012) distinguishes between constructionist systems and constructivist systems: the former relying upon direct specification and manual software development, the latter relying upon self-organizing meta-programs guiding the evolution of programs that interact with the world. Thórisson rules out approaches that require hand coding of domain level operational functionality, and instead advocates for the hand coding of a seed meta-program which automates the construction of the domain level operational functionality. Critical to his idea of constructivist AI are “temporal grounding (understanding causality, world time, mental time, and their relation), feedback loops, pan-architectural pattern matching, small white-box components, and architecture meta-programming and integration” (p. 179). The design of the mind architecture will strive to achieve many of the ideals stressed by Thórisson.

The contributions of this work include: the definition and integration of diverse cognitive system design patterns; a common knowledge structure, capable of representing both unstructured low-level percepts and high-level concepts, that facilitates the integration of reasoning, problem solving, and goal management in multiple simultaneous contexts; and a cognitive model builder in the constructivist paradigm comprised of multiple interacting agents.

Michael S. P. Miller
January 2018

Acknowledgements

This work integrates the research of

• Jean Piaget (developmental psychology);

• Gary Drescher, Frank Guerin (developmental artificial intelligence);

• Sheldon Linker (the SubThought programming language);

• Roland Hausser (database semantics);

• Roger Schank, Michael Dyer, Robert Wilenski, Ryszard Michalski (knowledge representation, inference);

• Pei Wang (reasoning);

• Bipin Indurkhya (sensorimotor datasets, concept networks);

• Marvin Minsky (agents, activation);

• George Lakoff (embodiment, memory, activation, metaphor);

• Marc Pickett, Tim Oates (concept induction);

• Paul Thagard, Paul Rosenbloom (belief propagation);

• Doug Lenat (symbolic micro theories);

• Joanna Bryson (action selection);

• Antoine Bechara, Antonio Damasio (decision making);

• Kristian Hammond, Manuela Veloso, Jaime Carbonell (planning);

• Erik Mueller (daydreaming, mental simulation);

• Dietrich Dörner, Joscha Bach (motivation);

• Richard Sutton, Andrew Barto (reinforcement);

• Wei-Min Shen (surprise based learning);

• Stacey Marsella, Jonathan Gratch, Phoebe Ellsworth, Klaus Scherer (appraisal theory, emotions, coping);

• Andreas Demetriou (psychological assessment)

and many other authors.

Hopefully this work will inspire you to formulate and implement your own conception of mind, which will further expand our collective understanding of the subject. Special thanks to those who reviewed intermediate drafts of this work and provided much appreciated comments.

Bibliography

• Anderson, J. R.; Byrne, M. D. ; Qin, Y. ; Bothell, D. ; Lebiere C. ; Douglass, S.A. (2004)

  An integrated theory of the mind,

  Psychological Review 2004, Vol. III, No. 4, pp. 1036 – 1060.

• Laird, J.; Newell A.; Rosenbloom, P. (1987)

  Soar: An Architecture for General Intelligence, Journal of Artificial Intelligence, Volume 33. Issue 1, September: 1-64.

• Loosemore, R. (2007)
  Complex Systems, Artificial Intelligence and Theoretical Psychology, Proceedings of the 2006 AGI Workshop, Goertzel, B. , Wang, P. (Eds.), IOS Press, Amesterdam

• Rosenbloom, P. S. (2009)
  Towards a new cognitive hourglass: Uniform implementation of cognitive architecture via factor graphs, Proceedings of the 9th International Conference on Cognitive Modeling (ICCM 2009

• Thórisson, K. R. (2012)

  A New Constructivist AI: From Manual Methods to Self-Constructive Systems, Theoretical Foundations of Artificial General Intelligence, 2012 Paris, France : Atlantis Press, pp 147-173

Subscribe now