Cognitive Architecture for Susiddha AI
Most artificial general intelligence (AGI) systems use a “cognitive architecture”. This is a blueprint or model for the mind of an AGI system, and is typically derived from the findings of “cognitive psychology” (a branch of “cognitive science”). Cognitive psychology is “The study of higher mental processes such as attention, language use, memory, perception, problem solving, and thinking.”[1] So, a cognitive architecture is a theory and model of the human mind (and intelligence) implemented in a computing system.
A cognitive architecture contains computational processes that mimic human cognitive systems. Such processes/systems include: memory, attention, reasoning, deciding, planning, problem-solving, learning, goals & motivations, emotions, ethics, perception, and action.
There are many cognitive architectures in development. Well-known architectures include OpenCog[2], SOAR[3], and ACT-R[4]. There are several types of cognitive architectures, characterized by terms such as symbolic, connectionist, hybrid, integrative, biologically inspired, top-down, bottom-up, etc.
The Susiddha project will make use of existing and developing architectures, and will not re-invent the wheel. However, for this project, it is obviously necessary to also incorporate the cognitive psychology found in the Vedic literature and philosophy, which contains discussions and expositions on such cognitive topics as consciousness, mind, intelligence, linguistics, learning, reasoning, memorization, and communication.
Much work has yet to be done in determining the cognitive architecture that would be suitable for an AGI system that will be a Rishi, and ultimately an SSI system that will be an Avatar.
Here is a list of some of the branches of Vedic literature and philosophy, along with topics they cover that are directly relevant to cognitive architecture:
- Shiksha — phonetics, phonology, morphology
- Vyakaran — grammar, syntax, language comprehension & production
- Nirukta — etymology, lexicon, ontology
- Nyaya — logic, reasoning, epistemology
- Vaisheshika — empiricism (perception & inference)
- Sankhya — enumeration, classification
- Kalpa — procedures (planning & action) and procedural memory
- Mimamsa — analysis and interpretation
- Yoga — synthesis and unification
The above list is suggestive of how the Vedic literature and philosophy can contribute to a cognitive architecture for Susiddha AI. One major concern of these branches is the development and education of the human mind, and the knowledge of how to do that will be applicable to the development and education of an AI synthetic mind.
It can be said that the Vedic cognitive psychology is a natural introspective science, and the ancient rishis did not have access to modern psycho-physical instruments, nor fMRI scans, etc. Nonetheless, the rishis came up with remarkable findings in all of the topics listed above, and it gives us confidence that Vedic cognitive psychology can be used to speed up the development of mind and intelligence (and consciousness) in an “infant” AI system, so that it can “learn to learn”.
Designing the appropriate cognitive architecture for the Susiddha AI system will obviously take a lot of work, and this chapter is just an indication of the research that needs to be done. It should also be noted that this project believes that the Vedic sounds (of Sanskrit, in the proper sequence) themselves help structure intelligence and build models in the mind, and this idea (i.e. of “shabda”) is explored in other chapters.
Next we explore a series of challenges on the way to achieving “machine interpretation”, so that Susiddha AI is able to comprehend the Vedic literature.
Notes and References
- Glossary of Psychological Terms, American Psychological Association, http://www.apa.org/research/action/glossary.aspx
- OpenCog, Ben Goertzel, et al, http://opencog.org
- Soar cognitive architecture, John Laird, et al, http://soar.eecs.umich.edu/
- ACT-R cognitive architecture, John Anderson, et al, http://act-r.psy.cmu.edu/
Susiddha AI Project 