CIS 5590.004. Artificial General Intelligence

Scope and Boundary

1. A(G)I Research

• McCullough and Pitts: artificial neuron
• Wiener: cybernetics
• Shannon: information theory
• Turing: intelligence as human-like behaviors
• Von Neumann: The Computer and the Brain

• Level of abstraction (structure, behavior, capability, function, principle)
• Selectivity in abstracting (Which one?)
• Position in the "science/technique spectrum"

• There are different routes to the same summit
• There are aspects to be integrated
• There is a correct definition

• "Look, Ma, no hands!"
• ELIZA Effect
• "Turing Test Considered Harmful"
• AI Effect
• Wishful mnemonics
• "The Bitter Lesson"

• Began at the fringe: initial publications and meetings (2006-2009)
• Popularized by the large companies: AlphaGo, ChatGPT, etc. (2012-2023)
• Current hot-spots: LLMs ("World Models", AI agents), Alberta Plan, Neuro-symbolic hybrids
• An objection to the notion of AGI and its preexisting responses

2. NAL and NARS

• A General Theory of Intelligence (2009), Chinese version (2026)
• From NARS to a Thinking Machine (2006), The Roadmap of NARS (2019)

• Categorical language
• Experience-grounded semantics
• Syllogistic rules
• Compositional rules
• Temporal/procedural/backward inference

• Realistic working environment
• Concept-centered knowledge representation
• Dynamic resource allocation
• Case-by-case problem solving
• Relative rationality

• Implementing top-layers of NAL (NLP, math, sensorimotor, self-control)
• Testing, tuning, educating
• Integrating with other techniques (e.g., DL, LLM, agents, robots)
• Applications (partial and customized)

Reading

• Non-Axiomatic Logic: A Model of Intelligent Reasoning, 2nd Edition, Chapter 15