Reasoning and Logic
1. Reasoning
Schools of formal model in AI:
Reasoning (inference): deriving new knowledge from existing knowledge
- Descriptive model (of human reasoning): psychological, linguistic
- Normative model (of valid reasoning): logical, probabilistic
Difference between the two types: Reasoning ability in Large Language Models, "What the Tortoise Said to Achilles".
Why to build an AGI in the framework of reasoning system:
- Generality (in representation)
- Justifiability (in steps)
- Flexibility (in processes)
2. Logic
Logic refers to laws/rules of valid reasoning/inference steps (from premises to conclusions), either formal (explicit) or informal (implicit).
A formal logic consists of a representation language (grammar rules), some inference/reasoning rules, and a semantic theory.
- Term logic: categorical sentences, syllogistic rules
- Classical logic, mathematical logic, propositional logic, first-order predicate logic
- Non-Classical logics: logic and types of reasoning, extensions and revisions
Logic and Artificial Intelligence and related debates
(e.g., Nilsson vs. Birnbaum
in Artificial Intelligence, Vol. 47)
3. Inheritance Logic, Layer 1
Reasoning from ideal experience.
- Grammar: term, inheritance copula, statement
- Semantics: experience, knowledge, truth-value, extension/intension, meaning
- Inference: inferring, matching
IL-1 can be implemented as a simple Question-Answering system using graph representation and algorithms.
Reading
- Non-Axiomatic Logic: A Model of Intelligent Reasoning, 2nd Edition, Chapter 2