CIS 5590.004. Artificial General Intelligence

Basic Inference

1. Inconsistency Management

Inconsistency and its consequence in propositional logic.

Proposed solutions:

Paraconsistent logic: to tolerant inconsistency without deriving arbitrary conclusion.
Belief revision: to recover consistency by removing minimal beliefs.
Non-monotonic logic: to draw defeasible conclusions and to update using new evidence.

2. Uncertain Inference

Traditional models:

Probabilistic and statistical inference: within a probability distribution, except Bayesian conditioning that is distribution update.
Fuzzy logic and other generalization of Boolean functions: triangular norms and conorms.

Type of inference:

Deduction, reference class problem
Induction, Inductive logic programming (ILP)
Abduction, explanations

3. NAL-1 Inference Rules

Local rules: revision and choice
Forward rules: extended Boolean functions, basic syllogistic rules
Backward rules: reversibility

Truth-value functions and Examples of NAL-1

Reading

Non-Axiomatic Logic: A Model of Intelligent Reasoning, 2nd Edition, Chapter 4