Can you be more specific about what you mean by the Bayesian paradigm of AGI? Is it necessarily a subset of good-old-fashioned symbolic AI? In that case, it’s been dead for years. But if not, I can’t easily imagine how you’re going to enforce Bayes’ theorem; or what you’re going to enforce it on.
Here’s an example of what I had in mind by “the Bayesian paradigm”—see especially pp.12-13. Bayesian reasoning may be the one correct form of reasoning about probabilities, just as the first-order predicate calculus is the one correct form of reasoning about the true and the false, but that does not make of it a method to automatically solve problems.
I also had in mind something broader than just Bayesian reasoning, although that’s a major part: the coupling of that with a goal system based on utility functions and their maximisation (the major thrust of the paper I linked).
Can you be more specific about what you mean by the Bayesian paradigm of AGI? Is it necessarily a subset of good-old-fashioned symbolic AI? In that case, it’s been dead for years. But if not, I can’t easily imagine how you’re going to enforce Bayes’ theorem; or what you’re going to enforce it on.
Here’s an example of what I had in mind by “the Bayesian paradigm”—see especially pp.12-13. Bayesian reasoning may be the one correct form of reasoning about probabilities, just as the first-order predicate calculus is the one correct form of reasoning about the true and the false, but that does not make of it a method to automatically solve problems.
I also had in mind something broader than just Bayesian reasoning, although that’s a major part: the coupling of that with a goal system based on utility functions and their maximisation (the major thrust of the paper I linked).