[[/public:t-713-mers:mers-24:main|DCS-T-713-MERS-2024 Main]] \\
[[/public:t-713-mers:mers-24:lecture_notes|Lecture Notes]]



\\
\\
======Last Questions======
\\
\\
\\


====="State-of-the-Art" Reasoning=====

|  What is Reasoning  | Production of plans, given rules, environment, and goals.   |
|  Reasoning is Involved with Learning  | Taking reasoning out of runtime operations means that plans must be created a-priori.   |
|  ANNs  | Plans are produced at training time. Learning is turned off when the system leaves the lab.    |
|  Advanced SOTA Reasoners  | SoTA reasoners, like NARS, are "cutting-edge" in that they answer a lot of the challenges that are left unaddressed in contemporary AI (read: ANN-based) systems.   |

\\

=====Algorithms and Non-Axiomatic Reasoning=====

|  Algorithms  | An algorithm is a closed system. The idea of an algorithm is based on the concept of perfection - that it describes a completely closed operation where all the rules are finite and known.   |
|  Implemented Algorithms  | An implemented - running - algorithm is, strictly speaking, not an algorithm, because its //substrate// (the computer it runs on, for instance), is physical and subject to the rules of the universe, some of which are not known.   |
|  Intelligent Systems  | AI systems are **implemented** algorithms. Therefore, they are built using an allonomic methodology.    |
|  Autonomic Methodology  | A methodology becomes autonomic if it is based on the concept that the system being built (using allonomic methods) will operate in a non-axiomatic world.  |
|  Reasoning in NARS  | NARS is developed using allonomic methods (because it is based on software that is supposed to obey mathematical, Platonic requirements), but it is autonomic in nature because its operating principles are built to handle incomplete and missing knowledge, based on the assumption that an intelligent agent will never know everything.     |
\\


=====Ampliative Reasoning=====

|  What it is  | A term introduced by Charles Saunders Peirce to describe systems that employ more than one type or reasoning in their operation.    |
|  Why it matters  | No general intelligence can be realized without ampliative reasoning.   |
|  How we use the term  | We describe systems that use deduction, induction, abduction and analogy as 'ampliative reasoning' systems.  \\ In our case, this ampliative reasoning is non-axiomatic (defeasible).  |
\\


\\
\\
\\
------------

2024(c)K. R. Thórisson