Position: Reasoning is a Learnable Rule-Based Process

Autonomous artificial intelligence is one of the most captivating and motivating topics today, both from a scientific and economic perspective. Historically, reasoning has been an area of focus for symbolic artificial intelligence, but recent advances have primarily emerged from deep probabilistic generative models. Despite significant interest and rapid progress in this field, the artificial intelligence generative community has yet to arrive at a clear operational definition for reasoning, often implicitly rejecting the historical approach to this topic in verifiable automated reasoning and logic. This situation generates ambiguity in the definition of reasoning, which in turn hinders quantifiable evaluation and progress toward reliable autonomous reasoning. It is proposed that reasoning be considered a rule-based process that can be learned, and operational definitions and a list of recommended practices for communicating research on reasoning in artificial intelligence are provided. This clarification is crucial as it enables progress toward more reliable and transparent autonomous reasoning, which can have a significant impact on the reliability and security of artificial intelligence systems in various applications. A clear understanding of reasoning in artificial intelligence is essential for developing systems capable of making informed and precise decisions, which can have a positive impact on areas such as automated decision-making and complex problem-solving.

Read the original article on arXiv cs.AI

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.