Language Models as a Knowledge Source for Cognitive Agents (2109.08270v3)

Abstract: LLMs (LMs) are sentence-completion engines trained on massive corpora. LMs have emerged as a significant breakthrough in natural-language processing, providing capabilities that go far beyond sentence completion including question answering, summarization, and natural-language inference. While many of these capabilities have potential application to cognitive systems, exploiting LLMs as a source of task knowledge, especially for task learning, offers significant, near-term benefits. We introduce LLMs and the various tasks to which they have been applied and then review methods of knowledge extraction from LLMs. The resulting analysis outlines both the challenges and opportunities for using LLMs as a new knowledge source for cognitive systems. It also identifies possible ways to improve knowledge extraction from LLMs using the capabilities provided by cognitive systems. Central to success will be the ability of a cognitive agent to itself learn an abstract model of the knowledge implicit in the LM as well as methods to extract high-quality knowledge effectively and efficiently. To illustrate, we introduce a hypothetical robot agent and describe how LLMs could extend its task knowledge and improve its performance and the kinds of knowledge and methods the agent can use to exploit the knowledge within a LLM.