Knowledge Engineering in the Language Model Era

In this talk, I contend that the emergence of large language models (LLMs) (e.g. GPT3, T5) as both a source of knowledge and an an effective tool for extracting knowledge provides new opportunities for knowledge engineering. First, by helping to address the cost knowledge acquisition, LLMs let us focus on other aspects of knowledge engineering - understanding the user, ensuring that the knowledge production lifecycle is reliable, dealing with curation. Second, the accessibility of LLMs means that more people could benefit from knowledge engineering practices. This raises the question of how to disseminate those practices. Here, I take the example of prompt engineering.