Generative AI: Useful Tools or Expensive Toys?

Purpose: To consider the uses and hazards of Generative AI systems, such as ChatGPT, in some common research practices.

Context: Generative AI systems now cover text, sound (speech and music), image, and video, generation, and combinations of these. A basic understanding of how these systems work and what they really do is needed to be able to judge well if and how they may be used in professional research practices, such as preparing research publications, funding proposals, presentations and talks, and computer code. Understanding how ChatGPT works and how the (so-called) Large Language Model (LLM) inside it is built does provide a good basis for deciding how and when good use might be made of any of the Generative AI systems. The important ethical issues and hazards raised by using ChatGPT in research practices are also common across the other kinds of Generative AI systems.