ChatGPT prompts for accelerating work in morphology and taxonomic revisions

Large language models are a novel class of artificial intelligence tools trained on very large text datasets and capable of processing natural language. One of these tools, chatGPT, is structured as a chat box, in which a user can input instructions in natural language and the computer responds. Especially in its latest iteration (GPT4, launched in 2023), this tools has yielded impressive results and called a large attention. In this talk we will briefly present how large language models work and what they can and cannot do well at the moment. In particular, responses from these kinds of models always sound correct, even when they are false, a phenomenon known as hallucination that needs to be avoided in scientific contexts. We will then show some examples of user prompts crafted to accelerate work in morphology, including parsing information between natural language descriptions and structured tables, standardizing terminology, interpreting ICZN rules in context, comparing descriptions in different languages, and producing keys, diagnoses and lists of specimens. We will discuss the promises and pitfalls of these prompts with practical guidance for taxonomists to make the best use of these tools to minimize effort in timing-consuming steps and be able to focus their attention on tasks in which expert humans excel machines.