Abstract: The interdisciplinary pilot project `Exploring novel figurative language to conceptualise Large Language Models’ is funded by Cambridge Language Sciences. This talk mainly concerns `slop', by which we mean text delivered to a reader which is of little or no value to them (or is even harmful) or is so verbose or convoluted that the value is hidden. Examples range from unnecessarily obscure legal agreements to academic papers in predatory journals. Intuitively, slop has negligible information density for the reader. Although the term `slop' is often used specifically for generative AI output, we argue that slop predates the machine-generation of text and has to be understood in the context of wider societal processes. Modern ways of delivering text are driving an increase in slop. In this context, systems incorporating LLMs have the capacity to be `supersloppers' --- tools for the creation and delivery of more and more pointless text.
Depending on time, I will discuss some other examples of figurative language and other ways of exploring aspects of LLMs. This will be a very informal talk - I am especially keen to hear about metaphors that other people find helpful (or unhelpful).
Ann Copestake, Lucy Duggan, Aurelie Herbelot, Amira Moeding and Eva von Redecker (2024). `LLMs as supersloppers' Cambridge Open Engage. doi:10.33774/coe-2024-dx12p
