Poking at Ever Larger Language Models

A gentle introduction for (digital) humanists

Language models have emerged as a powerful, fascinating but equally terrifying technology. For decades, interest in these models remained limited to a rather niche research domain within the field of natural language processing. With the recent advent of the so-called Large Language Models (LLMs), attention figuratively exploded. Terms such as GPT (an abbreviation for Generative Pre-trained Transformer) used to be strictly machine learning jargon, but now appear on the front page of newspapers or as a fashionable topic at dinner parties. 

The workshop explores language models (from small to Large) from a critical and practical perspective. We explain how these models are created and deployed, discuss their biases and (anticipated) societal effects, but we also show you how to access and apply LLMs for your own research projects or applications.

According to ChatGPT—currently the famous technology powered by LLMs—“Large language models [...] have emerged as a game-changing technology with the potential to revolutionize the field of digital humanities.” The question remains: will it, really? And if LLMs will “revolutionize” research in DH, in what way exactly? To answer these questions, we closely inspect the modes of access (e.g. open source vs commercial models), capabilities (e.g. in context learning) and types of interaction with LLMs (prompting).

This workshop will be a first experiment to create a better understanding of LLMs among researchers in the (digital) humanities. Some basic knowledge of Python and/or machine learning may be useful but is not necessary. For the practical part you will need a Gmail account to open Colab Notebooks as well as an OpenAI API key. It is recommended that you bring your own laptop, although some computers will be available if you need one.