Learning to Parrot – Workshop with Dennis McNulty – 30 September 6-8pm

FSAS is delighted to host this workshop as the first event in a series considering questions about our institutional relationship to digital technologies and the corporations they tie us to.

Learning to Parrot Workshop with Dennis McNulty, 6pm – 8pm, Wednesday 30 September. This event is free but booking is essential. 

We are living through an era where artificial intelligence (AI) is surfacing in all aspects of people’s lives. These systems are presented as all-knowing and far too complex for the average person to understand. Learning to Parrot is a two-hour workshop that stages a hands-on encounter with the assumptions underlying chatbot technologies such as ChatGPT. The aim is to demystify them and create a space for reflecting on the implications of their use. The title is a reference to the term Stochastic Parrot and the work of Dr. Emily M. Bender.

Practical Details

Participants will need to bring a laptop that can connect to the internet and be familiar with using it. The workshop will last about two hours.

Background

Artificial intelligence (AI) is widely understood to be a marketing term rather than a description of a coherent set of technologies. That said, when non-experts refer to AI these days, they are generally talking about chatbot-enabled text generation systems such as OpenAI’s ChatGPT or Anthropic’s Claude. These kinds of AI products are all based on a technology called Large Language Models (LLMs). As the name might imply, a language model pulls language apart into fragments so that computers can process it. In order to do that, the people who created the language model need to make some assumptions about what language is and how it “works”. As you might imagine, these assumptions have a profound effect on the language that LLM-powered chatbots tend to extrude.

Human writing emerged from a desire to capture knowledge and convey meaning. Over time, the process of writing and reading text has evolved into a rich and complex support for communicating, storing and generating ideas. LLMs are designed to circumvent any requirement to interpret the meaning of a text and instead, leverage the fact that writing takes the form of a sequence of graphic symbols. Like old-school SMS autocomplete, the language models are constructed and fed data so that they can predict what the next word in any given sequence of words is likely be. Chatbot designers believe that meaning will just emerge as a byproduct of generating a sequence of words in a plausible order.

This approach is described as statistical, in that word prediction is based on calculations of probability rather than any communicative intent. Consider whether you would be more likely to believe in the predictive power of a political poll based on the opinions of 10 or 10,000 voters? This statistical approach is the reason why the language models grew large. The engineers believed that analysing more text was the key to building models that could make better predictions about which word should come next.

In this exercise we will work through a very basic example with a short text to develop a sense of the limitations associated with modelling language as a sequence of words, particularly with respect to the implications this approach might have for the idea of meaning.

This event is free but booking is essential. Book here.

About Dennis McNulty

Dennis McNulty is an artist, researcher, music-maker, and instrument designer whose work grapples with know-ability and is often framed with respect to technologies such as language, diagrams or buildings. Learning to Parrot was developed in his role as a research assistant in Prof. Dan Kilper’s group at CONNECT based in Trinity College Dublin, where he explores the role of diagrams in cross-disciplinary collaboration, particularly in quantum networking research.

His artwork has been presented at the São Paulo Bienal, Liverpool Biennial, Performa Biennial, IMMA, Visual, The Dock and Grazer Kunstverein among others. McNulty is currently working on a collaborative Fingal County Council commission to mark EU Ireland’s presidency in 2026. www.dennismcnulty.com