Load "external" data in a browser-based text experiment. Document your experience working with the library, API, or data source and consider the following questions (stemming from the Excavating AI reading).
For this exercise I wanted to explore one of the word-based API’s. I went with Datamuse, but intend to explore WordNet as well.
In my class, Project Development Studio, I’m developing a game that explores and highlights the inherently political act of training computers to “see”. The work of Kate Crawford and Trevor Paglen, Excavating AI: The Politics of Images in Machine Learning Training Sets, turned out to be highly relevant for me this week. In the piece they trace the major historical landmarks in computer vision research, and detail the importance of why this history matters for how we understand many of the LLMs and Gen-AI tools we’re using today.
The circuit between image, label, and referent is flexible and can be reconstructed in any number of ways to do different kinds of work. What’s more, those circuits can change over time as the cultural context of an image shifts, and can mean different things depending on who looks, and where they are located. Images are open to interpretation and reinterpretation. - Kate Crawford and Trevor Paglen
For this exercise I decided to try and create an exercise which deals with the challenges of defining taxonomies to categorize words. I thought that this may help me sort out how to translate this kind of practice into a game. Although their work focuses on the act of labelling images, I think linguistic-relativity is equally relevant in the overall conversation of datasets and tagging them.
I started with the idea of defining “classes” as categories, and to use Datamuse ****to fetch related terms, thinking "apple" could be tagged with classes like "fruit", "tech", or "symbolic (knowledge)".