Twitter dreams of AI: understanding AI futurity through sentence embeddings

Abstract

This study uses a novel approach, Sentence BERT (SBERT) to examine Twitter (X) discourse around artificial intelligence (AI). As AI becomes increasingly prevalent, use, development, and conflict have unfolded rapidly, generating both a boom in dialogue regarding what a future with AI would look like and increasingly uncertain definitions of what constitutes AI. To better understand how this discourse emerged, I collected around 900k tweets from 2021 containing the term ‘artificial intelligence’ and utilized a two-stage topic model with SBERT to extract key themes. I then analyze tweets discussing medicine and ethics for specific discourses. While existing work around AI futurity has utilized traditional journalism or policy memos, discourse originating from social media remains understudied. I contend that computational text analysis and SBERT offer solutions to both the short-text nature of platforms such as Twitter and the volume of tweets. Thus, this paper shows how AI has created contested and fluid notions of futurity and expertise with a focus on medicine and ethics.