Generative AI

Summary

Generative AI (GenAI) is an umbrella term for a collection of deep-learning computer models that take large amounts of training data that they analyze computationally and then produce (or “generate”) similar outputs based on a large number of statistical calculations. Popular AI services, including ChatGPT, Midjourney, Gemini, Suno, Ernie, and Sora, will generate content, such as text images, music, and audiovisual material, in response to prompts provided by users. Substantial hype has accompanied GenAI, particularly as its capabilities and prevalence have grown and as promoted by its developers and partners. This positive discourse positions GenAI as both a potential solution to major social and economic challenges and as an inevitable part of everyday technological use. However, there are also numerous concerns and critiques of GenAI that highlight key issues around its impact as a technology and as an infrastructure. Such concerns include questions of bias in how GenAI is trained and how they may generate sexist, racist, ableist, and varying offensive and discriminatory representations; issues of accuracy and the capacity for GenAI to be used to create and share misinformation; the quality and value of work created used GenAI; uncertainty about copyright and the rights of creators whose work has been used to train GenAI; and extractivist critiques of GenAI and its environmental impact, the infrastructure and labor needed to physically maintain GenAI systems, and the data on which it is dependent. Such themes highlight that critical investigations into GenAI should not just consider the technology through its interfaces and outputs but also take into account its users, designers, infrastructures, and surrounding sociocultural, political, economic, and historical contexts.