- Author
- Ted Underwood
- Published
- 2 July 2025
- Kind
- Personal blog essay by a literary scholar
- Original
- tedunderwood.com/2025/07/02/a-more-interesting-upside-of-ai/
- Subject
- Writing with models
What it is
Ted Underwood is a professor of information sciences and English. The blog covers literary history and AI. This post tries to sharpen the case for AI as a cultural technology.
What it says
- Underwood says superintelligence replacing human agency is not a motivating upside for AI.
- Underwood argues human disagreement is load-bearing, so automated systems still need independent critics.
- The post frames AI as a cultural technology, like libraries or printing. It credits the term to Farrell, Gopnik, Shalizi and Evans.
- Underwood says generative AI is a second step change in mapping culture.
- Underwood offers Midjourney style references and Golden Gate Claude as early examples.
- Underwood expects maps of cultural space to be worth less than assistants economically. Underwood says it may not be a great short-term investment.
- The post grants that mapping and editing culture could be misused or become very dark.
What evidence it gives
The post offers argument by analogy to writing and printing. It cites other scholars, including Narayanan and Kapoor, and Farrell, Gopnik, Shalizi and Evans. Its two examples are Midjourney style references and Anthropic's Golden Gate Claude experiment. It has no data or tests of its own.
Interests and limits
The author states that the post is not a cost-benefit weighing. The vision is speculative and the author admits it may fail. The author is a humanities scholar, not an AI developer. The post names no interest in a product. Risks are acknowledged but not analysed.