Summary:

- The article discusses how large AI models, such as GPT-3, are not just technological innovations, but also cultural and social technologies that have significant societal impacts.
- The author argues that these AI models are shaped by the data and assumptions used to train them, which can reflect and perpetuate biases and power structures in society.
- The article emphasizes the need for a more critical and interdisciplinary approach to understanding and governing the development and deployment of large AI models, considering their cultural and social implications.