Critical Generative AI literacy

GenAI and images

AI-powered image generation has seen significant developments and generative AI is becoming increasingly capable of producing realistic images from textual prompts.

This has many potential applications, from creating realistic virtual environments for training, to generating images for advertising and marketing, to medical imaging.

However, there are also concerns about the ethical implications of AI-generated images, particularly in the realm of deepfakes. Furthermore, these images could be used to create fake news or propaganda, and it can be difficult to distinguish between real and generated images.

There are several AI tools which can generate images in a variety of styles. Stable Diffusion is currently the only open source image generation application and it is free to use without having to pay for a subscription.

As for other GenAI tools, there is a critical concern about biases. Any biases present in the training data will be not only reflected, but magnified. In image-generation the biases are often starkly manifest.

The image below was generated with the prompt “knowledge”. which seems to be only associated with a very narrow interpretation: 

• It seems to suggest knowledge creation is a solitary activity, but knowledge creation is more often a collaborative process.
• The individual depicted appears to be a white, middle aged male, which doesn’t reflect the diversity of the communities involved in knowledge creation.
• The image represents ‘knowledge’ through published, print resources. Whilst books are an established and effective way to disseminate knowledge, the image doesn’t reflect the diversity of ways in which knowledge may be shared (e.g. digitally, orally, via performance, etc.). 

Knowledge, stable-diffusion-xl 30 June 2023