You may have noticed that several of my recent blog posts have featured custom AI art work related to the topic. I’ve generated these using Microsoft’s Bing AI Image Creator, which uses OpenAI’s DALL-E text-to-image model. DALL-E can generate an image based on a text prompt; for example, the featured image on this post was ‘A 1950s style robot standing in front of an easel painting a bouquet of flowers in a vase’.
These are some of the other AI art images that I’ve used recently:
AI art is controversial. It can create images in a few seconds that would take a human artist hours or days to produce. And, in some cases, these image prompts can be told to create images in the style of a particular artist, depriving them of income from a commission. It’s also notable that models like DALL-E and Stable Diffusion have been trained on copyright works, without the rights holders’ permission.
With this in mind, I’m justifying my use of AI art on some of my blog posts because I’m not an artist myself, and as an individual blogger who doesn’t make money from blogging, I wouldn’t have the money to pay a human artist. Whilst I have over 5000 photos that I’ve uploaded to Flickr, there isn’t always a relevant photo to use that I have taken. For example, in my recent post on comment spam, I decided to generate the above image of a robot converting blog posts into a tinned meat product, because I don’t have a photo that represents that. And whilst I make use of screenshots where relevant, sometimes this isn’t appropriate.
Of the AI art generators that I have used, the Bing AI Image Creator seems to be the one that gives me the best results. Any images you create are saved in the cloud, and can be downloaded for re-use. And each prompt produces four images so that you can choose the one which looks the best.