AI Image Feedback Loop
Data artist Robert Hodgin recently created a feedback loop between Midjourney and ChatGPT-4 — he prompted MJ to create an image of an old man in a messy room wearing a VR headset, asked ChatGPT to describe the image, then fed that description back into MJ to generate another image, and did that 10 times. Here was the first image:
And here’s one of the last images:
Recursive art like this has a long history — see Alvin Lucier’s I Am Sitting in a Room from 1969 — but Hodgin’s project also hints at the challenges facing AI companies seeking to keep their training data free of material created by AI. Ted Chiang has encouraged us to “think of ChatGPT as a blurry jpeg of all the text on the Web”:
It retains much of the information on the Web, in the same way that a jpeg retains much of the information of a higher-resolution image, but, if you’re looking for an exact sequence of bits, you won’t find it; all you will ever get is an approximation. But, because the approximation is presented in the form of grammatical text, which ChatGPT excels at creating, it’s usually acceptable. You’re still looking at a blurry jpeg, but the blurriness occurs in a way that doesn’t make the picture as a whole look less sharp.
And we already know what you get if you recursively save JPEGs…
See also La Demoiselle d’Instagram, I Am Sitting in a Room (with a video camera), Google Image Search Recursion, and Dueling Carls.
Comments 2
This is actually also a fun party game that a group of humans can play -- one person writes a sentence, the next draws the sentence, the next writes a sentence describing the drawing, etc. There is also an online version called Gartic Phone that is excellent for remote work teambuilders.
Now THIS is an interesting play with AI tools that tickles my brain. (cf the Simpsons thing recently).
Hello! In order to comment or fave, you need to be a current kottke.org member. If you'd like to sign up for a membership to support the site and join the conversation, you can explore your options here.
Existing members can sign in here. If you're a former member, you can renew your membership.
Note: If you are a member and tried to log in, it didn't work, and now you're stuck in a neverending login loop of death, try disabling any ad blockers or extensions that you have installed on your browser...sometimes they can interfere with the Memberful links. Still having trouble? Email me!
In order to comment or fave, you need to be a current kottke.org member. Check out your options for renewal.
This is the name that'll be displayed next to comments you make on kottke.org; your email will not be displayed publicly. I'd encourage you to use your real name (or at least your first name and last initial) but you can also pick something that you go by when you participate in communities online. Choose something durable and reasonably unique (not "Me" or "anon"). Please don't change this often. No impersonation.
Note: I'm letting folks change their display names because the membership service that kottke.org uses collects full names and I thought some people might not want their names displayed publicly here. If it gets abused, I might disable this feature.
If you feel like this comment goes against the grain of the community guidelines or is otherwise inappropriate, please let me know and I will take a look at it.
Hello! In order to leave a comment, you need to be a current kottke.org member. If you'd like to sign up for a membership to support the site and join the conversation, you can explore your options here.
Existing members can sign in here. If you're a former member, you can renew your membership.
Note: If you are a member and tried to log in, it didn't work, and now you're stuck in a neverending login loop of death, try disabling any ad blockers or extensions that you have installed on your browser...sometimes they can interfere with the Memberful links. Still having trouble? Email me!