Last weekend, Ian's girlfriend Sarah took a photograph of the three little cousins together. They were in our front room round the coffee table. From left to right there's Zachary, Margot and Phoebe. Sarah uploaded the image into an A.I. facility and requested a Japanese Anime version of it. This charming picture is what emerged a few seconds later.
There are one or two significant differences between the anime version and the original image. For example, the A.I. facility has turned Phoebe's cuddly sloth friend Monty into a rabbit and Margot is now looking at the camera when in the real picture she was looking down at the book on the table.
I have tried to use a similar A.I. facility. This slightly dated but happy photo of Shirley and the little ones...
became this monstrosity...
AI can be sweet or freaky.
ReplyDeleteThat's a lovely photo of Shirley and your grandchildren. The AI version, just creepy and scary.
ReplyDeleteThe AI version of Shirley and the kids makes them seem scared while the real life picture is pure joy!
ReplyDeleteWell those three look as if they are having a great time. They are probably entertaining Grandpa.
ReplyDeleteWell we all do, if we could afford it. If he weren't expensive and dead I could imagine a version of the second image by Lucian Freud with you in the background in your underwear (children present after all, though that wouldn't have worried Freud), maybe reading a book or tapping away at your blog. Oh no! An AI challenge?
ReplyDeleteI agree it's a monstrosity, the real photo is much much better, the top one isn't so bad, I recognised all the cousins right away.
ReplyDeleteI think the 2nd AI photo has turned the cushion into the baby's long hair.
ReplyDeleteOh how I laughed. Shirley does look younger though, just very stressed, and the children look quite unhappy. Perhaps Bluey has been in accident.
ReplyDeleteI use AI quite often but it takes several attempts to get any decent results. Sometimes I just give up.
ReplyDeleteThis example clearly shows the limits of AI.
ReplyDeleteIn the first picture (the one Sarah had created), your three grandchildren are easy to recognise, although we know Margot‘s expression from other pictures as happy and serene, not with the slight frown her anime character shows.
In the second creation, it is obvious how the generative AI was unable to recognise the orange cushion for what it is, but took it for hair. That it could not get the smiles is a major fault.
Reading the various comments, I wonder why anyone actually bothers with AI - they won't be leaving lasting memories of how people really are. Perhaps it works better on other things?
ReplyDeleteThe first AI image is delightful, even if some of the features have been altered. The second one is a travesty, making everyone look unhappy, even fearful. What a shame.
ReplyDeleteNope. No AI pictures for me.
ReplyDeleteAn interesting question, that the Courts are starting to hear, "can anyone copyright an AI generated image? It will take time to get a clear answer.
ReplyDeleteTop one pretty good; bottom one not so good.
ReplyDeleteI like the real images much more. AI will just ruin what's real and what is not.
ReplyDeleteThe technology is so clever that we have at our fingertips.
ReplyDeleteI like the first image; and agree that the second is bad. I've so far only tried to get AI to create images by verbal descriptions, but then too the results vary a lot. I do find that AI can sometimes be useful to create illustrations when one does not have a photo, or does not want to use a real one. But I also discovered recently that one can't just ask it for a caricature of a named famous person... (That request got blocked.)
ReplyDeleteHa! That IS a monstrosity. I wonder why the AI thought everyone looked unhappy? I've tried to make some Japanese anime photos as well, with limited success.
ReplyDeletethis is a broad generalisation, but they reckon each AI image you generate, uses as much energy as fully chariging a smartphone.... just something to think about, innit?
ReplyDelete