[evolve2k]

Best activity I saw was say 10 front on pics of dogs faces. Print them out and for each image slice the image into vertical slices.

(The 10 original images are the training data)

Then you can make up dog faces by taking slices from each image. Then you can make sorta thinner dog faces (“beauty edits”) by just leaving out every second slice.

Nice practical analogy. Slice in one cow face cause the computer don’t know the difference (hallucinate), it’s mostly a thin dog but not fully some is just sliced in AI junk.

Also discuss how the final AI image is taken from the makers of the original images. Discuss ownership, it’s new but composite, etc.