Skip if you are feeling overly saturated with the next next next new AI thing. Amongst all the noise I still see curious things which pique my curiosity.
Friday, was a site/tool called Img2Prompt which claims to take as input, an image generated by Stable Diffusions and suggest a prompt likely to have generated it. How? Who knows. But it offers then a button to then generate a new image from that prompt.
Will it deliver a successful round trip?
But I can’t just share links without trying, so I put it through it’s paces. I tried both and image generated at Stable Diffusion Online and also just seeing what kind of prompt it would create from a normal image (guess what, I used a photo of my dog).
See the full details and examples here
Cut to the chase, I was underimpressed. These efforts of just putting some prompt in a box and getting something back that is marginally related seems no quite the future being touted.
I was reading this article today, so I wanted to share and see what are your thoughts. Basically, artist claim that AI basically makes indiscriminate use of the works of artists that are available on the Internet, without taking into account the rights of use of these works. it does not say that Ai uses those artworks directly, necessarily, but applies the same visual language that was created with so much effort by artists.
I think it is an interesting discussion and movement happening in the arts field today.
I too am interested! And I would hope we can have more discussions here as open educators wrestle with these questions.
In general, without a real understanding of how these generators work, it’s hard to say how previous works are trained. Don’t all artists absorb the works of those that come before them? If I go to a museum and study all the paintings of Emily Carr, get inspired, and adopt her techniques in my own work, how different is this? For some, the problem is that the works are being done by a non human entity.
Our intuitions and our rules and our past experiences are being stretched, because we do not not exactly how new works are generated. They just pop out of a black box.
Okay, Creative Commons is on this- they ran two webinars in October
and they are seeking community input to form a strategy on Generative AI in some sessions next week
Please participate and share what you learn (I might not make it).