If you saw yesterday’s post, about the updates I’ve made to my “Going to America” arrangement you might have spotted a rather grand painting in the background:

That took me all of five minutes to produce using, as you can see from the watermark, stablediffusionweb.com again. It uses the impressionism style and, unfortunately, it seems the site remembers the images I generate for a few days but not the prompts used to create them. This was something like “ocean liner sailing to usa with stars and stripes”.

One thing to note is, that although it was the blink of an eye compared to trying to create such an image by hand, I had to iterate through several progressions. My initial attempts began with the cinematic style and I mentioned showing a tuba player because that was the photograph I’d used with the first version of the music that I’d uploaded. I don’t think AI quite understands what a tuba is so I tried “concert band with stars and stripes” before deciding that I didn’t really want any musicians included. I switched to something about “sailing to america” (and changed to the impressionist style for a more painterly effect) and generated a couple with non-ocean going sailing boats before reaching the one you see above.

Yes, it is an “AI generated” image but it is important to remember that, although machine learning put the pixels together, the result you see still depends a lot on curation by natural intelligence. Credit – and blame – for AI creations still rests squarely on us humans.