I’m generally agnostic when it comes to technology. That includes what is currently being marketed as “artificial intelligence” or “AI.” Of course, whatever AI is, it’s really “intelligent” at all. Anyone who has attempted to use Microsoft Word’s “enhanced by AI” grammar and spelling checker already knows that. All AI does is patch together snippets from what it finds in its training data and then makes it seem like a human being did it. The latter is the hardest part and what makes it (sometimes) seem intelligent. But if the training data has errors, so will the “answer” AI produces. Hence, a grammar checker that is ungrammatical and can’t spell.
But I digress. I’d never use AI in writing fiction, or any other kind of prose for that matter. Word thrusts it on you, and its annoying and hackneyed–I write just fine without “intelligent” assistance. But, when it comes to drawing, I’m totally incompetent. I can’t even draw a recognizable stick figure. So, I occassionally will use AI together with Photoshop to do simple graphics for short stories i post on the web.
But I’ve recently discovered that you can use AI to retouch photos. For example, I have an old photo of myself taken when I was 37 (half my current age). I wondered what it would look like if Vermeer had been around to paint my portrait, so I asked Create.Microsoft.Com to render this photo into a painting “in the style of Vermeer.” The original and the AI product are below.
Pretty cool, huh? It sharpened the focus and took out the background, but it’s still recognizably me from that time in a galaxy long ago and far away.
The Oauchita Mountains in southeast Oklahoma are known for their scenic vistas. A few years back, my partner and I spent an October Saturday driving through them, and Gene took many snapshots. I wondered what it would look like if I took one of those snapshots and used the same AI engine to render it “in the Hudson Valley” style. The original and the rendering are below.
I liked that so much, I tried rendering it in the style of several different artists.
Okay, mabye the second row isn’t quite as cool as the first row or the Hudson Valley style, but they are still pretty cool.
Then it dawned me–I admit, I’m slow–that I could ask it to change the season! Here are three of those.
That covers all the seasons, since the original was taken in the late fall. All of that from one snapshot. Of course, Gene IS an artist, so all of these had with a really good photograph as the basic training data.
I tried–just to see what it would do–for a rendering in the style of Hockney. It said “something went wrong” and gave no rendering, to my great relief. Hockney is a living artist, and his works are all still protected by copyright. They have no place in any AI’s training data–unless, of course, the AI engine’s owners have his permission and have compensated him for the right to use his work in their commercial product.
Back to the main subject. I like all of these, although I have favorites. I’d love to know what your favorites are! Drop me a note in the comments!
Be First to Comment