There’s a lot of buzz right now around the emerging field of “generative artificial intelligence” – online, publicly available machine learning models that can create new works, either in text or visually in digital art. These advanced computing systems gather vast amounts of information from online sources and attempt to learn about it in similar ways to the way we think the human brain works.
There are other well known platforms like DALL-E by OpenAI but today I have been playing with a competitor called DreamStudio by Stable Diffusion AI. These platforms are free to use and you just have to login and you’ll get some free credits to play with – most of these systems use the concept of credits to stop people going mad and bombarding the system with thousands of requests – with the hype and fuss over generative AI these systems have been put under a lot of pressure. Then all you do is type a description of what you want it to create for you.
The next gallery was my first attempts at creating images “in the style of Toyohara Kunichika” of “ukiyo-e” or “kabuki actors” and it has made some interesting ukiyo-e styled images but they don’t in any way look like any of Kunichika’s work:
The next gallery of images was created by asking for “a landscape in the style of Hiroshige” (some “with people”) and I think it has done surprisingly well. Many familiar trademarks are there – the peaks of mountains, bokashi fades and the style of the trees.
And here is the same sort of attempt but in the style of Hokusai – you can see how it has tried to bring in elements of the famous Great Wave image:
AI sounds very interesting but we’re still in the early days, even when it sometimes looks as though the systems are “magical” and genuinely intelligent. You only have to use them for a while to understand that, while they can look intelligent, they often make very simple mistakes and can be very wide of the mark even when they are seemingly communicating with us coherently or convincingly.
In fact I think that the greatest danger isn’t that AI will take over the world – it’s that too many people will trust AI too much because it seems to be clever and innovative. We have already seen what happened with satellite navigation when humans trusted it too much and ended up driving into rivers – trusting the technology even when their own senses would have told them it was wrong!
So, the answer to my question is really “no, it can’t but it can create works inspired by ukiyo-e” and I think it will get better as the systems learn. As much as the technology holds great promise for creating new works, which might be suitable for use as graphics on websites etc, I think that it has a long way to go before it becomes really useful. Currently as the user you have to choose your words very carefully to create something you’d like and that doesn’t look freakishly weird. There are also issues to do with its output being a derivative artwork, based on someone else’s copyright, and what that means for intellectual property.