A sphinx cat and a pug boxer at a boxing ring
Usually I post about how I instruct machines to translate my ideas into scientific discovery. This time around I’d like to discuss something a bit different: How machines can help turn ideas into art.
About a month ago I was given early research access to the DALL-E neural net by OpenAI. Fiddling with it for a few hours and skimming through different examples I was mind blown by it’s capabilities.
But just as you start to wonder if DALL-E marks the start of the end for human creativity, you start to notice something very subtle.
The truly “artistic” creations of DALL-E have to be given a solid artistic context. You’d need to cue it with an artistic style or medium in order for it to generate something coherent. Otherwise its creations seem rather confused and half-baked. The range of references you can feed it with is huge - way more than my cultural education. That in itself can really give someone pause and make them feel small.
In the end however, DALL-E, like any other AI is a referencing machine. It’s like a guy that knows all the jokes and opening lines from all sitcoms and rom-coms ever broadcasted. It can be really impressive at a party but that guy won’t invent the next “friends” or “Seinfeld”.
At its core, it still lacks true “creativity” to invent new styles or references. Unless something fundamentally changes in the way we build AI this will always be the case.
Having said all that DALL-E is super fun. Checkout my digital exhibition “A Sphinx cat and a Pug”, co-created by DALL-E and myself!