I’ve been researching machine creativity and I ran into a few interesting articles that I wanted to share with everyone! If you don’t know, machine creativity is exactly what it sounds like it is: teaching computers how to create art in any form. It’s a very complicated field, because if you think about it, all computers are are mega-powerful calculators with lots of memory. How in the world could they write stories and poems and draw pictures?
The answer is through math!
I know that that may seem a little counterintuitive because math is literally one of the most concrete things out there and I think everyone would agree that art is the exact opposite. But there’s always an underlying logic behind all the art that we create; though our brains work in mysterious ways, every facet of a story/artwork we created is always the result of a decision tree—whether subconsciously or consciously, we always intend to make the decisions that we do when creating art. Decision trees are really good for computers—computers love decision trees since they’re essentially just big conditionals, and conditionals are only true/false, and that’s binary, so that makes computers very very happy
Anyway, I was researching some of the things that computers have created, and found about something called NaNoGenMo (National Novel Generating Month), which is like NaNoWriMo, but instead of you writing the novel, you write code to generate a novel. I think it’s a really cool idea and there were some very amusing examples in this Medium post. My favorite was “Twide and Twejudice”, which is Pride and Prejudice but with Tweets as the dialogue
But let’s be real, those novels really aren’t that comprehensible. I did a little more digging around and found something even cooler, a web app called word.camera, which takes in a picture and then outputs a poem based on what’s happening in the picture. (Just so you know, this is definitely not an easy thing to teach a computer to do!).
Machine creativity isn’t limited only to the written word, though; as I said before, machines have been to taught how to create all sorts of art. The coolest, in my opinion, is the visual art. For instance, Google DeepDream is a research project by Google which looks at any arbitrary image you give it and finds other images inside it. Here’s an example:
It’s kinda beautiful, right? And definitely a little surrealistic. Essentially, DeepDream reverses the process of recognizing objects in order to create objects that aren’t there.
Even cooler, Sony CSL is leading a project to develop AI that can create music—they’ve already generated a song called “Daddy’s Car”, which is...a little odd, but that’s understandable, considering it’s coming from a computer haha!
Anyway, this was very random, but something that I thought would be interesting to share In particular, this was interesting for me because I’m always trying to reconcile the two sides of my creativity, the science side and the writing side. I guess, in the end, the two sides aren’t all that different! In the end, whether I’m writing a fic or writing code, I am just stringing words together to create some sort of meaning.