Artificial Intelligence and Art

          It’s always really interesting to see artificial intelligence being used to make art (or at least artistically-inspired creations of some sort) because I feel like we often view art as a very human thing that a computer would never be able to fully understand.  There have been a few examples of this that I’ve seen, but the most recent one that I found in this article on engadget is an app called “WordsEye”.  The idea behind the app is really interesting; basically what it does is use natural language processing to analyze a written description you give it, and then it turns that description into a visual scene of some sort.  Now, as stated in the article, the app is limited by the models it has and the words it understands (you have to write in a very specific and precise way for it to understand you correctly), but some of the results are pretty interesting.  Theoretically, I suppose it could be used to bring ideas out of your head and put them into some sort of visual context, but since I don’t think most people think so specifically about what art they want to create, it’s probably just something cool to play around with (though, unfortunately, to do so you have to sign up for beta access as of right now).  As a side note, I think it would be really cool to take excerpts from books with really descriptive writing on scenery or people and see what this thing spits out. 

The real interesting thing here, however, is the idea of people creating “artificial intelligence” based around artistic or creative output.  This is hardly a new thing, but two other really great examples are Google’s recently released “Deep Dream” software and a program called “The Painting Fool” which was created by computer scientist Simon Colton all the way back in 2001.  Deep Dream was created using the same image-recognition AI that Google used for its reverse-image search and Photos app.  To describe it in the most simple was possible, it basically takes an inputted image/video and, using its trained recognition, it then reconstructs that image/video based on what it “sees”.  The results of this can be really amazing and show just how trippy and psychedelic a neural network’s vision of the world can look (although, to be fair, Deep Dream’s knowledge base is a bit limited).  The Painting Fool, on the other hand, was made to “paint” things based on its current “mood”.  As discussed in this Smithsonian article, it was set up for an exhibit where the program would decide its current mood based on a random newspaper article that it read, and then it would paint a portrait of the person sitting in front of the laptop it was running on in the style of that mood.  One thing that I thought was really cool was that the software would apparently sometimes be put in too “bad” of a mood by the article it read to even want to paint, saying things like, “I was in a particularly negative mood, because I was reading an article entitled: ‘Aftershocks rock Italy earthquake zone‘ in the world section of the Guardian newspaper, which was really sad, because it spoke of ‘terrified residents.’ So, I decided not to paint a portrait”.  When it was able to create paintings, however, it produced some pretty impressive results (as shown in the article).

While the technology has a long way to go, and there’s also probably the debate of whether or not the AI is responsible for the “art” created or the human that created it is, the current artistically-inclined artificial intelligences are pretty impressive and produce some pretty fascinating images.  I’m interested in seeing how this sort of technology will develop, and even more so, if it can ever go on to produce something beyond what people have been able to create thus far.  Who knows, maybe a truly sentient artificial intelligence of some sort would process things so differently than humans that it would be able to produce some really amazing, incomprehensible stuff.  Of course, this is just wild speculation and fantasizing, but it’s always fun to think about all of the crazy possibilities that artificial intelligence could unlock.