Art, what is it?
We have been writing a lot of articles about AI regarding factual research and applications. Recently I’ve been thinking about one thing that we haven’t explored yet. Creativity and art. What is the potential AI in this field? First I suppose we should agree on the definition of what Art means, so let’s keep it simple and take the one from the Oxford dictionary :
It is the expression or application of creative skill and imagination, typically in a visual form such as painting, drawing, or sculpture, producing works to be appreciated primarily for their beauty or emotional power.
As far as AI is concerned we need to be concrete and practical – meaning that art can be expressed through literature, music also performing or visual arts. Some art forms combine a visual element with performance like a movie, or words such as comics. An artist is therefore a person engaged in one or more activities related to creating art.
Many people feel, that whereas computers can excel in well-structured areas of problem solving like logic or algebra they will have little hope of ever producing truly creative work. For a work to be creative, it must be innovative which represents a challenge for computers. This is where I disagree and think that artificial intelligence and machine learning actually do have huge potential to drive a new generation of creativity.
Last year Google launched Magenta, a research project to explore how artificial intelligence can create art, and make that process easier for TensorFlow users. The group has about six researchers now, and will invite other academics to help try to solve the problem of creative machines. You can access the project on GitHub.
Meanwhile some artists have already started using AI in their work. Sam Kronick is an artist and his plan is to sketch rocks, without using a pen and paper. He and his artistic partner Tara Shi are going to do a 3D scan of the rocks so an AI can map their contours, learn to recognize them and then start generating its own rocks. In this process they are also figuring out the limits of computational creativity. Kronick and Shi are using a neural network, lots of rock pictures from Google Images, hundreds of thousands of search terms. Then, basically, their AI thinks in layers, figuring what what a rock is and how to identify it: the texture, colors on its surface, before trying to draw one.
Today – Artificial intelligence can already paint, write books and make music
Artificial intelligence has a long way to go before it can match the greatest painters of all times in creativity, but it seems they’ve gotten quite good at mimicking and remixing what they see. An interesting study published by researchers from the University of Tubingen in Germany, researchers described an artificial intelligence neural network capable of lifting the “style” of an image and using it to restyle another image, which is why these waterfront houses look as though they were painted by Picasso, van Gogh, or Munch.
Many of us are concerned that robots will take our jobs once artificial intelligence reaches its full potential but it’s a safe bet that no one has put “writer” at the top of the robot job list. Think again ! An artificial intelligence program has already co-authored a short story got through the first round of a national literary prize, it seems that no occupation is safe. The story is actually called The Day A Computer Writes A Novel. It didn’t make it to first prize, but it did come close. Officially, the novel was written by a very human team that led the AI program’s development. Hitoshi Matsubara and his team selected words and sentences, and set parameters for construction before letting the AI “write” the novel autonomously.
AI is also writing articles thanks to Wordsmith – an artificial writer developed by the North Carolina-based company Automated Insights that picks elements from a dataset and uses them to structure a “human sounding” article. As well as being able to use more emotive language, it varies vocabulary and syntax to make its work more readable. Founder Robbie Allen describes the program as a “natural language generation platform” which can be used to churn out data-driven articles faster than even the most nimble-fingered hack. Wordsmith also knows how to draft reports on marketing, business activity and financial results.
Now that AI can write, can it make music? At the Durham Moogfest festival Adam Roberts from Google was onstage putting this questions to the test. First a human plays a short tune then the computer follows up with a longer variation of the tune. Is the computer-generated melody an original piece of music? Researchers say yes, but add that humans still remain the ultimate artists. Researchers say the whole question of whether it’s art or not is irrelevant, and such questions should be debated within the art community.
This YouTube video is probably the first pop song written by AI.
The song in question was created by researchers at Sony, who used the company’s Flow Machines software to analyze a database of some 13,000 lead sheets from different genres around the world. The software writes its own melodies, and a human composer was drafted to turn material into a fully produced track. He simply inputted a desired style of music (in this case The Beatles). The lingering question, though, is what is “Daddy’s Car” actually about? Although the lyrics were written by a human rather than the software, it’s impossible not to detect the menacing threats of machine intelligence in the wording…
Tomorrow – Next steps
Above all, as of today there’s still a human side to AI produced art in both its creation and presentation. Will computers make art or will they just replicate human creativity ? Who knows, but apart from painting, writing a book or making music there are other art disciplines where AI could prove itself.
3D printing & architecture
What if artificial intelligence could design and make 3D objects ? Using a 3D printer an AI it will be able somehow leverage the image recognition power of deep neural networks to create new artifacts without human input. Just picture it! Lines of furnitures created by your personal AI. It could be created by innovating on shapes in relation to the volume of a room, colors, or other furniture to find a unique arrangement.
Artificial intelligence will also help us in architecture by creating new buildings with new ecological and energy standards while innovating on the architectural aspect. The advantage of AI here is obviously its propensity to correlate much more data than a human being much faster to create something innovative and new. Mixing 3D printing and AI will open new ways of building houses of maybe even cities. As you can see in the video below we already have a 3D printer building a house in 24 hours…
In sports, AI has been there for decades. We have seen robots playing chess. The sports world today is linked with technology by combining natural, athletic talent with analytics and even AI to get the best possible outcomes. Everything from reviewing player performance, improving skills, to even predicting the best actions for future players is being recorder. But next steps might be applying AI to other more creative sports, artistically speaking. Let’s take figure skating, imagine an artificial intelligence capable of creating new choreographies and finding the best combination between music and dance while correlating this to historical data of previous scores given by the jury members.
And my final example of AI creativity is films of a new kind. Have you seen this trailer made by IBM Watson?
There is also this weird sci-fi script written by Benjamin an LSTM recurrent neural network. The AI was fed with a corpus of dozens of sci-fi screenplays —mostly movies from the 1980s and 90s. Benjamin wrote the screenplay, complete with almost impossible stage directions. See the result below…
A film is a complex blend of music, dialogues, camera shots, sets, actors as well as costumes and potentially special effects. Artificial intelligence will eventually be able to create new scripts and direct the whole film from start to finish. Who knows maybe in a few years the Oscar for best director will go to an AI – if they get it right this time…