27 February 2018

Art, Imagination, and Creativity: Natural versus Artificial

Today, computational creativity explores the intersection between artificial intelligence, psychology, and natural language processing, largely based on neural network algorithms or machine learning. These branches of AI have benefited from a million-fold increase in computing power over the last two decades, a rate of change which is unlikely to stop into the future.

As early as 2012, IBM started a project with its Watson system to explore computational creativity. Since then the Watson team applied computational creativity to various domains. They have looked at how it can be used to develop new scents in the fragrance industry, create personalized itineraries for travel, and improve sports teams based on skills or strengths. In 2014, a collaboration with the Institute of Culinary Education lead to the successful debut of Chef Watson, at the annual South by Southwest in Austin, Texas.1 According to Lav Varshney, one of the system’s designers, instead of replicating earlier styles the goal here was not to solve a Turing test for cooking, but rather to invent new kinds of recipes.2

Released to the public in 2015, Google’s DeepDream was a computer vision program which used a convolutional neural network to find and enhance patterns in images via algorithmic pareidolia. (Pareidolia may be defined as the human ability to see shapes or make pictures out of randomness.) The following year MIT's Nightmare Machine appeared, consisting of frightening imagery generated by a computer using deep learning algorithms.3

In the 1990s, David Cope, a composer at UC Santa Cruz, created a program called Emily Howell, with which he can chat and share musical ideas. He describes it as “a conversationalist composer friend… a true assistant.”4 She scores music, he tells her what he liked or didn’t like, and together they compose. Fast forward to 2015, when Kelland Thomas, a jazz musician and associate professor at the University of Arizona School of Music, was granted funding to build a similar system capable of musically improvising with a human player, called MUSICA (for Music Improvising Collaborative Agent), under a Defense Advanced Research Projects Agency program called Communicating with Computers.5 According to Thomas, “We're trying to understand human creativity and simulate that." There are algorithms that can write hip-hop lyrics, for instance DeepBeat developed by Eric Malmi at Aalto University in Finland in 2015.6 In 2016, Margareta Ackerman’s ALYSIA (Automated LYrical SongwrIting Application) came to the attention of the popular press. Also using a computer as a collaborator, Ackerman came up with a system that could help write melodies.7 Also in 2016, Sony CSL Flow Machines showcased perhaps the first song to be composed by artificial intelligence, a pop song titled Daddy's Car.8 In 2016, even a computer generated musical appeared on the London stage, called “Beyond the Fence”.6

Robotic painting is an emerging genre of art, to the extent that Wikipedia now includes an entry on Robotic art. In 2013, there was an exhibition in Paris, called You Can't Know my Mind, which featured an artificial artist known as The Painting Fool offering free portraits on-demand, created by Simon Colton, a researcher at the pre-eminent Computational Creativity Research Group, Goldsmiths, University of London.9 Since 2016, RobotArt has been the sponsor of a $100,000 a year contest in robotic painting.10

​In 2011, the editors of one of the oldest student literary journals in the U.S. selected a poem called "For the Bristlecone Snag" for publication. However, it was written by a computer, but no one could tell. Zackary Scholl, then an undergraduate at Duke University, modified a program using a context-free grammar to auto-generate poems.6 The EU-funded What-if Machine project, 2013-2016, not only generated fictional storylines but also judged their potential usefulness and appeal.11 In early 2018, Microsoft unveiled a new technology called “drawing bot”, capable of creating images from text descriptions.12

Computational humor is another area of computational creativity. Dragomir Radev is interested in computational creativity, and trying to come up with systems that actually understand and generate funny text.13 Games By Angelina is the home of ANGELINA, the research of Michael Cook of the University of Falmouth, whose aim is developing an AI system that can design video games intelligently.14

Since 2017, Philippe Pasquier is teaching an online course in Generative Art and Computational Creativity, that introduces various algorithms from artificial intelligence, machine learning, and artificial life that are used for generative processes.15 Also in 2017, the World Science Festival in New York City featured a session on Computational Creativity: AI and the Art of Ingenuity, in which experts in psychology and neuroscience explored the roots of creativity in humans and computers, what artificial creativity reveals about human imagination, as well as the future of hybrid systems that build on the capabilities of both.16 Organized by the Association for Computational Creativity, the International Conference on Computational Creativity is the premier academic forum for researchers, which in turn has spawned Musical Metacreation workshop series. Metacreation refers to tools and techniques from artificial intelligence, artificial life, and machine learning, inspired by cognitive and natural science, for creative tasks.17

At the very root of “imagination” is not only the word “image” but also image itself. I believe most fundamentally the question is, how are images processed in the human mind, or brain, in such a way to lead to creativity? This then begs the questions, how are words converted into images, and how are images converted into words, in both humans and machines? And more generally, how are images processed in computers to lead to creativity?

  1. Stinson, E. “America's Next Top Chef Is a Supercomputer From IBM.” Wired (June 2015).
  2. Marcus, G. “Cooking with I.B.M.: The Synthetic Gastronomist.” The New Yorker (April 2013).
  3. Dormehl, L. “This AI generates fake Street View images in impressive high definition.” Digital Trends (August 2017).
  4. Hutson, M. "Our Bots, Ourselves." The Atlantic (March 2017).
  5. Misener, D. “New musical Beyond the Fence created almost entirely by computers.” CBC News (December 2015).
  6. Kane, K. “Algorithm and rhyme: Artificial intelligence takes on songwriting.” Palo Alto Weekly (April 2017).
  7. Needham, J. "We Are The Robots: Is the future of music artificial?" FACT Magazine (February 2017).
  8. Johns, S. “Artificial intelligence experts question if machines can ever be truly creative.” Imperial College London (January 2018).
  9. Arbesman, S. "Computational Creativity and the What-If Machine." Wired (January 2015).
  10. Perez, S. “Microsoft’s new drawing bot is an AI artist.” TechCrunch (January 2018).
  11. Weir, W. “Programming for laughs: A.I. tries its hand at humor at YSEAS.” YaleNews (December 2017).
  12. Parkin, S. “AI Is Dreaming Up New Kinds of Video Games.” MIT Technology Review (November 2017).
  13. Luckow, D. "SFU MOOC a new route for students." SFU News (January 2017).
  14. Rockmore & Casey. "Humans and Machines Making Beautiful Music Together." Slate (July 2017).