DeepDream: The art of neural networks

Gray Area Foundation for the Arts and Research at Google invite you to join us for a benefit auction and art exhibition of: DeepDream: The art of neural networks a special gallery show of artworks made using artificial neural networks. The entire limited edition collection of works will be auctioned during this special evening.

All of the proceeds from the auction and exhibition will be donated to Gray Area Foundation for the Arts

Program Schedule
  • February 26th, 2016
    6:00 - 7:00pm, VIP Reception
  • VIP tickets include a reception with open bar, hors d'oeuvres, exhibit tour with Artists and premium seating for the live auction stage program. Silent auction bidding begins.
  • 7:00pm, Main Event
  • After the VIP reception, doors open up for the main event, please arrive on time to ensure entry before 7:30PM.
  • 7:30pm, Live Auction and Stage Programming
  • We are proud to present stage programming with Master of Ceremonies Brady Forrest plus keynotes by Blaise Agϋera y Arcas and Josette Melchor. We are excited to offer the best artworks from DeepDream during a live auction led by Emily Quinn. Also enjoy music by David Last and continue to bid on the artworks as the silent auction will remain open for one hour after the live auction.
  • Saturday, February 27th, 2016
    10AM-6PM, Art and Machine Learning Symposium
  • On Saturday, we invite you for a series of talks and workshops on machine learning techniques for art creation. The exhibition is also open and free to the public. Register here.

Your attendance at this event will help support Gray Area's work in bridging the worlds of art and technology, just as the DeepDream project has. 

Gray Area Foundation for the Arts is a 501(c)3 Nonprofit Organization reviving the Mission District's Grand Theater as a Media Arts Center for a diverse community. Your donations and purchases at this event are tax-deductible minus the value of goods received.

Exhibition Overview

The training set strongly influences the imagery produced during DeepDreaming. Here a network trained on places was used (from MIT Csail Lab). Even at intermediate layers (as used in this series) basic components of buildings and landscapes emerge.

The art works in this exhibition are made using artificial neural networks (NN). NNs are a biologically inspired form of computing which, unlike classical computer algorithms, aren’t programmed directly by human operators but instead learn from large amount of example data. The networks used in this work are trained with natural images from the environment and learn to distinguish objects and parse them into high level features. Their usual application is that of image classification or object recognition.

Once trained, these networks can then also be used to generate new imagery, essentially “imagining” images based on the learned rules and associations.

DeepDream is one technique for generating new images using a trained artificial neural network. The process works as follows. A initial picture or simply random noise is shown to the network which will then visually parse and interpret it. We then calculate how to change the picture incrementally in order to enhance that initial interpretation. The process is repeated until an image appears.

The complexity of the interpretation that the neural network obtains varies where in the network this interpretation is intercepted. Neurons that are closer to the input image will respond to simple features while neurons deep in the network will respond to more complex ones. Thus depending on the depth of the neural network layer that’s targeted, different forms and features are obtained often leading to interesting new recombinations of knowledge elements the network has learned. Many different forms are possible from simple geometric patterns to psychedelic looking combinations of entire concepts.


The Instruments series was created using specific class visualization. The GoogLeNet network has in its training set a number of musical instruments. Here the network’s internal representation of Violins, Cellos and Saxophones were visualized, followed by a style transfer step to adjust the colors and texture of the image.

Mike Tyka

Mike Tyka studied Biochemistry and Biotechnology at the University of Bristol and went on to work as a research fellow at the University of Washington, studying the structure and dynamics of protein molecules. In 2009, Mike and a team of artists created Groovik’s Cube, a 35 feet tall, functional, multi-player Rubik’s cube. Since then, he co-founded ATLSpace, an artist studio in Seattle and has been creating metal and glass sculptures of protein molecules. In 2013 Mike went to Google to study neural networks, both artificial and natural. This work naturally spilled over to his artistic interests, exploring the possibilities of artificial neural networks for creating art.

Memo Akten

Memo Akten is an artist from Istanbul, based in London, UK. His work explores the harmonies and tensions between nature, science, technology, culture and tradition. Studying and working with complex systems, behaviour and algorithms, he creates data dramatizations of natural and anthropogenic processes, and combines conceptual work with investigations into form, movement and sound. Akten’s work spans many disciplines including video, sound, light, dance, software, online works, installations and performances. Alongside his practice, he is currently working towards a PhD at Goldsmiths University of London in artificial intelligence and artistic, expressive human-machine interaction. Fascinated by trying to understand the world and human nature, he investigates technology’s role in learning more about ourselves and how we connect with each other. He draws inspiration from fields such as quantum mechanics, astrophysics, molecular & evolutionary biology, abiogenesis, neuroscience, anthropology and philosophy.

Mario Klingemann

Mario Klingemann is a code artist and a skeptic with a curious mind. His interests are manifold and in constant evolution, involving glitch art, data visualization or robotic installations. If there is one common denominator it's his desire to understand, question and subvert the inner workings of systems of any kind. Mario Klingemann is participating in the Machine Learning Residency program at the Google Cultural Institute in Paris.

JT Nimoy

I am a silicon valley unicorn. I make software art. I communicate between engineers and artists. I think a lot about issues of programming for non-programmers. I am best known for my generative computational aesthetics for Disney's TRON:Legacy. My most contagious meme is BallDroppings. My most visible work is commercial. My art shows in serious galleries and museums. I believe creativity is a serious force of business and humanity, capable of much more than mere entertainment, consumer manipulation, and intellectual invigoration. It is nutritious, healing, magical, and an absolutely essential component in every single one of us.

James "Pouff" Roberts

James is an artist and musician based in Austin, Texas.

Samim Winiger

Samim is a compassionate alien & thinker. Game, Web, Music & Machine Learning Researcher. A Code magician and Narrative engineer. He uses tactical storytelling to deliver transformative narratives across mediums and technologies.

Jessica Brillhart

Jessica is the principal filmmaker for VR at Google. In 2009, she joined Google's Creative Lab where she spearheaded numerous award winning shorts and documentaries before joining the Cardboard team in 2015. Since directing, WORLD TOUR – the first VR film made with the Jump ecosystem – Brillhart has continued traveling the world, filming and experimenting, all in an effort to better understand and help inform others about this emergent medium.

Doug Fritz

Doug is an Artist & Engineer at Google. He tries to govern his life by three elements; the longing for love, the quest for knowledge, and empathy for the failure of the world. He feels this is something true for almost everyone he admires. For him the search for knowledge has always been encapsulated in the essence of a six year old with the power of recursion, able to ask why recursively ad infinitum, and he finds that absolutely beautiful.

Alexander Mordvintsev

Alexander Mordvintsev, Software Engineer at Google Zurich, creator of the DeepDream.

Michael Ishigaki

Michael Ishigaki is a virtual reality designer at Google. Michael resides in Piedmont, CA with his wife Erica. When he's not designing, he enjoys home renovation projects and cooking. Prior to Google, he worked at Hattery in San Francisco and studied Mathematics at the University of California, Berkeley.

Ross Goodwin

Ross Goodwin is a creative technologist, artist, language hacker, data scientist, current graduate student at NYU ITP, and former Obama administration ghostwriter. His work employs machine learning, natural language processing, and other computational tools to realize new forms and interfaces for textual expression.