In 2019, I exhibited this limited edition portfolio at the invitation of the United Nations at their annual AI for Good Global Summit in Geneva, Switzerland, and also at the Laguna Beach Festival of Arts. Each photograph of a vintage toy represents an application of artificial intelligence. The acrylic pane protecting the photograph is laser-etched with an excerpt of actual machine learning computer code that relates to the toy. The etched letters cast a shadow of the code on the photograph behind them.
Prediction. Framed dimensions: 42 x 50 in, 105 x 126 cm. The Python code accompanying this image is excerpted from TensorRec, a framework to create custom recommendation algorithms for TensorFlow using deep neural networks. TensorRec was developed by a former senior engineer from Spotify. The Zoltan toy in the photograph was produced by FAO Schwarz and popularized in the movie Big. Three framed prints sold, two available, at $3000.
Robotics. Framed dimensions: 42 x 50 in, 105 x 126 cm. The Python code accompanying this image is excerpted from Anki’s Desktop Security Guard application for its Cozmo robot. The application learns the faces of your friends and trusted colleagues. Cozmo then scans its environment. When it detects a face it doesn’t recognize, it sets off an alarm. The wind-up toy in the photograph was manufactured in 1956 under the name Ratchet Robot by the Nomura toy company of Tokyo, Japan. One framed print sold, four available, at $3000.
Space Exploration. Framed dimensions: 42 x 50 in, 105 x 126 cm. The Python code accompanying this image was excerpted from Google’s Exoplanet ML software, a neural network built to discover exoplanets (i.e., planets beyond our solar system) in light curves. The Exoplanet project is a collaboration by NASA and Google’s AI research team, and relies on data from the Kepler Space Telescope. This software successfully uncovered Kepler-90i, a rocky planet 2,545 light years from Earth. The wind-up toy in the photograph was manufactured by Chucklesnort Robots and is a tribute to the tin space toys of the 1950s and 60s. Five framed prints available at $3000.
Missile Defense. Framed dimensions: 42 x 50 in, 105 x 126 cm. The code excerpt accompanying this image was submitted by a Moscow-based contestant in response to a challenge from the United Kingdom’s Defence Science and Technology Laboratory. The challenge was to create software to label objects in satellite images, quickly and accurately. An application like this can help one country track the location of other countries’ mobile missile launchers. This code relies on a convolutional neural network. The cast iron toy in the photograph is of unknown provenance. One framed print sold, four available, at $3000.
Cognitive Medicine. Framed dimensions: 29 x 50 in, 72 x 126 cm. The Python code accompanying this image is excerpted from Xvision, a chest X-ray analysis application for TensorFlow that relies on deep learning and a convolutional neural network. The 1940s tin friction toy in the photograph was manufactured by Lupor Metal Products of New York City. One framed print sold, four available, at $3000.
Self-Driving Vehicles. Framed dimensions: 29 x 50 in, 72 x 126 cm. The Python code accompanying this image is excerpted from openpilot, an open source driving agent. The software communicates with a vehicle’s on-board diagnostics adapter to control the accelerating, braking, and steering actuators in most new car models. openpilot is claimed to be equivalent in performance to Tesla’s Autopilot. The tin friction toy in the photograph, circa 1950, was manufactured by APLS of Japan. A previous owner replaced the toy’s original human driver with a robot. Five framed prints available at $3000.