Human emotions from the perspective of an artificial intelligence
Science fiction novels and films usually paint a very one-sided picture of how rationally calculating and cold an artificial intelligence works. In fact, however, AIs are already an important part of our daily lives and it is therefore essential to actively shape the future of and with AI. So how about a new, positive vision of artificial intelligence in which an emotional dialog is created?
In addition to Kelp Forest, Savanna and Living Music, “Emotional Landscapes” is another project for the lobby of Intel’s headquarters in Santa Clara, where digital installations take on a whole new dimension with the 16m² interactive “Brand Story Wall”. With advanced hardware and software, Intel is laying the foundations for AI systems worldwide and leading the discussion on the ethical responsibility to use artificial intelligence. The installation impressively demonstrates how AI can understand us humans on an emotional level, painting a positive picture of a shared future.
A camera scans the active viewer’s face for 5 pre-trained emotions – happy, surprised, angry, sad and neutral – and adjusts the visualization on the wall according to the result. A laughing face sprouts colorful flowers, while a rage-distorted expression fills the scene with sharp and dangerous shapes. The facial recognition AI also reacts to “mixed emotions” in real time and merges different colors, lights and organic shapes on the canvas. This is made possible by Intel’s Open Vino Toolkit, which uses specially trained deep-learning models to recognize and process emotions on dedicated hardware.
The user, in turn, sees both his own image and the AI’s interpretation at the center of the composition and can playfully experiment with his expressions. In this way, a dialogue, an exchange, and a mutual learning of emotions and scenic expression are created.
Behind the scenes
Tools NewTek NDI, OpenVino, Unity3D
PeriodAugust - December 2018; January - March 2020
ClientIntel
TeamFelix Herbst, Paul Kirsten, Thi Binh Minh Nguyen, Marvin Podsendek, Mathias Nell, Andreas Schrader, Stephan Kloß