Project CIMON launches as the first AI assistant in space

July 2, 2018
 

On Monday, July 2, a SpaceX cargo mission will reach the International Space Station (ISS) more than 200 miles above Earth. Among the supplies arriving at the ISS is CIMON, a medicine ball–size colleague designed to work with Alexander Gerst, the European Space Agency astronaut commanding the Horizons mission.

CIMON, short for Crew Interactive Mobile Companion, is the first AI assistant in space. Built as a partnership between IBM, Airbus, and the German Aerospace Agency (DLR), this free-flying, zero-g-ready AI will be used in various scientific experiments and, in rare moments of downtime, can engage in friendly conversation with Gerst. CIMON joins a long list of space achievements for IBM, which has helped NASA and other organizations from the early space missions to the present.

Space is a uniquely challenging test for an AI-based assistant. The ISS is a highly complex environment—compact, crowded, noisy, and without gravity. Astronauts’ work is meticulous, sophisticated, and dependent on accuracy and speed. And the small crew works in isolation from the rest of humanity, so an AI assistant’s conversational ability and even personality may be factors in a missions’ success.

CIMON also represents IBM focus on applying AI technology to help people accomplish more in their work – from astronauts to teachers to customer service representatives and much more. Voice assistants like CIMON are designed to help people complete complex kinds of tasks. With Watson technology, CIMON is able to handle time-consuming tasks more efficiently, allowing Commander Gerst to be more productive with the scientific work he is carrying out. And in rare moments of downtime, CIMON is able to engage in conversation to relieve astronauts’ isolation.

 “Social interaction between people and machines, and between astronauts and a digital assistant equipped with emotional intelligence, could play an important role in the success of long-term space missions,” explained Till Eisenberg, CIMON project leader at Airbus.

On Earth, assistants of this kind have the potential to work in environments such as hospitals and nursing care.

Project CIMON’s mission is the start of an exciting future in which we can explore the boundaries of artificial intelligence deployment in space.

“This experiment should show how far it is possible to support the astronauts’ work,” says Dr. Christian Karrasch, CIMON project leader at the DLR. “Ideally, astronauts will be able to use their time better and more effectively. We are entering uncharted waters with Project CIMON and are working at the threshold of technological feasibility.”

CIMON’s AI has been trained to support all the specifics of three scientific experiments, each of which can involve 100 pages of steps. It has learned the setups and processes for these experiments down to the last detail; it knows, for example, exactly where and how a camera must be set up, or which handle is due to be activated next. It will follow along with the experiment Gerst is conducting, calling out next steps, showing videos of how to execute those steps, and suggesting solutions. The goal for CIMON is to increase the efficiency and effectiveness of the experiments.

CIMON was developed not by a space-focused team, but by a small group of IBMers who were interested in exploring the capabilities of Watson. Matthias Biniok, Sophie Richter-Mendau, and other non-specialists were able to create a working prototype in a few weeks using Watson services on the IBM Cloud.

CIMON taps into Watson services including Visual Recognition, Speech to Text, and Text to Speech as well as Watson Assistant, a versatile set of capabilities that can be deeply trained to respond to a wide variety of questions and commands. This service gives CIMON text, speech and image processing capabilities, as well as the ability to retrieve specific information and findings. These skills are developed based on the principle of understanding – reasoning – learning. All of the Watson services are accessible to developers on the IBM Cloud, and are being used by thousands of enterprises every day.