Envisioning Robot Teammates
The fast pace of advancements in artificial intelligence is only adding to the common fear of robots taking over our world. Many researchers, including Dr. Julie Shah, are working to build robotic systems that augment our abilities and work safely alongside us – with the aim of robots being our teammates.
Dr. Shah is a professor of aeronautics and astronautics at MIT, and lead of the Interactive Robotics group in MIT’s Computer Science and Artificial Intelligence Laboratory. Her aim is to develop robots that can augment what humans can do, and adapt and respond to human behaviors such that people will accept them as teammates.
At the MIT Museum, you can interact with a duplicate of the AI-enabled robot in Dr. Shah’s lab. The robot’s name is IRGO, which is (taking some creative liberties on spelling) based on the initials of the lab name “Interactive Robotics Group,” and inspired by the Latin saying “Cogito ergo sum,” or “I think therefore I am.”
In the museum, there are 3 types of interactions you can have with IRGO:
…Physically move the robot through the steps of a task, so it can learn the motions.
…Set a dinner table in order to train the robot to also do the task by example and feedback.
…Make a (model) sandwich, where the robot can communicate with you to achieve the task together.
As you team up with the robot, you are contributing to MIT research. Dr. Shah’s team hopes to stress test the system with lots of interactions between the robot and museum visitors, in order to gather insights into how to improve the system. For example:
If a person finds it cumbersome to physically move the robot, researchers can work on adjusting the algorithm to make the robot more adaptive to the force from people as they are trying to teach it.
If the robot has difficulty setting the table even after three demonstrations by its human teammate, researchers could brainstorm ways to adjust its algorithm to acquire more information from the person via asking questions.
If researchers see that people want to be proactive in the sandwich demo and directly tell the robot to move to a given location, researchers may investigate new algorithms to allow the robot to listen to people, in addition to talking.
At the end of the day, Dr. Shah and her team envision this kind of robot teammate being implemented in settings ranging from hospitals to factories to homes. The algorithms used in these interactions in the museum can provide further insight into making that big goal happen: an effective robot teammate that is aware of human co-workers, knows the best sequence of tasks to avoid conflict, communicates when needed, and gets the job done—as a team.