Over the course of the semester, our trainees are reviewing webinars in their given fields and preparing abstracts to help colleagues outside their discipline make an informed choice about watching them. As our program bridges diverse disciplines, these abstracts are beneficial for our own group in helping one another gain key knowledge in each other’s fields. We are happy to share these here for anyone else who may find them helpful.
Personal AI-based robots as lifetime human companions
Shigeki Sugano D. Eng.; Sethu Vijayakumar Ph.D.; Tetsuya Ogata, D. Eng.
9 March, 2022
Science webinar series
View on the Science AAAS website (requires registration) >>
Analysis by Miles Norsworthy:
This webinar was presented by three different professors working within the field of robotics and artificial intelligence. Dr. Sugano’s portion focused on many of the physical requirements of robots interacting both with humans and independently. Dr. Ogata’s presentation explored the obstacles and advancement of deep learning by robots. Finally, Dr. Vijayakumar spoke of the degree of autonomy needed for robots as well as future bodily prosthetics and exosuits.
Dr. Sugano’s presentation expressed the need to go from “hard” to “soft” robotics for both safety and function. A hard robot is what we might find within a factory with its components made of steel, powered by pistons and gears, and filled up with various openings between actuating parts made perfectly for human fingers to be caught in. The soft robots proposed by Dr. Sugano are instead more continuous with a type of artificial skin and powered by the pumping of liquid into flexible portions of the robot to produce safer, and more flexible robots that more closely mimic the function of biological agents. With this soft but precise means of locomotion, Dr. Sugano aims to have robots performing highly precise work such as invasive surgery by 2050.
This 2050 goal set by Dr. Sugano is shared by Dr. Ogata, who seeks to produce robots with the ability to truly understand their environment and the task at hand. Using deep predictive learning, these soft robots need to identify mistakes made during a task and use that information to alter future task completion to improve performance. A crucial application of this learning is the ability of a future robot to make decisions when given a novel circumstance. This future robot must autonomously identify what needs to be done, muster the tools necessary to accomplish the task, and then use those tools to accomplish a task while the conditions that affect said task vary throughout the procedure. Needless to say, this is an incredibly difficult challenge for robotics and, if accomplished, will alter how work is done by 2050.
Finally, Dr. Vijayakumar asks the question of when and to what degree should robots have autonomy. When should a human instead make the decisions? Given the overall webinar’s theme of robots acting not just alone but alongside humans this is an important question. Dr. Vijayakumar also broadens this question beyond robots and into prosthetics and exoskeletons. If one is using an exoskeleton at what point does the exoskeleton make the decision to move in such a way as to avoid injuring its occupant and at what point does the occupant get to override the exoskeleton’s autonomy? One can imagine a future where unconscious patients are brought into the emergency department by their exoskeletons that detected a health emergency.
2050 is still many years away and much work needs to be done in robotics and artificial intelligence but the work of the three presenters shows us that an exciting new future not just of robots but of robots and humans working together for a safer and more efficient world.