Artificial intelligence (AI) enables scientists and engineers to create autonomous technologies that can function on their own while adapting and responding to changing environments and scenarios.
A common application of autonomous computer systems is in transportation. Today's consumer vehicles use automated, assistive technology, but they are not fully autonomous. The Society of Automotive Engineers and the U.S. National Highway Traffic Safety Administration have developed a six-level classification system to describe levels of vehicle automation, ranging from zero (the system provides momentary driver assistance) to five (the system is fully automated with no human driver needed). Currently, only vehicles at levels zero, one, and two are available for purchase in the United States.
Self-driving cars might make the news most often, but they are not the only place AI is being explored for use in vehicles. With future advances in AI, autonomous vehicle systems could take over the sometimes dangerous and tedious jobs of flying planes, delivering cargo, and even operating ambulances—and, some researchers argue, they could do it more efficiently and safely than humans. Improvements in autonomous technology will also create new opportunities in medicine, scientific research, and robotic space exploration.
Humans in the Loop
Though they are designed to be independent, some of these future technologies may still be monitored by and collaborate with humans. Such systems are said to have "humans in the loop."
"The question of whether we want a technology to be truly autonomous is up for debate, because we still want it to have alignment to human values: our safety concerns, our requirements," says Anima Anandkumar, Caltech's Bren Professor of Computing and Mathematical Sciences. "In so many scenarios, achieving full autonomy is probably too hard or even impossible with current technologies. But having humans be able to close the gaps of AI, and vice versa, can work very well. I think our goal should be those hybrid systems that combine humans and AI together."
"There are going to be huge risks if we unleash this autonomy without having human protocols governing it," says Mory Gharib, the Booth-Kresa Leadership Chair and director of Caltech's Center for Autonomous Systems and Technologies.
In this video, Gharib and Anandkumar discuss the definition and promise of autonomous technologies.
Caltech's Center for Autonomous Systems and Technologies (CAST) has emerged as a hub for autonomy research. The 10,000-square-foot facility opened in 2017 with the goal of promoting interdisciplinary research on autonomous systems. Work conducted at CAST has led to the development of Neural Fly, Neural Lander, and Neural Swarm, test beds for deep-learning techniques that can help drones fly and land in windy conditions and coordinate the machines' movement in swarms.
Caltech is also home to the Robotic Assisted Mobility (RoAMS) Science Initiative, which seeks to develop personalized robotic-assistance devices, such as exoskeletons and prostheses, by using AI to connect the device to the human operator's nervous system.
John Dabiri, Caltech's Centennial Professor of Aeronautics and Mechanical Engineering, has developed autonomous ocean-going drones that swarm and use the oceans' currents for navigation. The drones are intended to monitor conditions in oceans either on earth or on other planets.
Research conducted at Caltech is making its way into the commercial realm as well. Gharib founded Toofon, a company that is developing an autonomous heavy-lift drone for firefighting and supply-chain applications.