Sep 9 2015
Jeff Clune studies how to harness the power of evolution to make artificially intelligent robots.
Jeff Clune, left, a UW assistant professor of computer science, recently received a National Science Foundation CAREER Award. He explains concepts related to harnessing evolution to produce artificially intelligent robots with doctoral students Joost Huizinga, center, and Henok Mengistu, right. (UW Photo)
To do so, he and his students recreate the process of evolution in computers, wherein simulated robots compete to make it into the next generation, and only the smartest survive. The brains of these simulated robots can then be downloaded to real robots, where they will eventually help society in innumerable ways, from putting out forest fires and finding earthquake survivors to cleaning our homes.
The research program headed by Clune, an assistant professor in the University of Wyoming’s Department of Computer Science, just received a massive boost in the form of a Faculty Early Career Development (CAREER) Program Award from the National Science Foundation (NSF). The $507,465 grant not only will pay for the students and equipment necessary to conduct the research, but it also significantly validates the value of Clune’s research. The NSF’s most prestigious awards for new faculty, CAREER grants are given to those who exemplify the role of teacher-scholars through outstanding research and excellent education.
Evolving more complex computational brains
Animals seen in the natural world are amazingly capable. They soar in the skies, swim in the seas and sprint over rough terrain. Their brains are arguably the most complex, impressive designs we know of. How did evolution produce them? Can we recreate that process in a computer to evolve capable, intelligent robots? Those are the questions that wake up Clune each morning.
“One thing we know,” Clune says, “is that the brains of animals are incredibly structurally organized. For example, they contain regular patterns of neural wiring and a hierarchical organization of separate, modular processing units.”
Clune and his colleagues recently have shed light on the biological mystery of how such structural organization evolved in animal brains and other networks in nature, such as protein, metabolic and genetic regulatory networks. They also have shown how to exploit that new knowledge to create computational simulations of evolution that produce those same properties in the brains of robots. His CAREER award funds the next phase of research, which is to investigate how these properties improve the intelligence and capabilities of robot brains.
Specifically, Clune and his students will test whether regularity, modularity and hierarchy, separately and in combination, improve the ability of robots to (1) learn to adapt to new situations, (2) continue to function even when damaged, (3) deal with unreliable data from their sensors, and (4) generally become smarter, measured as the ability to solve challenges of varying complexities.
“We think that encouraging the evolution of structurally organized computational brains could help improve the ability of robots to do all of these desirable things, but we have to conduct the experiments to find out,” Clune says.
Clune’s research has been featured in the world’s top scientific journals, such as an article he co-wrote on robots adapting to damage, which was recently featured on the cover of the journal Nature. Videos summarizing the research from his lab can be viewed at www.evolvingai.org.
Teaching the next generation
NSF CAREER awards must contain a substantial educational and outreach component. Clune’s grant will enable him and his students to continue to run and further improve the Laramie Robotics Club. Free to attend, the club teaches children a love for programming, science, technology, engineering and math (STEM).
Each week, students in the club get together and play with robots (see pictures and videos at www.LaramieRoboticsClub.org). They get to design their own challenges, such as having robots solve mazes or play robot tag. Because they want their robots to do something fun, they learn the programming and math required to accomplish their goals. Thus, they get to learn how exciting programming and robotics can be. Within the first year, one former member is now an engineering student at UW, and a former mentor is at the Massachusetts Institute of Technology on a full-ride scholarship.
“We wanted to create the learning opportunities for Laramie children that we wish we had when we were kids,” Clune says. “Growing up, my Ph.D. students and I had a thirst for technical knowledge, but no one around us knew how to provide the information we craved. We want to change that for the next generation.”
Clune’s educational goals extend beyond Laramie. His grant helps fund his plan to create a turn-key guide to “How to create a robotics club” that will be free for any community in Wyoming or beyond.
“We’re figuring out how to make this work the hard way,” Clune says, “and we want to spare everyone else that effort. We want other towns to be able to hit the ground running with a list of robots to buy, a set of lesson plans and activities that kids enjoy working through, and glitch-free software that they can easily install and learn to use.”
Clune’s Ph.D. student Roby Velez has volunteered long hours to help get the club off the ground, and this grant will provide funding for Velez and other graduate students to be able to dedicate more time to refining the club’s learning materials.
Computer programming and robotics represent a huge segment of the current and future economy. As the Seattle Times reports, there is a well-documented shortage of graduates in computer science nationwide. The Bureau of Labor Statistics projects that 70 percent of all newly created jobs across all STEM fields during this decade -- across engineering, the physical sciences, the life sciences and the social sciences -- will be in computer science.
“Any efforts to teach Wyoming children a love for programming, science, technology, engineering and math will ready them for the world they will grow up and work in, and it is rewarding to help them be able to be the leaders and innovators in a fast-paced, technology-driven economic and scientific world,” Clune says.
Clune has a bachelor’s degree in philosophy from the University of Michigan, a master’s degree in philosophy and a Ph.D. in computer science from Michigan State University, and was a postdoctoral fellow at Cornell University funded by an NSF Postdoctoral Fellowship. His research is frequently covered by the international media, including National Public Radio, NBC, Discover, the BBC, the Economist, Wired, National Geographic and The Atlantic.