It's been decades since artificial intelligence (AI) left science fiction to join science fact. AI and machine learning, a sub-branch of AI, are key technologies in the smart phones we carry, the software systems that protect us from fraud or predict the weather, in robots and drones. With every passing year the fields of AI and machine learning continue to evolve and new levels of complexity are introduced to more of the machines we use in everyday life. Dr. Melanie Mitchell, Professor of Computer Science at Portland State University, is one of a large community of researchers worldwide building upon theories of AI and machine learning and contributing to the cutting edge of innovation enabling machines to preform tasks similar to those humans do. Vision—in the sense of recognizing and relating objects—is one of these tasks. Dr. Mitchell, along with her colleagues and students, is developing tools to teach next generation machines to learn from digitized images.
When the charge-coupled device (CCD) was invented at Bell Labs in 1969 it enabled computers to capture and digitize images. Four decades later some computers and other machines can capture, process, and analyze images and assimilate that data to make the computational equivalent of inferring relationships between objects. Scientists like Dr. Mitchell are standing behind these incredible technological advancements.
"We're trying to get computers to learn and interpret visual scenes," Dr. Mitchell said. "Our work focuses on photographic imagery at the moment, although I have collaborators working on video—moving images."
The applications of this technology are vast and the potential benefits are clear. Autonomous vehicles could reduce automobile accidents. Retinal scans could replace security codes and login names. Medical and topographical imaging could help doctors, police, and firefighters. Computer vision could improve any number of imaging fields, from microscopy, to x-ray imaging, to MRIs.
Of course, teaching a computer to see is no easy task. Seeing takes place between the phenomenological world and millions of synaptic pathways where neurons fire at the speed of light, sending messages back and forth between different regions in our brains, connecting vision, memory, language, and other brain functions to create and learn from an experience. Achieving this order of complexity in computer vision may still be some time off, though the innovations that will make it possible are being developed and studied around the world in labs like those of Dr. Mitchell.
Dr. Mitchell's interest in computer vision, machine learning, and AI in general overlaps with one of her other fields of interest: complex systems. After receiving her Ph.D. from the University of Michigan, Dr. Mitchell took a position at the Santa Fe Institute where a group of scientists from disciplines across the academic spectrum had gone to push the boundaries of the science of complexity. In 2009, Oxford University Press published Mitchell's award-winning book on the subject: Complexity: A Guided Tour.
"I really think that it's important for scientists to share their work with the broader public," Dr. Mitchell said. "There is a responsibility to help people understand science and how scientists think."
A professor at PSU, Dr. Mitchell is also an External Professor at the Santa Fe Institute. Through the Santa Fe Institute and in partnership with PSU, Dr. Mitchell is teaching a massive open online course (MOOC), "Introduction to Complexity," that in its first quarter reached over 56,000 students from countries around the world. In order to help students, educators, and persons interested in learning about the subject of complexity, Dr. Mitchell recently received a grant from the nonprofit John Templeton Foundation through the Santa Fe Institute to launch the Complexity Explorer, an online learning tool and repository for research and information relating to the complex systems field.
"The mission of the Complexity Explorer," Dr. Mitchell said, "is to develop a website that has education materials related to complex systems for all levels of learners, ranging from high school to post-graduate education and professional development. It's meant to fill what we think is a real need for different materials for people who want to use these ideas in their classrooms, learn on their own, or develop online courses.
"In general, I'm really interested in innovative approaches to education, like flipped classrooms where students watch lectures at home and solve problems together in class, and like these MOOCs."
AI has come a long way in the past decade. Robots are playing soccer—don’t count on matches selling out any time soon, though; cars are driving on their own; face and speech recognition have come a long way, and devices like the upcoming Google Glass will change the way we interface with the Internet. Dr. Melanie Mitchell is one of the scientists working to bring these changes to the world while at the same time teaching the next generation of computer scientists and helping the rest of us understand just how exciting and engaging this innovative branch of science has become.
Authored by Shaun McGillis
Posted March 26, 2013