Dhireesha Kudithipudi and Christopher Kanan are professors with different kinds of students. Their classrooms are labs, and the students are high-tech computers they are “teaching to think.”
Kudithipudi, a professor of computer engineering, and Kanan, an assistant professor of imaging science, are part of a growing group of RIT researchers working in a field broadly known as artificial intelligence, or AI. It’s essentially about building increasingly complex algorithms—the rules that govern their operating systems—so that the machines can perform tasks that normally require human intelligence, including making decisions.
Both professors and their colleagues in RIT’s Center for Human-Aware Intelligence believe their research, with multimillions in funding from the National Science Foundation, National Security Administration, Air Force Research Labs and other prominent organizations, could lead to breakthroughs in everything from health care to energy management to cybersecurity.
“There is an exuberance in this field,” said Kudithipudi. “And the work and research we are doing today is to improve human lives tomorrow.”
A new center for AI research
AI at RIT got its start a year ago when more than 200 faculty, students and staff participated in the Move 78 Retreat—a combination conference and strategy meeting. The retreat brought together for the first time individuals teaching and researching AI in classrooms and labs across campus. Move 78 leaders asked the group to focus on three areas:
The Move78 initiative flourished. Faculty from engineering, computing, business, science and liberal arts developed the Center for Human-Aware Artificial Intelligence to consolidate resources and foster a culture of AI research among faculty and students. There are now more than 40 separate AI-focused courses and numerous faculty-researchers in different tracts: Deep Learning for Vision, Machine Intelligence, Brain-inspired Computing and Big Data Analytics. Work being done at the center builds on national priorities for AI where the economic impact is projected to increase global GDP by nearly $16 trillion by 2030, according to a 2017 report by PricewaterhouseCoopers.
“It will come as no surprise to those familiar with RIT that we already have quite a robust portfolio regarding AI research and development. Our AI activities have been growing quite rapidly, and that is a trend I am sure will continue for the foreseeable future,” said Ryne Raffaelle, RIT’s vice president of research who helped to establish the new center. “AI is somewhat a natural for us as it brings together a number of our well-established strengths in computing and cyber technology, with things like visual perception, speech recognition, language translation, and decision-making and automation.”
The center has four focus areas: Brain-Inspired Computing (Kudithipudi’s speciality), Machine Learning and Perception (Kanan’s area of expertise), Automation and Human-Centered AI. Projects will address society’s challenges in the areas of manufacturing, cybersecurity, sustainable systems and energy development, and improved medical care and technology.
“There is a huge opportunity for researchers in this domain – but the tricky part is that researchers working in this area need to be truly interdisciplinary,” said Kudithipudi, director of the center. Kanan, who teaches in RIT’s Chester F. Carlson Center for Imaging Science, will serve as associate director.
That interdisciplinary work is happening:
Ray Ptucha and Andreas Savakis, assistant professor and professor, respectively, from the computer engineering department in the Kate Gleason College of Engineering, are training computer network systems to recognize images and information toward improving autonomous vehicle technology, and are advancing high tech sensors and imaging technology used in drones, medical devices, security and “smart” facilities, respectively.
Cecilia Ovesdotter Alm, associate professor and computational linguist in RIT’s College of Liberal Arts, is leading the Language Science @RIT initiative. One key area is its spoken language processing which can be used for applications such as speech recognition and speech synthesis. She and Reynold Bailey, associate professor of computer science in the B. Thomas Golisano College of Computing and Information Sciences, are working together on an NSF Research Experiences for Undergraduates program grant in computational sensing.
A variety of applications of artificial intelligence topics are investigated by professors across the Golisano College, including Reynold Bailey, Matt Huenerfauth, Richard Zanibbi, and Linwei Wang. In collaboration with researchers from RIT’s National Technical Institute for the Deaf, Huenerfauth’s lab is also investigating how automatic speech recognition technology could be used to produce captions automatically for one-on-one or small-group meetings between deaf and hearing participants.
Ferat Sahin and Eli Saber, both professors of electrical engineering, are developing real-time street and traffic sign detection using deep learning, and vehicle detection using LiDAR and RGB imagery, the latter refers to colors used for computer displays. Sahin and his Ph.D. students are working on better human-robot collaboration approaches for industrial robots through intelligent sensing, machine learning techniques and visual feedback.
“There is so much happening that even as a professor it has been insanely hard to keep up. I see this becoming much more interdisciplinary. It’s not just a computing thing anymore. New algorithms can recognize faces better than people can, and these same technologies can be applied to video gaming and robotics. This technology is really moving fast. It has led to a flood of people in the field, including students, and a huge demand in industry,” said Kanan, who is working with Kudithipudi on training neural networks to continually learn and adapt over time once new data and images are integrated.
Building a skilled AI workforce
The research at RIT is also focused on building a workforce skilled in AI.
Kudithipudi’s Neuromorphic AI Laboratory has nearly 20 undergraduate, graduate and doctoral students involved in research she is conducting through grants from DARPA, the National Science Foundation and Air Force Research Labs. Over the past decade, her teams have designed energy efficient, intelligent platforms specific to image perception, pattern recognition and predictive analysis.
“Our lab has a unique strength in this area because we not only design algorithms that are inspired by the brain, but we also build the hardware prototypes that emulate these processes inspired by the brain,” she said. “Brain-inspired computing, also referred to as neuromorphic computing, has been a very important sub-field within AI. Most of the AI algorithms today are designed with performance or accuracy as a target, but neuromorphic computing is part of a unique dimension, and we are trying to deploy these AI algorithms in the most energy efficient ways.”
Recently, Kudithpudi, whose students have gone on to receive doctoral degrees and find careers in companies such as Intel, NASA and Google, was honored with the Digital Rochester Technology Woman of the Year Award. The award was in recognition of her important research in AI and neuromorphic computing, but also for her approach to mentoring students, helping them explore and learn so that they become individuals who can make a difference.
One of those students was James Mnatzaganian, who worked in her lab in the Kate Gleason College of Engineering developing a mathematical framework for hierarchical temporal memory’s spatial pooler—essentially brain-inspired architectures for computing systems. His master’s thesis in this area received top honors at RIT and by the Northeastern Association of Graduate Schools.
“Dr. K encourages her students to think for themselves. She does not assign research topics, but instead requires that students seek out topics, within the relevant domain, to research,” said Mnatzaganian. “I joined the team with the intent of having a strict hardware focus. I left focusing not on hardware, but on algorithms with software implementations. This was a turning point for my career. I realized that I instead wanted to focus on software. She helped direct me to the field that I have since chosen to make my own.”
Today, Mnatzaganian is a senior software engineer with Carbon Black Inc., a big data cyber security firm, where he uses the data science skills he learned to improve the company’s security efficacy.
The center also wants to extend its reach beyond the university.
RIT is training technology managers, scientists and researchers about machine and deep learning during programs led by faculty-researchers from the center. Participants will participate in hands-on exercises to build AI models.
Work with national and international organizations will also expand through the center, and the possibility of new degree programs in each of the research areas is a possibility—and a necessity. With countries investing millions in developing different AI technology, it is important to keep up and surpass competitors. But to Kudithipudi, Kanan and the researchers in the center, new discoveries would be shaped not only by potential economic impact, but by taking into account ethical implications. They are asking the important questions of—what is AI being used for and what can it be used for?
“If you look at where AI stands today, we have a lot of algorithms that are very powerful in a narrow set of tasks,” said Kudithipudi. “People say, ‘If you give me a big enough computer, lots of data and a deep enough network, then you can have super intelligence at your hands.’ The reality is far from it, but there is so much potential. The benefits and future successes of AI will come from several aspects as multiple disciplines—and people—come together to develop new application domains. The success in this field has to be measured by how it can improve human lives, rather than any one, single metric.”
Note: Video available for this story