Dr. Brian Sadler
The Senior Research Scientist for Intelligent Systems at the Army Research Laboratory (ARL) in Adelphi, Md., Dr. Brian Sadler, featured speaker for the fall UAH Distinguished Lecture series.

The Senior Research Scientist for Intelligent Systems at the Army Research Laboratory (ARL) in Adelphi, Md., pronounced the greatest threat from Artificial Intelligence (AI) to be deception by adversarial forces rather than a ‘Skynet’-style takeover. Dr. Brian Sadler made the remarks as featured presenter for the fall Distinguished Lecture Series at The University of Alabama in Huntsville (UAH), a part of the University of Alabama System. The researcher’s talk highlighted his expertise and future vision for a number of cutting-edge fields, such as machine learning, signal processing and multi-agent autonomous systems.

Sadler is a Fellow of ARL and a Life Fellow of the Institute of Electrical and Electronics Engineers (IEEE), and has been an IEEE Distinguished Lecturer in Signal Processing, Communications and Artificial Intelligence and Machine Learning. He has been an Associate or Guest Editor for several international research journals including the IEEE Transactions on Signal Processing, IEEE Transactions on Robotics and the International Journal of Robotics Research. He received Best Paper Awards from the IEEE Signal Processing Society, as well as a number of ARL and Army research and development awards and an Outstanding Invention of the Year Award from the University of Maryland. He was the recipient of the Presidential Rank Award in 2021. His research expertise includes multi-agent intelligent systems, networking and communications and human-machine integration. He has over 500 publications in these fields with 20,000 citations.

“Following more than 100 years of analog, and then 50 years of digital, we have now entered a new era,” Sadler noted. “I don’t much like the term Artificial Intelligence. I think ‘machine learning’ is better. We entered the machine learning era around 2010, although there is no hard boundary, and there are many steps leading up to this.”

The pace of these advances is breathtaking, yet the power to wield the new technology could be sobering in the wrong hands.

“What’s the definition of intelligence?” Sadler pondered. “If it’s something a human can do, then what at first seems remarkable will eventually seem mundane. But if you are building something very powerful and you don’t quite know how it got the answer, you must be careful with the answer you get. How worried should we be? We’ve overcome a lot of challenges as a species. I’m not so much worried about AI taking over – I’m more worried about bad actors taking advantage of it.

“Machine learning and AI are adept at working with natural signals, such as speech, text and imagery, signals that we struggled with in the digital era,” Sadler noted. “When you can create signals that appear natural, then we have to worry about deception. This may be the number one security threat facing our way of life.”

Signal processing is an engineering and statistics subfield that focuses on analyzing, modifying and synthesizing a large variety of signals. These include man-made, such as communications and radar signals, and natural signals, such as sound, voices and visual images. A current challenge is to improve the theoretical prediction and classification of large datasets originating from natural signals. “Things that are natural are inherently non-linear,” the researcher explained. “Mathematically, we struggle with non-linear.”

Facing the biggest challenges from these technologies

“A new national defense challenge is open source,” Sadler said. “There is so much code out there now. That changes the adversarial problem. The cyber beast is huge. We have to learn to fight with cyber and misinformation. We do not take these things lightly.”

The field of multi-agent autonomy is similarly challenging. Take driverless vehicles, for example, that must deal with what the field calls ‘corner cases,’ which are any set of unusual circumstances the vehicle encounters that is beyond the design of its hardware and software, whether due to external factors or coding mistakes. These arise due to things like adverse weather, road construction and unexpected driver or pedestrian behavior.

“There is a risk-reward with autonomous systems,” Sadler said. “We have the technology right now to build an autonomous vehicle that can drive from New York to San Francisco without human intervention. That’s a significant reward. But how high a risk are you willing to take? You’re never going to encounter all the corner cases out there, even if you keep driving another million miles, and another, and another. And you can’t just tell a machine ‘obey the rules of the road.’ That’s too broad. That’s not how we drive. We collaborate. I need the person who is going to use the system to tell me what the policy is, and the policies have to be very specific and contextual. And how do we test these things? For the military we define the rules of engagement, and these can change rapidly. This means that an autonomous vehicle must account for the current rules. The commander needs a risk-reward knob that can be tuned on-the-fly in order to accommodate his or her current threat.”

Taking a peek over the horizon, the researcher depicted a future where these systems are ubiquitous, noting that “everything is convergent – technology converges. We are moving into a world that is all systems. Like the system in your pocket, a cell phone, which is extremely complex. We used to have this Einstein vision of what is brilliant, one person working alone in the basement. These were our heroes. Today, my idea of brilliance is cohesive interdisciplinary teams working together. Intelligence is inherently collaborative. Beyond the machine learning era, I would say the next step is collaborative intelligence. And the combination of cognitive science and machine learning will emerge.”

The Distinguished Lecture Series was conceived to enhance community collaboration with UAH, the agencies on Redstone Arsenal and industry partners in Cummings Research Park. The series aims to raise awareness and foster a better understanding of current events and future trends, and how these activities are positively influenced through the region’s government, corporate and academic partners.


Kristina Hendrix

Elizabeth Gibisch