Photo: Lillian Griffith

A double-major in computer science and philosophy, Todd Griffith ’91 has a unique perspective on AI.

Discovery Machine

by Katie Williard

Todd W. Griffith ’91 and his wife, Anna, decided to start a tech company shortly after the birth of their first child. But it wasn’t the desire for flexibility or a novel parenting gadget that drove their decision. They were simply “too tired to resist,” he says.

Discovery Machine, based in Williamsport, Pa., grew from sleepless nights and a Small Business Innovation Research Grant from the Defense Advanced Research Projects Agency (DARPA). “Using the power of artificial intelligence (AI), we create interactive cognitive training agents,” Griffith says. “But we know that not all expertise can be captured in data.” Using a hybrid approach, Griffith’s company combines interrogable AI (which can be questioned and form a response) with in-person instruction.

Twenty-two years of business has resulted in a well-tested methodology that is now being used by the Air Force. “By interspersing the virtual instruction and simulations with supervised flying practice in T-6 training planes, they are able to reduce the time to the pilot’s first solo flight. What was taking nine months is now taking three,” says Griffith.

“We’ve gained a lot from technology, but we’ll never replace real human interaction.”

During development, Griffith’s team aimed to understand how expert pilots process information and emulate it with an AI model. Embracing John Boyd’s observe, orient, decide and act loop, Discovery Machine created an AI instructor — the Virtual Instructor Pilot Exercise Referee (VIPER). Now, $10 million simulators with in-person oversight have been replaced by virtual reality headsets and gaming chairs.

The hybrid method is the perfect fit, Griffith says. “We’re never going to replace the story about flying through the hills of Afghanistan and having to pull up at the last minute. That has to come from the pilots that did it. The software allows instructors to teach and then say, ‘Go practice this maneuver 100 times.’ Then VIPER gives the learner the feedback they need.”

The company evolved slowly through the two-decade tech boom that continues to expand. “We grew organically, and have seen the ups and downs,” says Griffith.

But his philosophy remains the same. “Human expertise is necessary for creating intelligent systems,” he says. “We’ve gained a lot from technology, but we’ll never replace real human interaction.”

Back to Full Issue

Table of Contents