---

At the 1982 NCAA championship, an 18-year-old Michael Jordan sinks a 16-foot jump shot in the final seconds of the game — giving the Tar Heels a one-point win over Georgetown. It’s the play that kickstarts Jordan’s transformation from college ballplayer to global phenomenon.

In the 2023 movie “Air,” Matt Damon’s character replays that final sequence again and again, trying to persuade Nike’s marketing vice president, played by Jason Bateman, to make Jordan the spokesperson for a new line of basketball shoes.

“They’re down by one, there’s under half a minute to go … Why is the ball going to the 18-year-old skinny freshman from Wilmington, North Carolina?” Damon asks. “He knows he’s getting the ball. The play is drawn up for Jordan.”

Gedas Bertasius likes to show this scene in classes and talks. For him, it’s more than sports nostalgia — it’s a case study in how humans read video: how we anticipate, infer intent, and extract meaning from motion. As a computer scientist at UNC-Chapel Hill and a former college basketball player at Dartmouth, he’s building AI that can do the same.

“Even though existing tools can now organize and tag video clips, video analysis still requires significant time commitment from coaches,” he explains. “I’m trying to build the technology that can automate that even more. There’s no way one coach can review 30 hours of video footage between consecutive games. They don’t have time to do that.”

From that challenge, his research goals follow naturally: to create AI-driven tools to review game footage, predict athlete career trajectories, and deliver personalized coaching for learning any skill — from basketball to cooking. Basketball is the proving ground; human learning is the horizon.

“AI is not just about making machines better,” he says. “How can we develop technology to make people better?”

Practicing: where basketball begins

Growing up in Lithuania, “basketball is like our second religion,” Bertasius says. It was always in the background — a game on TV, his family talking plays, the sound of a ball echoing through a gym. At age 8, his parents enrolled him in the Sarunas Marciulionis Basketball Academy, an elite program led by professional coaches.

“In our first year, we didn’t even play basketball,” he says. “We did lots of different physical exercises, played games like dodgeball, and learned how to dribble, pass, and shoot.”

When he started competing, Bertasius played small forward — the team’s Swiss Army knife. They do it all: shooting, ball-handling, defending, rebounding. And despite the name, small forwards aren’t small. Bertasius is six-foot-five.

Success followed: a spot on the Lithuanian National Team, multiple international championships, and a European silver medal in 2008. Two years later, he crossed the Atlantic with a Division I scholarship to play at Dartmouth. The court had carried him far. It would also carry him somewhere unexpected.

Pivoting: from court to classroom

Like many college athletes, basketball sat at the center of Bertasius’ world. But another curiosity tugged at him: how people learn so quickly.

“I think that’s one of our most fascinating capabilities,” he says.

He arrived at college planning to study math and economics. Then a sophomore-year computer science course derailed his plan — in a good way. Programming felt like a new kind of playbook. He built a virtual betting recommendation system that pulled game statistics from the internet to predict outcomes. A professor took notice and joined him on the project.

The experience clarified a fork in the road. If basketball wasn’t a lifelong career, what would be? He switched majors and dove into machine learning — drawn to the idea of translating the speed and adaptability of human learning into code.

“Research and basketball share some general characteristics,” he says. “Both take persistence — you fail, you push through, and you repeat a skill hundreds of times before it finally clicks.”

Rebounding: bouncing into research

Bertasius pursued a PhD in computer science at the University of Pennsylvania, focusing on computer vision — teaching computers to interpret images and videos in ways that echo human vision and understanding.

He set out to build a personal assistant that could support users as they moved through the world. His early vision was a pair of wearable smart glasses that could offer verbal and visual guidance, reminding someone of the steps in a recipe or helping them learn new physical skills.

To gather the right data, he became a walking experiment: a helmet with a GoPro mounted on top, recording daily life.

“Everybody around me just looked at me like, What’s wrong with you?” he says with a laugh. “But I wanted to use computer vision to understand human behavior, things like actions, skills, intent. Video was a natural medium to capture that.”

The helmet drew stares, but the idea behind it was simple: Our lives are sequences of actions. If machines can read those sequences — the way a coach reads a game or anticipates Jordan’s winning shot — they can help us learn faster, train smarter, and maybe even see possibilities we’d otherwise miss.

After graduating in 2019, this project landed Bertasius a postdoctoral research position at Facebook, where he worked on general video understanding and built augmented reality tools for Oculus headsets.

“Facebook probably had one of the strongest computer vision teams anywhere in the world,” he says. “Seeing the problems they worked on and how they were tackling them had a significant impact on my work.”

Scoring: advancing AI research

Today, the same instincts that once helped Bertasius scan a court — where the pass is headed, how the defense might collapse — shape the questions he asks as a researcher. How do you teach a computer to notice the right details at the right time for the right learner? How do you turn hours of raw footage into meaningful insight?

It all begins with data. Bertasius and his lab partner with coaches to obtain game footage, pull content from YouTube, and record their own videos at basketball gyms, soccer fields, and more. They then label everything, teaching their model how to recognize what’s happening — using terms like “pick and roll” or “lateral shuffle” to describe the actions unfolding on the screen.

“Honestly, it’s not glamorous,” Bertasius says. “We’re just setting a lot of different values and seeing how the model learns under certain conditions.”

Processing these high-action videos in real time is no small task. Players often look similar, their movements are fast and layered, and bodies constantly block one another from view. The AI must be trained on exceptionally detailed descriptions, and the final system has to map each step in sequence to the larger goal, understanding which actions actually matter at any given moment.  

But the outcome could be a game-changer — in multiple ways.

“The goal is to build an AI technology that can perform complex video analysis similar to how humans do it, whether it’s a sports analyst or a medical researcher who needs to navigate multiple information sources to find answers to complex questions,” Bertasius says.

He wants to create a ChatGPT-like program that can offer real-time commentary on sports videos. Users might ask: Why did the player do that? What was the point of that strategy? The AI would answer instantly, explaining the play as it happens.

This technology could also help scouts and coaches predict an athlete’s career trajectory — insight that could shape draft decisions, training plans, or even how a team structures its long-term strategy. Machine learning models can scan hundreds of hours of footage in minutes, surfacing patterns a human reviewer could easily miss.

“We’re starting with sports, but I think this sort of decision-making capability goes way beyond that,” Bertasius says.

That broader vision is why he’s developing a virtual personal coach for anyone learning a new skill. Users wear camera-equipped glasses while performing a task — chopping onions, shooting free throws, playing guitar — and the AI analyzes their movements in real time, offering tailored feedback on how to improve.

Bertasius is also exploring how this technology could support classroom learning. He’s fed recorded lectures into these models and then asked them questions about the material, testing how well the AI can interpret, summarize, and reason through complex instruction.

In 2025, this technology won the Multi-Discipline Lecture Video Understanding Challenge at a global AI conference, outperforming models from OpenAI, Google, and Anthropic.

“We can use this technology to scale office hours by giving every student instant, personalized help,” Bertasius says.

When asked what ties these goals together, his answer is simple.

“I think it’s about empowering people,” he says. “Whether it’s an AI companion that helps you understand a basketball game or a virtual coach that helps you learn new skills, democratizing this expertise is the most impactful thing I could accomplish with my research.”