In an AI World, Can Sports Stay Human?

Billions of dollars spent, thousands of hours invested, livelihoods and long-term health risked over an epic clash—all hinging on a referee guesstimating the exact location of the ball at the moment its carrier crashes to the ground. Should the spot appear to be close to a first down, a precise measurement is performed.

Stan Honey, the inventor of the yellow first down line now ubiquitous across football broadcasts, has always found this funny.

“It is interesting how accurate the measurements of the chains can be and then how approximate the original spotting of the ball,” he says.

For most of football history, the solemn duty of measuring first downs was carried out by so-called chain gangs: people in bright vests lugging around giant orange sticks. But before this NFL season, the league installed six 8K Hawk-Eye cameras in each stadium to optically track the ball’s position, employing the same technology that has become familiar in other sports. In the blink of an eye, the Hawk-Eye system can judge precisely how far the ball is from the first down line. On Sunday, when the Patriots and Seahawks kick off, the NFL’s virtual measurement system will be used for the first time in a Super Bowl.

Honey is an expert on first down technology if there ever was one. When his first down line launched in 1998, he and his company, SportVision, not only changed how we watch football, but seemed to break down a wall between sports reality and our mediated experience of it.

“We began to hear from people saying, ‘I'm taking my kids to the game, and they're looking for the line on the field,’” says Bill Squadron, who co-founded SportVision with Honey and led its business side.

In the years since, precision tracking technology has become part of the fabric of sports worldwide. It’s difficult to imagine tennis now without Hawk-Eye cameras acting as line judges (the French Open is the only Grand Slam event where humans still do the job). In soccer, the technology polices the goal-line and offsides. Cricket’s Indian Premier League has so thoroughly integrated Hawk-Eye cameras that it has a designated TV umpire sit beside two Hawk-Eye operators during matches. And here at home, next season, Major League Baseball players will, for the first time, be able to challenge umpires’ ball and strike calls, appealing to robo-umps that employ similar technology.

As technology continues to advance, sports leagues are constantly weighing how much it should be incorporated into their games—how much they should let the real world of sports merge with the technological one. Just as we wrestle in broader society with how technology and AI will change our world—and what jobs they will replace—the same questions surface in sports. Will AI one day replace referees? Possibly even some coaches?

Technology has of course been changing football strategy for years—albeit in ways that are often opaque to fans. Every player wears tracking chips in their shoulder pads and, across the NFL, teams have started to embrace AI as a way to seek strategy advantages from the mountains of data they collect. Because of the competitive nature of the league, teams are secretive about their practices.

Honey believes that the NFL’s new measurement system is neither good nor bad, simply the next step in an evolution. He acknowledges that traditionalists might miss the drama of watching the chains get stretched out, with all the players crowded around, straining for a look (the new Hawk-Eye graphic simply doesn’t measure up). At the same time, Honey says, “Younger fans might just be pulling their hair out, you know, waiting for the chain gang to wander along.”

The question, to Honey, is how far we want to evolve and how fast. There is a trade-off between the imperfect humanity of the game and the undeniable benefits of efficiency and accuracy. As robo-umps enter MLB, for instance, gone will be the thrill of watching a pitcher’s strike zone expand as he gains steam. But maddening controversies over obviously botched calls—like this one in Game 3 of the most recent World Series—will also be a thing of the past.

It’s hard not to wonder, what’s next? And will we want it?

As the senior vice president of innovation and media at the United Football League, Scott Harniman’s job is to push technological advancements in a league that has essentially become a trial lab for the NFL. In March, the UFL launched a new division called FAST, or Football Advancement through Sports Technology. Under the program, the UFL partners with outside companies to incorporate their tech into the game. The UFL selected three partners in 2025: an AI-powered body camera company called Mindfly, a tracking outfit called SkillCorner, and Sportable, a company that puts sensors in balls.

Mindfly is an Israeli company that started out making body armor cameras. Whenever body-cams have been attached to athletes in the past, the picture always comes across jerky (and usually mildly nauseating). Mindfly came up with a design that places the camera right in the center of a player’s chest—drawing comparisons to Marvel’s Iron Man—and uses AI to present a stabilized, smoothed out picture.

“This Israeli team, ex-military, they had never even seen shoulder pads,” Harniman says. “So we went through several iterations just on the product side.”

This past season, according to Harniman, the UFL was able to deploy the camera on at least one player in 25 of its 43 games.

The UFL is also tinkering with tech that could further change the game itself. Right now, the chips Sportable puts in the league’s balls provide data like spin-rate and speed. The question is when chips like that will allow for automated spotting of the ball—an achievement Harniman calls a “holy grail.”

The challenge is pinpointing not just where the ball is at a given moment, but where it is when its carrier is declared down. “That is such a precise measurement and the only way that it will ever be solved with technology is by marrying a sensor and optical tracking system,” he says. “We’ll continue to work on that, but we’re not close to solving it yet.”

He adds, “I think it’s still a two-year curve.”

Harniman may be impatient, but two years would be lightspeed. Whether robo-spotting is two years away or 20, Dean Blandino says it would “have a tremendous impact.” Blandino, who helped launch the NFL’s current version of instant replay and served as its head of officiating from 2013-2016, now holds the same role for the UFL. He finds the technological advancements exciting, but believes the key will be how the tech is deployed. “We have to make sure that we maintain that human element of, it’s played by people, it's coached by people, it's officiated by people,” he says.

In other words, the games must still feel like they’re being decided by athletes. The current NFL playoffs have been full of tense, dazzling play, but also questions over whether replay was appropriately deployed, not to mention further litigation of the NFL’s most torturous question: What is actually a catch? (Give the wrong answer and you could be banished from the city of Buffalo for eternity.) Frame-by-frame Zapruder-style breakdowns like the one inspired by the Denver Broncos’ overtime interception of Buffalo’s Josh Allen is not what most people signed up for when they became sports fans.

Of course you have to get key plays right, Blandino says, but how precise do we need to be on a random second-and-8? He thinks about the NBA, where you often see a defender knock a ball out of bounds, which would normally mean the offense gets the ball back. But after a five-minute replay delay, we learn that the ball grazed off the offensive player’s fingernail and possession goes the other way. In baseball, we now have the phenomenon of a base-stealer easily beating the catcher’s throw, only for his body to slide off the bag for a millisecond while the tag is being applied—and to be called out on replay. Is that what fans want? As new advancements come online, how do we ensure that sports don’t get bogged down in technicalities?

Blandino believes that requires the most human of things: subjective judgment.

“Getting it right is number one,” he says. “But to what extent do we have to get every play right? Or are there the four or five plays during the game that really impact the outcome that we’ve got to get right? I'm more in the four or five play camp.”

There is one group trying to reach as many people as possible with their football AI innovations. It also happens to be backed by one of the world’s most powerful technology companies.

Every Thursday night during their NFL broadcast, Amazon airs a “Prime Vision” simulcast, based on Next Gen Stats. The broadcast shows the game from a behind-the-QB perspective, displaying all 22 players at once. The screen is full of color-coded signals pointing out different AI-driven insights. Amazon’s models will identify likely blitzers, receivers probable to be open, and predict whether a defense will play zone or man coverage. A “pocket health” feature flashes colors in the backfield to cue viewers into how much danger the quarterback is in.

Sam Schwartzstein—who played center at Stanford, captaining its 2012 Rose Bowl champion team before moving on to a career in tech and football—serves as an analyst for the broadcast. Along with other members of Amazon’s football team, including on-air talent like former NFL players Ryan Fitzpatrick and Andrew Whitworth, he works with a group of scientists specialized in AI in Amazon’s Tel Aviv office to develop new models and features.

Schwartzstein says that Amazon’s model for predicting blitzers is “better than me” at identifying them pre-snap (and he considers himself very good). As various tools have been developed, he has leaned on his former Stanford teammate Andrew Luck, once regarded as one of the NFL’s most cerebral quarterbacks and now the Stanford football general manager. “He was really fascinated by our coverage ID model, which is able to pick up man or zone, and we even have the ability to predict cover zero through cover six,” Schwartzstein says.

Luck became fixated on a play where an offensive player went in motion and a linebacker followed him across the formation—usually a tip-off that the defense is playing man-coverage—but the model correctly predicted zone. What was the model seeing that tipped it off? “He couldn’t figure it out,” Schwartzstein says.

Players on the field tend to focus on a single (or small handful) of keys before any single play. But football fields are full of interrelated tells—many of them are just beyond the capabilities of any single human to identify and decipher in the mere seconds before the ball is snapped. “The safety’s depth, the cornerback’s depth, the lineman’s alignment can all be other triggers to indicate, hey, this is not going to be man [defense],” Schwartzstein says.

There are obvious competitive advantages to be gained from these sort of insights. It’s against NFL rules for teams to use technology like this during a game—seeking to keep the human element of coaching, the NFL does not even allow teams to view video replays on their tablets on the sideline, just photo stills. But AI models can certainly inform teams’ preparation. Amazon does not share its information or models with teams, Schwartzstein says, but squads across the league are racing for their own advantages.

It sounds crazy to ask, but could AI one day be calling, or at least suggesting, plays? Helping design game-plans? Schwartzstein believes the NFL will be attentive to making sure the game is not taken over by tech. “I think that it won't intervene so much during the game that it will make it like we’re watching the spreadsheet,” he says. “We always joke, the game is not played on a spreadsheet.”

Honey and his company, SportVision, also invented Major League Baseball’s K-Zone, which superimposes a hitter’s strike zone over broadcasts. It launched on ESPN in 2001 and, not long after, Honey recalls a dinner with MLB executives. He asked whether MLB might ever consider using something like K-Zone to officially call balls and strikes. He was told that the league needed to wait for younger fans to come along who would be more accepting of technology. But not only that: “Newer generations won’t accept the fact that a game could be affected by a mistake,” he recalls being told.

Now, a quarter century later, automated balls and strikes are finally coming to MLB.

Honey admits he is torn over all the advancements. “When we raise kids, we try to communicate that life isn’t fair,” he says. “Just pick yourself up and get on with it.”

“As technology does a better and better job, we kind of lose the metaphor in sport that, you know, life isn’t fair,” he says.

Squadron, Honey’s co-founder, calls his former partner “indisputedly the greatest” sports technologist in history, but disagrees with him on this point. “I think we get enough of that lesson in you know, a great receiver dropping a ball over the middle,” he says.

Both views are, of course, valid—and speak to the impossibility of satisfying all fans as technology gets more and more bound up in the games we love.

During Week 14 of the NFL season, when the Bengals and Bills played a game in the snow at Buffalo, the elements interfered with the Hawk-Eye system, knocking it out. All season long, chain gangs have remained on the sidelines as backups. Finally, one was being called into duty. Technology has taken our games to amazing places, but it may also sometimes fail.