Robots are like us: struggling with baby goats, car nightmares, and LA-to-Miami Beach culture shock

While the public often imagines autonomous robots as cold, calculating machines, the reality of deploying them in public spaces reveals a technology deeply concerned with social acceptance and survival. MJ Burk Chun, the co-founder and vice president of product design for Serve Robotics, addressed the Fortune Brainstorm AI conference with the argument that robots are just like us.

The ‘long tail’ of the baby goat

The trouble often begins when the machines leave the controlled environment of a simulation and enter the “wild” of city sidewalks, Burk Chun said. During a deployment in Los Angeles, the delivery team found that the real world was “even more dynamic than we expected.”

In one specific instance, a robot froze, “thoroughly confused about the pet baby goat” standing in its path. While the robot’s sensors could identify a human pedestrian, the goat represented a “long tail problem”—a statistical outlier that standard training data had not prepared the AI to encounter. Like a person seeing something inexplicable on their morning commute, the robot simply didn’t know what to make of it.

Nightmares on Main Street

It isn’t just confusion that plagues these droids; it is also fear. The intersection of two streets is described as “one of the most dynamic places in our cities,” filled with high-velocity vehicles that pose an existential threat to small delivery devices.

“Robots have nightmares about cars,” the executive said without elaborating on how she can tell when a robot is having nightmares, or what those might be like. “Cars are also very scary for robots.”

Robots must constantly calculate the risks of sharing public space with heavy machinery, she explained. To cope, engineers have to spend significant time determining if a robot is “safe enough to cross the street,” assessing everything from pedestrian light signals to the status of the ground.

Coast-to-coast culture shock

Perhaps the most relatable struggle for any human who has relocated is the difficulty of adjusting to local culture. The robots, it turns out, are not immune to this.

The company found that the “conservative routing” algorithms optimized for Los Angeles—designed to handle “very high traffic high-speed intersections”—did not translate well when the fleet expanded to Florida. In Miami Beach, drivers tend to “cruise” rather than the Angelenos who race to make a turn, meaning the robot’s hyper-cautious LA programming was out of sync with the local rhythm.

“The future really is already here … it’s just not evenly distributed,” Burk Chun said, paraphrasing the great science-fiction writer William Gibson, who first began popularizing the concept of cyberspace back in the 1980s. (Neuromancer is a particular Gibson classic.)

“It is also quite amazing how each city expresses itself in the way people walk,” Burk Chun said. “Not just the sidewalk infrastructure, but also how people drive.” She said every city expresses a unique “flavor” that a robot has to learn when it moves there, just like a human.

A guest in the neighborhood

Underpinning these anxieties is a strict social contract. “Robots don’t have rights to be on sidewalks, people do,” Burk Chun asserted. This philosophy dictates that engineering decisions must be “socially aware,” prioritizing human comfort over robotic efficiency.

Because “more people will walk next to the robot … than we’ll ever get a delivery from a robot,” the machine is viewed as an ambassador. If the robot fails to “deliver delight” or provide value to the community at large, it is viewed as a missed opportunity to build a harmonious future.

To earn their keep, these robots are doing more than delivering lunch; they are working as municipal inspectors. Using advanced sensors, they collect data on “missing curb cutouts” and “hidden potholes,” sharing that information with cities to help repair physical infrastructure.

For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.