Robots learn to ask humans for help
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Sarah Grillo/Axios
Delivery robots roll through cities with ease — but can't open doors or push elevator buttons, driving new AI that lets them ask humans for help.
Why it matters: People and robots will increasingly interact in the real world — at work, stores and even on the street. But the technology needs to be seamless to avoid chaos and frustration.
Driving the news: Serve Robotics, with 2,000 sidewalk robots across 20 cities, is partnering with T-Mobile to introduce new AI-powered conversational robots that interact, engage and think in the moment.
- A prototype of Serve's next-gen delivery bot, which they call Maggie, is friendly and engaging — and not afraid to ask for help.
- "Excuse me, could you push the button for me?" Maggie asks a pedestrian at a busy crosswalk. "Thank you," the robot says, as the pair cross the street together in a promotional video.
The big picture: Until recently, the biggest challenge has been designing robots to safely navigate the physical world — to see and avoid obstacles, for example.
- Now, as large language models have advanced, robots can add real-time comprehension and communication.
What they're saying: "We have solved the final pieces of the puzzle needed for robots to fully integrate into society," Serve Robotics' CEO Ali Kashani tells Axios.
Yes, but: Giving language and comprehension capabilities to robots takes a ton of battery power and computing effort.
How it works: T-Mobile's edge AI network helps offset those computing demands.
- It provides low latency, enhanced security and localized data processing, avoiding trade-offs in robot battery life, operating time and added hardware cost.
What's next: So far, Maggie is Serve Robotics' only conversational robot, but Kashani is convinced they'll be everywhere eventually.
- "This is the direction the world is going to move, and we're trying to be at the forefront of it."
