The rapid rise of autonomous delivery is transforming how goods move through cities, and at the forefront of this revolution is Serve Robotics. As the company behind the cute, self-driving sidewalk robots navigating city streets, Serve Robotics is redefining last-mile logistics with a focus on AI, efficiency and safety. Uber and Nvidia are among the publicly held company’s investors.
The AI Innovator recently spoke with Ali Kashani, the visionary leader steering Serve Robotics into its next phase of growth. We discuss the challenges of scaling autonomous delivery, the evolving role of AI in robotics, and what the future holds for this innovative industry.
What follows is an edited version of that conversation.
The AI Innovator: Can you tell me how the company was started?
I started this as a project, or really a division inside Postmates, back in 2017. … We were d working on sidewalk robots as an independent division. Postmates was acquired by Uber in 2020 so we joined Uber, and then I spoke with Uber about spinning this into an independent company, which they agreed, and we became independent at the end of 2021 so we are about four years old as a company. And then a few months ago, we took the company public on Nasdaq.
How does your system work? Where are the robots stored? How does it get to the customer?
It’s actually pretty cool when you think about it. The robots have their own depots. We have multiple depots in Los Angeles right now and there are new cities coming online soon.
So the robots leave their home in the morning, they go to work, they stay out all day, and they come home at night, and while they’re out, they can accept jobs from Uber or other partners when they’re available, and basically they hang out in a busy part of the city where the restaurants are until they get a job.
Once they get a job, they go pick up from the restaurant, which gets a notification on their tablet that there is a robot outside so they go load the robot.
It’s really straightforward (to operate). You don’t need any onboarding. You don’t need to know how it works. You can just walk to a robot and operate it, and then the robot goes to the customer. You can track it on your app, just like a driver.
You get a message when the robot is there. You can press a button on the Uber Eats app to unlock and grab your food, and then the robot would be on its way and you don’t have to tip, which is a value proposition for customers. They love that nobody touches your food.
It’s actually more reliable to use these robots than human couriers today. It’s more likely to get the right food to you. So people like them. Merchants like them because making fewer mistakes means less headache for them as well.
Will people steal the robots from the street?
We’ve been around, operating in even rough areas of Los Angeles for over five years. There was one time when two individuals actually attempted to steal a robot. … They picked up a robot, they put it on the back of their truck, and then the robot jumped off and came home in one piece, completely unharmed. Within half an hour, the individuals were arrested.
How many robots do you have in LA and where do they operate?
Last year, it was 100 robots, but the numbers are evolving. We’re growing our fleet. Today, we deliver for about 900 restaurants today, just in Los Angeles.
We are in neighborhoods … where restaurants are concentrated, the business districts, basically, and then in a couple mile radius, you have a higher density population. Those are the places where we operate. We are in downtown LA, we are in Hollywood, West Hollywood, Koreatown, there’s a long list. We just went to Long Beach. There is more, and it’s expanding fairly quickly.
This year, we have 2,000 robots that we are under a contract with Uber to deploy on their platform, and we plan to do it by the end of this year. They’re manufactured by Magna International and assembled by them in Detroit.
What cities are you expanding into with those 2,000 delivery bots?
Expand in LA, we’ve talked about a few others like … Dallas and other potential targets. Some of them are going to happen sooner than others.
You’ve reached Level 4 autonomy. How do you handle real world obstacles, like people, pets? Tesla has not even reached Level 4.
There are a couple of important advantages that we have. One is we don’t weigh two tons (like a car), and we don’t move at 30, 40, 50, 60, miles an hour. That means we have 3,000 times less kinetic energy than a car. Kinetic energy is what makes cars unsafe. If they come into contact with a person, they transfer that energy, which our bodies can’t handle. So the impact of a car could be fatal.
With our robots, even if they come into contact with someone, the most likely scenario is that people keep walking. It’s a very different safety profile.
On top of that, our robots can stop at any time when they need help, and this is what makes them Level 4. They can stop and ask for help, and a human can remotely help them, and that makes the whole problem so much easier to solve.
By my estimate, self-driving cars need about 99.9999% of accuracy just to match human level driving. It’s roughly about one in a million. It has to get it right in all but once in a million times that they cross an intersection, because if they do more than that, that’s the rate where human errors cause accidents.
For our robot, … we don’t cause an accident if we stop, and that makes all the difference. That means that the AI we need to enable these robots to work at Level 4 is orders of magnitude simpler.
You said that the bot can ask for help. How?
We can connect to any robot at any time. We see what they’re seeing, see the sensors, health reports and how it’s doing. How’s the battery, tire pressure, everything? We can check that remotely. So if a robot needs assistance, a person can actually assist it by telling it what to do.
How fast do the bots move?
They can move up to 11 miles an hour.
How long does it take to get your food?
The average time from us picking up something to dropping it off is like 18 minutes or so. It depends, obviously, on the distance. Our median delivery distance in LA is 1.3 miles, so it’s not very far.
How much of human intervention do you have for these bots?
This is the stuff that is obviously super-confidential. What we do share is that one person can be in charge of multiple simultaneously moving robots.
What does the human operator do?
For example, we had a case where a robot came across police tape. It didn’t know what to do, so a human operator stepped in. Or the robot had to navigate a construction site. (However,) when we’re crossing intersections, we want (human) supervisors to be involved. That’s temporary. Eventually, we would not need them involved. But right now we have them involved.
Does the robot operate on the street or sidewalks?
Sidewalks. … Imagine a really cute shopping cart with eyes and personality.
What was a challenge you had to overcome before developing and scaling this technology?
We are pioneering something here. We are making something that no one has ever made, and we are bringing robots to public environments. One of our first questions was, how do we make sure that people like them? Because suddenly this new thing is showing up in your world, you’re just walking by, doing whatever you do every day, and suddenly a robot shows up. Is it going to be a hindrance, or is it going to be something that adds to your day?
Especially in the U.S., we had our work cut out for us because of Hollywood – there’s such a negative perception about robots and a dystopian future. On the other hand, it’s also really cool and exciting. So we kind of took it as a challenge. We have this opportunity to make this really fun and something that adds to your day. I think we succeeded.
We have some super fans, like children who absolutely love the robot. They go with their parents looking for robots. Like every week, I have pictures of people kissing the robots, and generally the reactions are super positive. But that was obviously an important problem.
… It is also making our cities safer. We can remove cars off the street. About one every four cars that’s on the street is basically doing a local shopping trip or errands. It doesn’t need to be there. As we say, you don’t need a two-ton car to move a two-pound burrito. By removing (cars), we are making cities safer. …
A million people die every year from car accidents.
We can make cities cleaner. We can take some of the space that belongs to cars and give it to people, and we can bring the cost of last mile down for the customer. This is one of those things that makes a ton of sense when you start thinking about it, but if you execute it poorly and people react negatively, you can have setbacks. So we wanted to get this right, and I feel like we did pretty well. This was one of those areas where I’m very happy with the outcome.
Beyond Uber, do you have plans to expand outside of food delivery maybe in the next three to five years?
Absolutely. There are a lot of other things you can deliver. Food is just something we eat three times a day. So it’s the most common (use case) and it’s a big market but also kind of the most expensive, because you want the food in 30 minutes. So it is the most expensive kind of last mile delivery.
But imagine medications, pharmacy, parcels, groceries. There’s also reverse logistics: How many times do you buy something and then you have to return it, and go through that whole nightmare? Or there could even be new ways of shopping. Let’s say I order shoes on Amazon and I have to wait days for it to show up. … Instead of ordering it from Amazon, I could have ordered it from a local store. It could have shown up in a robot in three different sizes. You could try it on right there, pick the one that fits and put the other two back in the robot to return.
There are some really cool things that could become possible all of a sudden, and it would really improve local commerce, which I think is a really fantastic outcome. So there are a lot of other things we can do with these robots once they’re out there, and they’re making the cost of last month substantially lowered.
Do you see a time when sidewalk robots will fully replace human couriers? Or do you think there’s always going to be a hybrid?
If you look at most examples of automation, humans stay in the loop one way or another. You just make humans more productive and capable. This was true even with ATMs, which is a classic example. With automated teller machines, everybody was really concerned that tellers were going to go out of a job. It actually increased the number of tellers because banking became cheaper and more affordable as a result of ATMs, and banks opened more branches.
It turned out there were a lot of people who were not using banks. So once these ATMs came, they started using banks and it actually resulted in more tellers having to be hired as a result of the benefits of the ATM.
I believe it is going to be very similar (with delivery bots). The cost is so high now for the last mile. If you order something from China to your home in the U.S., it costs $2 to get it, but if you order it from your local Chinatown, it would cost $10.
If you reduce that cost, you’re going to have this effect of massively increasing the market, which means humans are going to be needed on every part of that stack, basically from the businesses that need to serve those customers to the delivery, to the running of the robots.
I don’t think it’s ever going to be robots taking over the whole thing. It’s always going to be complementary. The market is going to be just much bigger and there will be a lot more people involved.
Be First to Comment