• 5 months ago
As CEO and co-founder of Starship Technologies, Ahti Heinla has spent a decade proving that the future of urban delivery isn't airborne or underground: it's autonomously rolling down your street. With 10+ million autonomous deliveries worldwide, Starship is the world's only physical AI proven at scale in real-world robotics. In this Founder's Story, Heinla reflects on the conviction it takes to build hardware at scale before the world knew it needed it, what true autonomy demands in the messy reality of city streets, and what Starship's journey reveals about Europe's potential to lead the next wave of intelligent mobility.

Category

🤖
Tech
Transcript
00:18okay hello hello hello again black stage we've got an amazing conversation coming up for you
00:24now I know that everyone here will have been speaking and saying AI hundreds and hundreds of
00:30times today we know that we're stepping into a world of AI that's very exciting but this gentleman
00:37here is already optimizing it and deploying it on our streets hands up if you know where Milton
00:45Keynes is anyone no what about Northampton these are British cities so thank you very much Ahti
00:54very normal bog-standard cities but also Tallinn Hamburg and multiple sites all across the US
01:02the future of AI isn't happening in the future it's already here it's arrived on six wheels
01:10you just perhaps weren't paying attention right now autonomous robots are delivering groceries
01:16hot food spare parts right navigating the chaos of the real world and all of us humans one pavement
01:23at a time not hundreds of deliveries guys not like the one or two robots you're seeing on
01:29the shop floors here at VivaTech not thousands millions of robots VivaTech Black Stage please
01:38be upstanding for Ahti Henler the CEO and founder of Starship Technologies please
01:45now audience I'm sure you all want to know what these robots look like right so please watch your screens
01:53watch your screen and see you next see us
01:53that's a nice day
01:53I'm sure you all want to know what these robots are doing
02:07you can see us
02:08that's a nice thing
02:08clink-a-intensive
02:08and there's the main thing
02:08share a lot more
02:08you can see us
02:09that's a nice thing
02:09and we're doing
02:10and you can see us
02:10you can see us
02:10there's a lot of
02:12we can see us
02:13we can see you
02:13can see us
03:23Amazing, wasn't it?
03:25Thank you so much for you obviously showing us how and where these robots, these delivery
03:30systems are all going.
03:31Now, before we get into the nuts and bolts of what they do and where they go, let's discuss
03:36you.
03:37You're a disruptor.
03:39Hands up if you know his background.
03:42No?
03:43Yes.
03:43Wonderful.
03:44Fantastic.
03:45So, this is a gentleman who is behind the tech of Skype and then you launched and founded
03:51Starship Technologies.
03:53You put technology in play while others were still thinking about it.
03:58So, talk to us a little bit about your journey and what is the link between these two sort
04:03of big brands that you have created?
04:06Is it perhaps a need to connect people?
04:09Tell us about yourself.
04:10Yeah.
04:11So, I am an engineer by training and by background.
04:15My parents were both software developers.
04:18They met at the workplace.
04:19My mother taught me how to program computers when I was 10 years old and then I started
04:25building all sorts of different technologies and being part of a number of different teams.
04:31Probably best known of it is Skype.
04:34I was part of the founding team and it helped bring about one revolution in the telecom industry,
04:41effectively changing how people talk to each other, how people communicate over long distances.
04:47And nowadays, of course, this is very, very common.
04:50Everybody has video calls all the time.
04:52But back then, when we started, this was really a disruption.
04:56This was really something that a lot of people thought we were crazy or a lot of people, wow,
05:01I mean, phones already exist.
05:02I mean, why do you build this product?
05:04Is anyone going to use it?
05:06So, we obviously believe that this will change the world and it did change the world.
05:11And I am right now doing another revolution in the world and that is in the world of transportation.
05:17Where there is also, actually, you know, there's a revolution about to happen or happening right now,
05:23which is automating delivery, automating transportation.
05:26When you look at science fiction movies, you typically don't see a future where there is lots of delivery couriers,
05:36you know, riding on bikes or whatever they do.
05:39You see, you know, you see things flying to you, right?
05:42You know, you see things created out of thin air.
05:44And that's the future we are building.
05:46And we are, this future is here right now.
05:48We are doing millions of deliveries with thousands of robots right now.
05:53We are, we're not operating in Paris.
05:56That's the reason why you're not seeing them outside of everywhere.
05:59But if you are, for example, in the Slush Conference in Helsinki,
06:03I mean, in Helsinki, we are operating, we have hundreds of robots in Helsinki and everybody sees them all the
06:08time.
06:10What is the public reaction when they see these robots on the streets?
06:15Do you think people are still surprised today?
06:17I mean, audience, quick show of hands.
06:18Are you surprised when you see these delivery robots on the streets?
06:22Do you wonder where are they going?
06:23What are they doing?
06:25Yeah.
06:25So is that surprising for you?
06:27I mean, you've been building all of this.
06:29You're making, and this is the thing about what he does, it's making the kind of technology,
06:34the futuristic technology that we all see all around the world,
06:37it's available for everyday uses and everyday people, isn't it?
06:41So talk to us a little bit about how all of this is working and where they're going to.
06:46Right.
06:47One of my biggest surprises when building this product and this technology was actually how easily people were accepting robots
06:56on the streets.
07:00I actually thought, you know, in the beginning that, you know, there might be some discomfort or some people might,
07:05I was thinking maybe some people would call the police when they see an autonomous robot, you know, driving on
07:10the street.
07:11And the aliens were coming or something.
07:12We have done more than 10 million deliveries so far and nobody ever has called the police.
07:17And it's just actually, it's actually shocking to me even how accepting people are with robots.
07:25And we at Starship, we see that effectively because there's lots of robotics companies in the world, I think.
07:34There's lots of talk about physical AI.
07:36We are probably the only physical AI company that is actually at scale in commercial deployment, you know, having thousands
07:45of robots around.
07:47And we see the data and we see the experience at scale.
07:51So, not just like one sort of one pilot or one test, you know, somewhere, but in hundreds and hundreds
07:58of cities right now.
08:00Like, you know, in some of the countries where we operate, we do about 25% of all same day
08:06grocery deliveries right now.
08:07So, we have grabbed significant part of market share.
08:11There are the tens of cities where robotic delivery is the norm.
08:16It's not an exception.
08:18It's not like, you know, wow, I got delivered with a robot.
08:21But, you know, soon it's going to be, wow, it's really, you know, it's strange, you know, an unusual thing.
08:26I got delivered with a person.
08:27Like, the default is going to be the robot in a few years, I think, everywhere.
08:33Now, Tallinn obviously makes sense because you're from Estonia.
08:36But why have you chosen the cities that you're currently operating?
08:40I mean, obviously, like we said, it isn't just a few robots.
08:42These are millions.
08:43So, where has the data, where and why has the data sent you to those particular locations?
08:49We are deployed right now, I think, in, what is it, seven countries or, you know, I lose track because
08:56we're launching in a new country every couple of months.
09:00And we started out in our commercial operation.
09:04We started out in the UK in the city called Milton Keynes.
09:07Now, Milton Keynes is a very unusual city.
09:11It's not an average city in the UK.
09:14It's kind of what's called like a planned city with very sort of wide sidewalks and it's a very sort
09:22of spread out and, you know, it's a very unusual city.
09:26So, we chose that city for our first commercial launch in 2018 because we wanted to launch in the easiest
09:35place possible.
09:36In the place where it is the easiest for robots to drive.
09:40Now, we have graduated beyond that.
09:43Right now, we are more than 100 cities and it's no longer the easiest places.
09:50So, next city we launched in was a city called Northampton, which is not actually very far away from Milton
09:57Keynes, but it's a totally different city.
09:59It's a, it's a more like a traditional old, you know, you know, English city with, you know, some, you
10:06know, cobblestone streets somewhere and so forth.
10:08So, the, you know, now we can do all sorts of different environments and we have deployed in Finland, for
10:17example, all of the top 20 cities in Finland, plus about 50 or 100 small towns and villages as well.
10:26So, we are no longer selective in where we go.
10:29We just grab market share and we just deploy in tens and tens of cities all the time.
10:35Now, there was something really interesting in what you said that you started, obviously, you know, two different cities in
10:40the UK, Milton Keynes, Northampton.
10:42One has wide streets.
10:44So, easy for, obviously, the deliveries we made for the robots to run along those streets.
10:49There are six wheels to go across those streets.
10:51You said Northampton is a little bit different.
10:53It's more traditional.
10:54So, when we, I mean, I'm sure everyone has seen some of the other robots, a few robots.
10:59They're the ones that almost look like humans.
11:02And, obviously, the ones at Vivotech are doing okay.
11:05But, you know, you'll see these videos where the ones that are, the robots that look more like human, if
11:09they try to, you know, if the floor, essentially, that they're walking on isn't flat, they sort of fall over
11:15or they have an accident.
11:18So, when you mention the two different kinds of cities, one has wider roads, one is more traditional, is navigation
11:25of the robots, is that something that is difficult for you?
11:29Or have you inbuilt that into the technology, into the wheels, into the structures of the autonomous delivery as well?
11:35A great question.
11:36Even, actually, the autonomous driving software and technology for the robots to make the right choices of whether to turn
11:43left or right or whether to go faster or slower and, you know, which way to go.
11:48You know, this is the key in this technology.
11:50Like, when I said that, oh, we're building this future where transport is automated, people have had that dream for
11:57a long time.
11:58For a long time, people have wanted, you know, to have these flying cars and, you know, and transport automated.
12:05People have wanted that for a very long time.
12:07It's not that we invented this idea.
12:09We were just the first company to actually build a system that actually works.
12:13So, it has been difficult, overall, for the humankind to build autonomous driving technology.
12:24When, I think, you know, when Google announced, I think in 2009, I think they announced that they will have
12:30a self-driving car on the market in five years.
12:34So, now it's like, you know, 17 years later and, you know, you could say way more cars, you know,
12:40are on the market than they are.
12:41We're expanding, but it's taken, has taken longer than people expected, you know, at first.
12:46It's very, very difficult to build autonomous driving technology.
12:49Our robots look like, you know, it's like a simple box, but there's so much technology inside it.
12:55And the intellectual property, the technology that we have built there, that is the main asset of the company.
13:01That's the main asset of Starship as a company.
13:05There's lots of companies that have tried to do this, but not gotten it to work.
13:11So, that's why we are the leading company, because we made it work.
13:15How did you make it work, then?
13:18There's a lot of technology and a lot of data.
13:21Data is very, very important, because it's actually pretty easy to have, to build a robot that drives 100 meters
13:30autonomously in a straight line.
13:35And maybe it has an accident during that time as well.
13:37That's actually pretty easy.
13:39But the level of reliability you need to build in, in order to actually deploy commercially in the real world,
13:49is staggering.
13:52Like, you know, for example, our robots, we have thousands of robots driving all the time.
13:59And collectively, our robots, for example, they cross roads about a million times per week, collectively.
14:07So, our robots do a million road crossings per week.
14:10And that's the scale that we are operating at.
14:14And imagine if our robots have, you know, made some mistake in the middle of the street.
14:19We're crossing the street, you know, 1% of the time.
14:23Like, we would have so many accidents or incidents or issues.
14:28You know, we couldn't really, we wouldn't be able to operate.
14:31So, the level of reliability needs to be much, much, much, much, much higher.
14:37And it takes a lot of data, a lot of experience to iron out all of the corner cases to
14:42see that, you know, okay, we did now 1,000 road crossings.
14:46You know, 998 of them were completely nominal, but two of them had some sort of an issue.
14:52Let's fix the issue and just go on like years and years and years like that.
14:58And then you get a reliable system that's actually reliable enough to deploy in the real world at scale, doing
15:06millions of deliveries.
15:07That's why we have gotten there.
15:09We are the leading physical AI company in the world.
15:11We're the only one that has deployed at scale.
15:15Like, we're kind of on par with Waymo.
15:16Like, Waymo is also kind of in the Western world, like, the only company that's actually deploying autonomous vehicles at
15:25scale.
15:25They're just going city after city and thousands and thousands of cars.
15:28So, we do the same with the transportation.
15:31They do people transportation.
15:32We do goods transportation.
15:34So, would you say reliability is what keeps you and your team up at night, making sure that, you know,
15:39every robot gets to its destination,
15:41even if there is a small percentage or 1%, that there could be a problem along the way for it?
15:46I would say that kept us awake at night seven years ago.
15:51Now, I would say that is solid, even though we are improving the tech all the time.
15:57But there was a phase in the life of Starship when I, as the CEO, I stood in front of
16:03the company every week on our weekly all-hands meeting.
16:07And I reported what is our safe at the number.
16:10We quantified this, you know, how reliable are our robots.
16:13And my message to the company was that we need to get this 250 times better.
16:18I mean, 250 times better.
16:21250x.
16:22And so, I stood in front of the company every week and said, okay, now we have 250 times to
16:26go.
16:26Now we have 200 times to go.
16:28Now we have 170 times to go.
16:30And I did that for two and a half years.
16:32So, two and a half years of improvements.
16:34And then we got to the goal.
16:35Then we got to the goal that our robots were safe enough, reliable enough to deploy in the real world,
16:41to do millions of deliveries, which we are doing right now.
16:43What were some of the problems then that you had about two and a half years ago that you fixed?
16:49So, overall, this sort of AI technology is pretty hard to actually get reliable.
16:57If you think, like right now, over the last couple of years, of course, the world has kind of realized
17:02in a way that, you know, oh, now computers can do things they couldn't do before.
17:06So, now there is a lot of intelligence kind of available in computers, in technology that wasn't there before.
17:14And when you use AI behind your screen, you know, you ask like an LLM a question and you think,
17:22wow, this is fantastic.
17:23I mean, it's just the answer is like 90% of the time, the answer is kind of correct, but
17:27it still makes 10% of mistakes.
17:29You know, it doesn't know how many hours are there in strawberry and all of these things, right?
17:34Now, the level of reliability we need to have when navigating the physical world is much, much higher.
17:42Like 10% error rate is a complete non-starter.
17:48So, it's by far not as easy as putting chat GPT in the robot and asking, should I turn left
17:55or right?
17:56Like your robot will crash, you know, within the first hundred meters if you do just that.
18:01So, obviously, there's a lot of AI technology in the robot, but it's a lot of care has been taken
18:06to make sure that the robots are safe to deploy.
18:10And they are.
18:11We're doing millions of deliveries now.
18:13How are you tracking every single delivery to make sure that it gets the destination?
18:18Can you take us through some of your processes?
18:21So, we used to deploy robots at first, like this is a long time ago.
18:26This is like, you know, eight years ago.
18:28We deployed in a way that, you know, that we had, you know, human oversight.
18:34At some point, 10 years ago, we actually had people following robots just, you know, physically, right?
18:40You know, right now, this sounds a little bit like unthinkable, but back then, that was the state of the
18:46industry.
18:47The autonomous vehicles were also driving in a way that they had safety drivers in the cars all the time,
18:54you know, just for safety.
18:55You know, somebody with like a hand on a button or something that, you know, red button, stop, you know,
19:00emergency stop.
19:00You know, that's how also we operated eight years ago or 10 years ago.
19:05So, we did actual deliveries to actual people, thousands of deliveries in this way.
19:09I have done many deliveries myself, you know, walking behind a robot.
19:14But then, we became actually the first company in the world to deploy robots without any human oversight in 2017,
19:24where we started driving without people following robots, also without remote oversight.
19:31There was also a phase where we had a remote oversight that there wasn't any physically following the robot,
19:36but there was always a person watching on a screen, you know, remotely in an office, you know.
19:44So, nowadays, that also doesn't happen.
19:46We do have a remote assistance service so that when there's something unusual happening,
19:51then, yes, a person can take a look remotely on what's going on.
19:54And, you know, Weibo has the same, everyone has done it the same way.
19:59But this is rare, like 99% of the time, robots are driving by themselves.
20:04So, you have no human oversight, really, now.
20:07Now, some of the conversations, I'm sure everyone's been seeing them all across VivaTech,
20:11some of the conversations I've had on stage when it comes to the deployment of AI and how we're using
20:16it.
20:18There's the question of humanity.
20:20Where is humanity's place in all of this?
20:23But when it comes to autonomous delivery, you're providing a function,
20:26a really important function.
20:29So, if we think of perhaps, I don't know, a pandemic,
20:31when people couldn't have people-to-people contact anymore,
20:34I can see these six wheels, these robots coming in and making, you know,
20:39very important deliveries or, you know, whether it was somebody needs milk
20:42or someone needs some, you know, medical equipment or something is,
20:46it would be a very important function.
20:48But should people be scared then that this is the future
20:53and that there isn't necessarily a human element involved?
20:59I think delivery is actually, or transport, is a pretty tiny part of the humanity, I would say.
21:09You know, we are looking to automate a lot of it,
21:12but I don't actually think this takes anything significantly away from humanity.
21:18We do grocery delivery and food delivery.
21:22Most people don't really enjoy doing grocery shopping.
21:25They do it for a practical purpose, for a utility, because they need to eat.
21:30But they would rather hand it off to a machine, which is what we are enabling.
21:38So, talk us through what's inside the box then, in terms of what are you delivering,
21:42who to, where to, you know, what is it, and perhaps what comes next after that?
21:49Are you scaling, you know, obviously this is something that you're scaling,
21:52but are you also scaling what goes inside it, what is being delivered?
21:55Right.
21:56So, we do grocery delivery today and we do hot food delivery.
22:02So, grocery and food, that is what we typically deliver.
22:07And we work with almost all of the delivery apps in Europe,
22:14means Uber Eats, Vault, Vault, Foodora, and Just Eat as well in the UK.
22:23And so, there's fewer delivery apps that we don't work with
22:28than the delivery apps that we do work with.
22:32And we also work with many of the major grocery retailers in Europe.
22:41And we are still covering, overall, if you think the whole, you know, global market,
22:46we are covering a tiny, tiny percentage of it.
22:49In some places, like in Finland,
22:53we do about one quarter of all same-day grocery deliveries right now.
22:57So, everybody in Finland knows our robots.
23:00They are completely commonplace, completely mass market.
23:04Like, all major customers in, all major potential customers in Finland
23:09are either working with us or talking to us.
23:14And the same revolution will happen in other countries as well.
23:19Like, I think most of the Western markets will kind of switch to robots
23:26in the next couple of years.
23:29We'll have robots do a significant percentage of the delivery.
23:32I think, you know, human couriers will not disappear,
23:35but a lot of the couriers or a lot of the deliveries will be done by robots
23:40because the technology completely exists right now.
23:43And, of course, we are making it better all the time.
23:47But this is a solution ready to come to your city, wherever you live.
23:54So, it's coming soon, and you've got big partners, big players in the system.
23:59But was there a time when you were in the sort of early stages
24:03that it was really hard to convince people to partner with you,
24:07for them to believe that this is efficient and it will work?
24:11Absolutely.
24:12I think five years ago, seven years ago,
24:18the delivery apps and retailers were interested in this technology,
24:23but they didn't really think that it's solved right now,
24:28that this is really a solution for them.
24:31Particularly for retailers, the issue is the cost of delivery.
24:36A lot of grocery retailers do not offer delivery.
24:41And why do they not offer delivery?
24:43Very simple.
24:44It's not economical for them.
24:46It's what they call margin drain.
24:50It erodes the profits that they are making.
24:53It's loss-making for them.
24:55That's why they are not offering delivery.
24:59But people want delivery.
25:00They just don't want to pay as much for it as it actually takes.
25:05Now, robots is the solution.
25:07And with our robots, we are actually offering them to make delivery
25:11instead of a margin drain, to make it something that they call margin driver,
25:17something that helps them make money and helps the consumers,
25:22their customers, get what they want, which is delivery.
25:26That's really interesting what you say, that the retailers don't want to have it
25:33because obviously it costs them money.
25:35Now, I don't know about anyone else, but when you do your online shopping for groceries,
25:38obviously where your autonomous robots aren't available,
25:41there's always an add-on.
25:43You put everything into your basket and then there's maybe like a five pound or something,
25:47sorry, or I could say five euros because we're in Paris, obviously.
25:50There's like a five-year add-on for the home delivery.
25:53So where would you come in in this ecosystem?
25:56How do you make it easier for the retailers so they're not spending?
26:00But then how do you make sure that the end user, the customer,
26:03is also not having to maybe spend too much money on this autonomous delivery robot?
26:08Yeah, we are a B2B company in terms of our business model.
26:11So we don't sell directly to the end consumers.
26:15We sell to the retailers and delivery apps.
26:17So we have a pricing agreement with them and generally we are offering them cost savings
26:23that then they are passing on to the consumers as well.
26:28Now, let's come to the slightly more, well, you're talking about scale.
26:33So we're talking about scale.
26:35We're in Europe.
26:36What's the one word everyone starts talking about?
26:38Regulation, right?
26:39So where does regulation help you and where does regulation fail you right now?
26:44Or make things difficult, I should say.
26:46I think probably in a conference here, there's lots of people on the stage.
26:54And when the topic of regulation comes up, probably everybody is going to say that Europe
26:59does too much regulation.
27:02I also do say that.
27:07However, I wouldn't actually say that my problem as a CEO of Starship is that regulators want
27:14to have some scrutiny on our robots.
27:16Of course they do.
27:17I mean, safety is important.
27:20It's just like cars have regulations.
27:23There are technical inspections and lots of standards and so forth, the certifications.
27:29There should be something like that, of course, for robots on the sidewalk as well.
27:35Our robots are much smaller than cars.
27:37Cars, you know, they travel slower.
27:39They're much, much safer than the cars inherently, right?
27:43You know, then they're going to, like, kill a person, right?
27:46But, of course, there needs to be some sort of guardrails.
27:49It shouldn't be completely, you know, free for all that everybody drives whatever robot
27:53they want, you know, on the streets.
27:55So that's actually not my problem.
27:57I would say my only complaint, again, you know, for regulation is actually that it's slow.
28:03But overall, I see regulators really are very good at actually embracing technology like this.
28:10My experience with regulators is completely positive just except for one thing at a time.
28:16Like, it sometimes happens, for example, that, you know, we are discussing with regulators
28:22in a country where we are not operating right now.
28:26But, you know, we submitted an application, you know, for a permit.
28:30And they said it takes six months to review that application for a permit.
28:35I mean, six months.
28:36I mean, it's just like, I mean, it's insane, you know, from my perspective.
28:42I mean, it's not just this one permit that we need.
28:45Like, you know, it's like a second larger expansion after that and so forth and so forth.
28:50So I would say these sorts of timelines are ridiculous from my point of view.
28:55But other than that, I have actually only good words to say about regulators in Europe.
29:01I think they're actually very forward looking, you know, very much looking to change regulation
29:06for the new technology to come.
29:09So I'm actually very, very pleased on the regulation in Europe.
29:13I don't think there should be no regulation on robots.
29:16I don't think so.
29:17No, no, no.
29:17There should be a regulation.
29:18It just should be about 10 times faster.
29:22But lots of countries have changed the regulation over the last decade to make sidewalk robots
29:29completely legal and regulated in the right way.
29:32You know, many countries, you know, like Finland, Estonia, US, I mean, UK is moving on this.
29:38Germany is moving on this.
29:40I think this progress is very, very, very much happening.
29:43So what will the future be?
29:45Will it be, I mean, obviously, you want it to be autonomous delivery, but will robots take
29:50over the future?
29:51Should us humans be scared or should we be happy?
29:54What's the future?
29:55I think it's not necessarily me that wants delivery.
29:58It's our consumers and customers that want to get delivery and affordable delivery.
30:05I don't think people should be scared at all.
30:09Like our robots are doing a very, very narrow task in the world, which is just transporting
30:15one thing to another.
30:17So that's not scary at all.
30:20Thank you so much.
30:22We're exactly on time.
30:23It's been a pleasure.
30:24Please be upstanding for him, please.
30:26Thank you, audience.
30:28Thank you, everybody.

Recommended