A bit of a technology ‘man for all seasons’, Harri Holopainen started his career in computer graphics on a Commodore 64. He worked on smart card payment systems, co-founded a small graphics software company, and even designed and implemented a prototype online gaming world, a subject he did his university thesis on. Upon graduating, he and his partners grew their computer graphics software company, Hybrid Graphics Oy, until NVIDIA stepped up and bought the company in 2006. Harri later struck out on his own again, as a partner at Love of Technology Strategies, and co-founder of Microtasks, a microwork company.
In 2013, Harri stepped into the world of machine learning and robotics, at ZenRobotics, a company that builds smart robots for waste sorting and recycling. Founded in 2007, they are at the cutting edge of applying AI-based (they call it “ZenBrain”) robotics to sorting all kinds of trash. Their mission is nothing less than defining Next Generation Recycling. They have two main products, a ‘fast picker’ that is aimed at traditional mixed recycling streams, and a ‘heavy picker’ that can sort construction and demolition waste materials. The latter makes up to 6900 picks per hour using multiple sensors and can be found in Scandinavia, throughout mainland Europe, China, Japan and Singapore, and even in the U.S. There is even a system running on wind power, in Sweden.
Over the last nine years, Harri has served at ZenRobotics as Robot Lab Head, Head of Technology, and now, CTO. He describes himself as a generalist, having worked on VC rounds, defined product strategies, negotiated licensing agreements with Ericsson and Nokia, headed R&D development teams, and even hand-built critical robot components. But as he notes now, “Lately I’ve also been up to my elbows in trash.”
Read the podcast transcript here
Eve Picker: [00:00:09] Hi there. Thanks for joining me on Rethink Real Estate. For Good. I’m Eve Picker and I’m on a mission to make real estate work for everyone. I love real estate. Real estate makes places good or bad, rich or poor, beautiful or not. In this show, I’m interviewing the disruptors, those creative thinkers and doers that are shrugging off the status quo, in order to build better for everyone. If you haven’t already, check out all of my podcasts at our website RethinkRealEstateForGood.co, or you can find them at your favorite podcast station. You’ll find lots worth listening to, I’m sure.
Eve: [00:00:53] Harri Holopainen has a mission. To define Next Generation recycling. A bit of a technology ‘man for all seasons’, Harri started his career in computer graphics on a Commodore 64. He moved onto smart card payment systems, co-founded a small computer graphics software company, and even designed and implemented a prototype online gaming world, a subject he did his university thesis on. But in 2013, Harry stepped into the world of machine learning and robotics at ZenRobotics, a Finnish company that built smart robots for waste sorting and recycling. And there he helped build their A.I. based ZenBrain robots, which sort all kinds of trash, first as a robot lab head and now as CTO. Harry describes himself as a generalist. He’s worked on VC rounds, defined product strategies, negotiated licensing agreements, headed R&D development teams, and even handled critical robot components. But lately, he says, “I’ve been up to my elbows in trash”. You’ll want to hear more.
Eve: [00:02:23] If you’d like to join me in my quest to rethink real estate there are two simple things you can do. Share this podcast and go to RethinkRealEstateForGood.co where you can subscribe to be the first to hear about my podcasts, blog posts and other goodies.
Eve: [00:02:55] Hi, Harri, thanks for joining me today.
Harri Holopainen: [00:02:58] Hi Eve, glad to be here.
Eve: [00:03:00] So I was really fascinated reading about your career. You did early work on a Commodore 64 and technology has defined your career. So, like everything from computer science to online gaming, I’d really love to hear about this trajectory and what the common thread is for you.
Harri: [00:03:22] Well, I think the common thread has always been in working with technologies that at some point will have an impact in everyday life and that sounded quite sort of absurd actually when we started to work with computer graphics, but then along the way came computer game consoles that started to bring home computer graphics into living rooms. And then you’ve got PCs and finally, you got mobile phones and I remember the first things, when there was a big customer asking for us that they would like to have these graphics very advanced graphics on a cell phone display that had, maybe, I don’t know, 80 times 60 black and white pixels. And we were thinking kind of like, yeah, like, what’s this is never going to go anywhere. But then again, a couple of years later, we realized that it’s the user interfaces of these mobile phones that actually will require quite sophisticated graphics. And this sort of graphics portion of my life ended in 2006. We sold the company that we founded to Nvidia, who was then and is still the number one graphics software and hardware company. And also the movie Avatar came out. And I remember seeing Avatar, and my first realisation was that, OK, so I have been in computer graphics long enough, so my work is done. Time to find something else to do.
Eve: [00:05:13] Interesting. So, you moved on to many other things and you have ended up in machine learning at ZenRobotics.
Harri: [00:05:26] Yes. And I’ve been here now for eight years, and this is, well, I would say that the primary sort of thing that comes to mind is that nobody really is against robots picking up our trash, and everybody agrees that there’s quite a lot of trahs out there and it ain’t going to recycle itself. So, it’s kind of a no brainer thing to do to apply robots there. And that’s also another kind of technology, which has been in people’s imaginations for over 100 years. But then this idea of smart robots actually doing something useful outside assembly lines, it still hasn’t quite happened. And I feel that this waste sorting is one big step towards that direction.
Eve: [00:06:24] So ZenRobotics sorts waste, all sorts of waste. And what’s your role there?
Harri: [00:06:33] My current title is CTO. The first thing that I did in the company was that I made the first prototype of the current type of robots that we currently use. Made it big for the first time. And then after that, I have been working in basically, since we are sort of a smaller company and need to move fast, so the research is very fast paced activity. So the research is the things that you think you can sell in 12 months’ time. And basically, those are the projects that I’ve been spending almost all of my time in, and it involves things like mechanisms for grouping waste and, of course, mechanisms that actually can survive in a waste plant and then also a lot of higher-level software to make the robots really earn its pay at the customer site.
Eve: [00:07:32] Let’s step back a bit. So ZenRobotics is a company that basically sorts waste using robots.
Harri: [00:07:40] Yes.
Eve: [00:07:41] And I read somewhere about the ZenBrain. What’s the ZenBrain and what are the products you’ve developed to sort waste
Harri: [00:07:52] ZenBrain is basically the collection of technologies that use a variety of sensors to look at the waste on a conveyor belt and then recognize the objects on the belt and then figure out how the robot could actually grip the objects on the belt. And then finally, the pieces of software that tell the robot to move over. So that the object from the belt actually ends up in the correct place, and the first application area was construction and demolition waste. And there are the objects can be quite large. I think we are talking about maybe 30 kilograms of maximum weight for pieces of concrete and stuff. And then the second robot that we have done is a robot designed for handling packaging and light waste. And the difference there is that that robot is much faster. But of course, it doesn’t need to lift 30 kilos because most of the things that it picks are things like hamburger cartons and plastic bottles and things like that.
Eve: [00:09:09] So they have different brains. So, what’s the problem that Zen Robotics is trying to solve? Why was the company launched?
Harri: [00:09:17] The company was founded by two old friends of mine and then some waste sorting experts. And the first slogan for the company was that’s basically “let’s do something cool with robots and A.I.” And then they try to figure out what that might be. And they actually did quite a lot of, sort of, small-time projects. I think there was discussions about going to fisheries and make a robot that picks up the dead fish from those containers before they make all the other fish sick. And that’s an interesting challenge in gripping that fish. And then there was another project done for a nuclear power company where the challenge was to, Recycling of these fuel rods that was apparently required some, some high level A.I. So then, at one point my friend, who has often trouble getting sleep in the evening, he was basically just at his home watching TV and there was Discovery Channel on showing images about these staggering piles of waste that’s, that you can find in all around the world. And then he realized that how about applying A.I. to make a robot that actually can sort waste? And it sounded very easy because, of course, you have these industrial robots, and they are not really that expensive. So that’s problem solved. And then there’s already, back then there was equipment that was used to identify materials on a conveyor belt. So, just put those two things together and we will have a robot that sorts trash. And it didn’t turn out quite, to be quite that simple.
Eve: [00:11:16] Simple, yeah, that’s what I was going to say. Sounds simple, but probably not.
Harri: [00:11:20] Yes.
Eve: [00:11:21] So I’m really fascinated by the whole construction industry and how this might impact it. And have you seen a change in approach to recycling materials over the years? And how readily is this being adopted in real estate projects or demolition projects or anything like that?
Harri: [00:11:42] Back when we started, the idea of using robots to sort construction and demolition waste was quite sort of novel. And when we were discussing people, then there was this category of people who were forward-looking. Back then they quickly realized that, actually, this makes a lot of sense and also so that it’s, it should be also quite profitable. And today we are in a situation where pretty much all the recycling industry agrees that robots are one important piece in this puzzle of getting circular economy work. So there’s quite a big, sort of, change in overall attitude. And of course, on the practical side, the waste industry is, first of all, it’s quite conservative. It’s not really the kind of an industry that immediately jumps into all the new things out there. And also the existing waste processing plants are quite large and expensive. So, even if today we would invent something completely sort of ground-breaking, then it would take quite a lot of time before the customers could actually employ it because these new breakthroughs, they don’t make any practical difference. If you have a 20 million, year old plant that you just built last year, and it’s incompatible with that. But now, actually this year, we have seen an opening of two new plants, one here in Helsinki. It’s about 30-40 million Euro plant, and it’s designed around robots. And most of all, the plant is designed to recycle waste so that none of it goes to landfill. And that’s quite a fantastic sort of starting point.
Eve: [00:13:43] That’s amazing. Yeah!
Harri: [00:13:45] And there’s also another plant in Switzerland that’s opened, also this year, and they are employing robots to recycle actually concrete and other inert materials. As you may know, cement industry is one of the biggest CO2 polluters. And the point of their plant is that they will take in concrete, stone and all the other inert mineral materials and then recycle it into something that can be used to make a concrete with less cement in it.
Eve: [00:14:18] Interesting.
Harri: [00:14:19] And that’s also a kind of plant and process that you can’t have without robots because there’s no other way to sort that kind of material.
Eve: [00:14:29] So what countries are at the forefront of this Next Gen recycling trend?
Harri: [00:14:35] I think that the waste industry itself is quite interesting because it’s especially, in C&D, it’s quite a regional industry and there’s a lot of regional differences. And that means that there is not that much competition globally. Because obviously, if you do C&D sorting in Finland, it would be completely unfathomable to just not be competing with companies in the US, for example, because you can’t transfer the waste itself, nor you can really transfer the end results of the recycling. And so, our customer, first customers ended up being the first adopters, essentially, all around the world, which is and has been quite challenging because we are a small company in Finland and our then first customers were, well, one of them was, well, a couple of them on in Central Europe, then one in the U.S., then I believe we have one in Australia and then one in, I think, Singapore or Japan.
Eve: [00:15:49] Oh, interesting. So, I’m Australian, you know, so that’s thumbs up for Australians. So, your company is in Finland, but when you say that customers, do they buy these robots from you? Is that what you’re selling?
Harri: [00:16:03] Yes, we sell the, basically the robots and then our customers are the companies that operate waste sorting facilities. And of course, we are in close cooperation with the companies that design these waste processing plants and processes and equipment.
Eve: [00:16:27] Ok. It’s really interesting. So, you have a fast picker and a heavy picker. And you describe, the heavy picker is really used for the construction industry, and the fast picture is for light boxes and things and like, what’s next? I mean, there must be other pickers in the, I’m a Picker too, but that’s not what we’re talking about. There must be other pickers in the works, right?
Harri: [00:16:58] At this point we have about, I think, maybe 60 arms around the world in production and we are currently scaling up. And it’s really no problem for us of identifying potential new use cases because there’s basically one new potential use case coming up every week. And there’s the, yeah, there’s like, for example, textile recycling is one big area where there are very few existing solutions. And then there’s obviously scrap metal and all that entails.
Eve: [00:17:38] Salvage yards, yeah.
Harri: [00:17:40] Yeah. And then recycling processes for cars and electronics. And there’s the recycling process for used batteries. Like practical problems like if you have a facility that recycles lead acid batteries, then it’s rather straightforward because you strip out the plastic shell and take out the lid and then basically, you’re done. And but then again, in that pile of batteries, you have a used lithium-ion battery, if you put that battery in that process, it may explode there, and that’s going to be a big problem for them. So that’s a typical kind of place where this added complexity of basically the everyday products out there will pose these interesting new challenges to companies that are already recycling things. And then there’s obviously, there’s a potentially very large amount of waste categories that are not really yet recycled at all because there is no economic way of doing it.
Harri: [00:18:51] And construction and demolition waste, there are other ways to do it than with robots. One thing to separate, for example, wood and light plastics from stones is to dunk them in water and then skim what floats. And that kind of works but of course, it makes everything wet, and soon that pool of water itself will be contaminated. And then, of course, there’s manual waste sorters are what are currently used in the quality control of municipal waste and also in construction and demolition waste and pretty much every sort of waste process where there is a significant sort of operation going on. And of course, one of our entries to the market has been that we will reduce the number of manual sorters required. Well, the possibilities are, of course, endless and unlimited. So that has never been our problem. So this picking and sorting is the easiest thing that makes a difference and has commercial value. But of course, after you have a robot that’s good at picking these things, why not use the robot to tear them apart as well?
Eve: [00:20:10] One thing that springs to mind, I saw a amazing show where a woman had an architect design a house for her and they used the wings of a decommissioned airplane for the roof, which was just fascinating, you know? But the fields of decommissioned airplanes are just crazy. I don’t know if anyone’s tackling those.
Harri: [00:20:30] Yeah, that’s also, and I would think that that entails a massive amount of manual labor. I guess a similar use case is decommissioning of ships, which I believe basically happened by, I don’t know, stranding them on a beach somewhere and having them [???]
Eve: [00:20:49] And then they just rust.
Harri: [00:20:51] Yeah. Or then there’s like 200 guys that come with, I don’t know, pliers and angle grinders and that, and put it into tiny pieces and.
Eve: [00:21:02] Interesting.
Harri: [00:21:03] Very, very manual, intensive, and very hazardous work.
Eve: [00:21:07] So I have to ask, what is the economics of this look like for someone who wants to deconstruct a building manually using a robot? Is it cheaper than sending out a crew?
Harri: [00:21:17] Well, I think if you have a building that needs to be decommissioned, then today I’m not really sure if our customers use the robots as a unique selling point, because the point of the robots for our customers is basically just to be able to give you a better price because there’s less, the operation has less cost. And of course, especially in the municipal waste, the regulatory bar is obviously rising constantly, and that obviously applies also to C&D sorting. That means that there are higher sort of regulations for the total operation of demolishing a building because you can’t demolish a building and then just dump it somewhere. So at the end of the day, that, at least it will mean that the prices of putting stuff in landfill, they are quite steeply rising and that forces the operators of these recycling facilities to make their processes more efficient.
Eve: [00:22:30] Interesting, so can you tell me what your team looks like? And you said you’re a small company? What does that look like?
Harri: [00:22:38] In the early, earlier days when a lot of the stuff that we had to do was quite sort of exploratory in nature, then I think I maybe had a 10-person team at that point. And I think we are about 60 persons at the moment. And then nowadays, when our focus is on delivery and maintenance and making sure that our customers get basically, professionally built and maintained equipment, then that means that the role of sort of rocket science is something that is luckily less needed today than five years back, when we still had problems in making sure that the robots actually keep working. And now, at the moment, we are focusing on making sure that our first about 50 customers are happy. And also, my team is now basically focusing on measuring and estimating the performance of the robot. And that’s actually quite a fascinating problem because one thing that people really don’t realize about waste is that waste is extremely hard to measure. The only thing that is easy to measure is to drive a truck on a weigher and notice that there’s 20 tons of waste in the truck. But then again, measuring what’s inside that container. The only known way of measuring it is actually to have some guy come over and take a peek.
Eve: [00:24:10] Interesting. That’s the manual bit, right?
Harri: [00:24:14] Yes. And that’s currently a quite a massive blocker in the waste industry, because if you think of an industrial process, it works because it’s measured. Whereas in the waste industry, it’s a bit difficult to even notice whether the process is actually working well or not. So, if you have a facility that sorts plastic, let’s assume, let’s say that this facility provides 10 tons of HTP plastic a year. So how do you know that there’s actually 10 tonnes of plastic instead of nine tons of plastic and one ton of other stuff? Well, you don’t really know. And of course, you will know if you have a process that really dislikes these contaminants, then you notice that something went wrong when you put into that HTB plastic in the process and you notice that there’s an explosion, then you notice that maybe there was a couple of these nice lithium-ion batteries inside that 10 tons of HDP. And of course, that’s too late. And in order to prevent that, there’s manual checks that are done more or less sort of consistently and the problem of this manual checking is that it’s expensive and it’s also very difficult to get a statistically relevant measure of basically a pile of waste by just a guy eyeballing it. And connection with robots is that the robots actually do look at every single object that comes under their sensors, and they take a really hard look at it and they may determine whether it should be picked or not. And that means that the robots actually can tell you quite a lot of what the customer actually had flowing in his waste process. And there are also some other sorting equipment that can tell that but they are not quite widely used yet, and they definitely are not used at the front gate of these waste processing facilities. So whatever people put in the waste basket that will at some point end up in one of these facilities, and no one really knows what the stuff is, we see one glimpse of it, and we are working in making sure that the robot can actually tell something useful of the waste itself. And over time, it may be that the knowledge of the waste itself, that might even be more valuable to the customer than the sorting result.
Eve: [00:27:00] So, yeah, I always wonder about sorting residential waste, which, I can’t imagine is an exact or efficient process, I think most people probably ignore the guidelines for recycling, and everything ends up being dumped in one place, so it feels like all that waste you’ve got to go back to the beginning.
Harri: [00:27:20] That’s an interesting question about how much people should be sorting at home. And I guess the extremes are that, especially in the US, there’s, in a lot of public places, there’s a big container where you dump everything, and it says that it’s sorted somewhere else. And then another extreme was that I was skiing in Austria some years back and that flat that we rented, it had nine garbage bins.
Eve: [00:27:50] You know, that’s very common in Germany, too. My husband has shared photos of me of these recycling bins and even more so there’s limited hours when you can put glass in them because it might disturb the neighbors.
Harri: [00:28:03] And you need to have nine of these in your kitchen. So, they’re under the sink there’s three, and I don’t know beside the sofa, there’s two and there’s a couple of in the cupboard over there and it’s just complete insanity because if you have nine categories to think of then it just, it’s ridiculous. It will just get people annoyed. And it’s also not efficient at all, because the problem is that you need to have nine different trucks visiting your home, or you have, need to have one truck that has nine compartments. And they all fill at different sort of pace,
Eve: [00:28:41] And you have to have someone who’s diligent enough to fill them properly, right? Yeah, the human element.
Harri: [00:28:46] I don’t mind that because of course, we’ll happily sell robots that fix those issues later on at the plant. But I personally, I think that there’s like, first of all, this bio stuff, leftover food and that should be kept separate because that’s really a nasty thing because it will foul up everything else. And then after that, well, I would say that glass is quite straightforward. Uh, in Finland and other European countries, at least we have this, and I guess in the US too, there’s
Eve: [00:29:26] Some places, not everywhere.
Harri: [00:29:28] Yeah. You’ll return empty bottles, and you get some money back.
Eve: [00:29:32] Yeah.
Harri: [00:29:34] And so that makes sense. And then cardboard and paper, probably. But then if you put people starting to sort of recycle different kind of plastics, then it’s just not going to work.
Eve: [00:29:48] But even the paper like, yeah, some people argue that they put the dirty pizza box in the paper recycling, but it’s dirty, it’s got food in it.
Harri: [00:29:58] Yeah, yeah there’s a lot of this. My wife has also lived in Germany, and she also lived in Switzerland for a while, and they are absolutely sort of fanatic about what the neighbors put in the trash.
Eve: [00:30:12] So recycling is a really big business, and maybe your robots have to develop a sense of smell as well. In the ZenBrain,
Harri: [00:30:20] I felt that for a long time we have all the technology that we will ever need. So, the technology is are not really the difficult bit. The difficult bit is actually finding a customer who can make a business out of a process that has a robot. And for these new areas where they are no working large scale solutions, it’s going to be really hard because they would need quite a massive capital to set up a shop that would produce enough of these, whatever resulting fractions that would be, where the volumes would be so high that using those fractions would be a business for someone else. So, if you want to recycle textiles, I guess recycling textiles itself is not necessarily that hard. Uh, but the problem is that exactly what are you going to recycle, what are your fractions and what’s going to happen to those fractions? And that’s, what are you going to do with, for example, cotton that has been reclaimed from textiles. Do you, like, it would be really stupid to like, incinerate it. It would be even more stupid to put it into a landfill. There’s a company that does these sound insulation panels out of the reclaimed fiber.
Eve: [00:31:45] There’s a company in Pittsburgh that makes fabric and is done very well out of plastics.
Harri: [00:31:49] Yes.
Eve: [00:31:50] So actually, plastics from Haiti, so they’re very, very specific. I don’t imagine they have robots sorting that in Haiti, but that’s what they do. Yeah, interesting. So just to round up, what are some of your favorite success stories, you know, people where things really change because of one of your robots?
Harri: [00:32:14] I would really say that this recently opened facility at one of our customers, Remeo, here in Finland. It has 12 robot arms, and the plant is designed not to send anything to landfill. That’s quite a remarkable achievement.
Eve: [00:32:31] Yeah, that is.
Harri: [00:32:33] And the plant is brand new and it’s quite, sort of, well it’s something else. I’ve seen a lot of waste processing plants and all of them are fascinating in their own manner. But this is something new and it’s enabled by robots, and it has taken us basically 10 years of work to get there.
Eve: [00:32:52] Interesting. So that’s a glimpse of the future for sorting waste. Nothing goes to a landfill.
Harri: [00:32:59] Yep.
Eve: [00:33:00] Well, thank you very much. You’ve been heard to say “lately, I’ve also been up to my elbows in trash”.
Harri: [00:33:07] Yes.
Eve: [00:33:08] So I’m just wondering, are you having fun? Is this interesting work?
Harri: [00:33:13] Yeah. Waste is fascinating because going to a waste plant, well, the first thing you notice basically might be the smell, but the big thing in these waste facilities is the conveyor belt. That’s where the waste is flowing and it’s just mesmerizing. And you’ll see all of the, basically, by-products of humans living, and for some really completely inexplicable reason when we go at the site where our big robots sort construction and demolition waste, there’s, like, uncanny amount of shoes on the belt. Yes. And I just, at some, we looked at data on one of our sites in Norway, and that was only for one day, and I just basically had to calculate the rate of shoes appearing on that line. And the conclusion was that if that rate holds for a month, they will have a ton of shoes. And it’s really like, absolutely amazing because if you go on the belt, it goes like half a meter per second and there’s a shoe and then, whoa, that’s a shoe, and then wait for 10 seconds or a minute, hey, there’s another shoe. But you can’t figure out how many shoes there actually are over a one day, or one week, or one month of production. And that’s the kind of things that’s really…
Eve: [00:34:42] Really fascinating.
Harri: [00:34:43] You never get bored.
Eve: [00:34:45] No. So, I have to ask, are there more women’s or more men’s shoes?
Harri: [00:34:50] We haven’t really made statistics, but I’m actually absolutely positive that at some point, our A.I. will have this built in function in detecting shoes.
Eve: [00:35:02] This is really fascinating. Well, thank you very much for joining me. I really enjoyed it, and I can’t wait to see what you scale up to.
Harri: [00:35:11] Thank you.
Eve: [00:35:11] Wonderful.
Harri: [00:35:12] Yeah, me neither.
Eve: [00:35:17] Smart brains building smart robots to sort trash in very smart ways.Eve: [00:35:24] You can find out more about this episode or others you might have missed on the show notes page at our website RethinkRealEstateForGood.co. There’s lots to listen to there. A special thanks to David Allardice for his excellent editing of this podcast and original music. And thanks to you for spending your time with me today. We’ll talk again soon but for now, this is Eve Picker signing off to go make some change.
Image courtesy of Harri Holopainen, ZenRobotics