Our website use cookies to improve and personalize your experience and to display advertisements (if any). Our website may also include cookies from third parties like Google Adsense, Google Analytics, Youtube. By using the website, you consent to the use of cookies. We have updated our Privacy Policy. Please click on the button to check our Privacy Policy.

Transcript: How Artificial Intelligence Can Help Create a More Circular Economy

Transcript: How Artificial Intelligence Can Help Create a More Circular Economy

SEE THE SHOW NOTES AND LISTEN AT: How Artificial Intelligence Can Help Create a More Circular Economy

Katie Whalen [00:00:06] Hi, I'm Katie Whalen and join me each week as I talk with experts around the globe about Circular economy. You'll find out what's being done to make it a reality and if it can really solve the problems it promises. It's time for Getting in the Loop.

[00:00:26] Hi, everyone. Thanks for tuning in today. Katie here. I'm really excited about today's episode. We're talking about artificial intelligence and the Circular economy with Johanna Reimers of Refind Technologies. This week, as you may know, I'm at the World Circular Economy Forum in Helsinki and I'm hoping to meet some of you in person. If you're also here, let me know by sending an email to Katie@intheloopgames.com. The organizers are also live streaming the conference, so be sure to check that out and stay tuned for my recap of the World Circular economy Forum. And stay tuned for a recap in an upcoming episode. When I get back from the World Circular economy Forum, I'll be recording new episodes with some exciting guests as there is still a lot. I think we can talk about related to Circular economy. Stay tuned for episodes related to circular design, circular investments, as well as practical reflections on how companies actually implement circular strategies like repair, reuse and remanufacturing. If you have any questions related to these topics or you have ideas for other topics or guests you like to hear from, let me know. Send me an email to Katie@intheloopgames.com and I'll try to see what I can do. Who knows, maybe you'll be on an upcoming episode. Now onto today's show.

[00:01:49] Today on the show, we're talking with Johanna Reimers, who is CEO of Refind Technologies. Before co-founding Refind in 2014, Johanna worked as an I.T. consultant and project manager. In this episode, we're learning how A.I. can help create a more secure economy as Johanna shares all about the visual recognition software her company has developed. It can be used to store anything from batteries to even fish.

[00:02:16] Thank you so much for coming on the podcast, Johanna.

Johanna Reimers [00:02:20] Thank you. It's nice to be here.

Katie Whalen [00:02:22] Excellent. Could you give us a brief introduction to Refind Technologies?

Johanna Reimers [00:02:28] Yes, sure. Refind is a technology provider and we use cameras and deep learning software to teach machines and systems to recognize things based on their looks.

[00:02:42] And our first application for this technology was a battery sorting machine, which we call the optical battery sorter, and we have sold it worldwide since 2012. Then we have expanded the applications to also recognize different sort of- different kinds of used electronics and more specifically different types of phones. And in 2015, we also started looking at recognizing different types of fish species. So the recognition task is, as you can understand, is very generous. Well, capability and it could be used for many different areas. But I think there are many more promising application areas than the ones that we have tested it on so far. So I see I see potential for other types of image analysis applications.

Katie Whalen [00:03:44] Yeah. So there's a lot of different types of applications and we'll dove into that. Maybe you could tell us a little bit at first about what these different applications that you mentioned. The phone one, for instance, how does this work?

Johanna Reimers [00:04:02] Well, it works similarly as a human brain would work. So if you want to teach someone how to recognize different types of things, then you will show it images of how it looks and you will say this is this type of phone model, for instance, it's an iPhone 5 or it's an iPhone 4 or it's a Samsung, this and this. And it's good if you show variations of this particular identity like the backside and the front side, and maybe if there are different, if it can have different looks depending on how how much it has been used.

[00:04:46] So maybe it's a bit worn and torn or scratched. Any kind of variation will have, at least when we're talking about machines, it's one thing to teach a human something like it already has some sort of awareness about, but that a machine it could see a look at it as a blank paper, some say. So you have to show it many different samples of the same type of the object or category that you want it to recognize. So that is what we have done for phones in this instance and then also for batteries. We have shown, and when I say show, it means that we have taken photos of and then labeled these photos of different battery looks and battery looks- You recognize them by their labels. So we have taken thousands and thousands of images of different batteries because that is eventually how we will learn how to recognize things.

Katie Whalen [00:05:55] Yeah, so just today, I'm clear and how it works, so you get that, you know, a collector collects different types of phones or batteries and it comes into the technology, the machine and then it identifies it and is able to sort that based on you having taught it to recognize these different types of batteries or phones.

Johanna Reimers [00:06:24] Exactly. So the first step is building a- build up a database of on a test set us so to say. You need- In our case, we need a lot of images. We need a lot of labeled images. And that will be our test set. We then show these images to the neural networks, the software behind it, and they get exposed to these images millions and millions of times until it has learned how. What is the characteristic features of each of these different types of identities? The identity is up to you. What you decide. I mean, we have a phone model could be one identity, but it could also be like a closer product class as a phone. It could be the identity. It depends on what you want to select.

Katie Whalen [00:07:20] You've also been applying this to other things besides phones and batteries, you mentioned the fish sorter as well.

Johanna Reimers [00:07:28] Yes. Yeah. And that was something that- Well, it did not really come naturally after having worked in the electronics recycling industry. We did not look for a fish application ourselves, but we were asked- We got a request from the Nature Conservancy, which is a very big environmental organization. And they were working on a project where they needed to I identify different fish species is to be able to count the fish in the sea or make some sort of estimation of which fish species. Do we see an increase or decrease in and how big do they get, which is a measurement of how old they are. It was a very time consuming task to both take images of fish. And we're now talking about dead fish. Fish catches and have experts going through these images and identifying the different types. You could imagine that the fishermen themselves would be able to do this, but it's not it's not so easy when it comes to the Indonesian waters where this took place. It's it's one it has one of the largest variations or the largest fish, the riotous populations in the world. So there could be hundreds of subspace names for for the different snapper, some groups.

[00:09:01] So that requires a lot of expertise to know the Latin name of different fish. And they asked us, could you recognize these automatically or through your software somehow? We said, yes, because I think that's an even better application than anything man-made. I would say because nature has a way of making things unique. If there is a purpose to it, whereas man will copy things that the way that makes it a bit more difficult sometimes to recognize things.

[00:09:41] So I would say that I would say that the fish recognition this is one way. It's not a simpler task, but it has less time. And just in the form of of that, there are no illegal copies of certain brands and so on.

Katie Whalen [00:10:02] No knock offs. No knock off fish.

Johanna Reimers [00:10:04] No. Then. Then there is of course there are fish that imitates others and then are similar in that sense. So- well there are some challenges, but it's in a way it's a much more straightforward task.

Katie Whalen [00:10:22] So there are many, many, many potential applications, you know, fish and batteries. And it seems like, you know, using this learning technology, there could be potential for this solution and other similar solutions to contribute to a more circular economy. What can you tell a little bit about like how you see it fitting fitting in?

Johanna Reimers [00:10:47] Yeah. I mean, for the moment, I mean, circular economy, it's of course a very popular thing to talk about these days and if it comes down to two resource efficiency, I would say. We have always had the mission to reduce waste and the vision that one way to one way to do this is that we would use more automation or automation that has been the means to increase efficiency when it comes to production. So why shouldn't it be the means to also be more efficient? And what happens after end of first use or or whatever happens after an item has been disposed of?

[00:11:33] And we've I think definitely it can be AI or machine learning can be used for enabling more circular economy to take place so to say. But it's just as much as any other automation technology can be used in novel AI or machine learning is- Well, it's part of the automation flaw. I would say, as I'd say when it comes to industrial application. And it's definitely an area where you want to automate things that you have been done manually before. And when it comes to circular economy, I would say that. First of all, the driver is for circular economy today. Well, it's this is what I said is resource scarcity and then it's also the reduction of the environmental footprint. But whatever you do, it always has to work financially. And then if you can use artificial intelligence in a way to save costs, then that won't be the main reason for why it is used indefinitely to save costs. But I would say that it's difficult to say where it will make most have the most effect at this moment.

Katie Whalen [00:13:09] Yeah, I heard you speak at the Circular Materials Conference last- Was it March?

Johanna Reimers [00:13:14] Yeah, I did.

Katie Whalen [00:13:15] Last march in Gothenburg and you- Something that stuck in my mind. Maybe I'm getting this completely wrong. But you were saying that you could use it to identify- you could use these types of technologies to identify like how broken like a phone was or sort of like what was the  level of repair needed for like a computer or something? Or even just like the what make and model, which I think from a lot of the research that I've been doing in terms of looking at repair and extending product lifetimes a lot of the time and money from and challenge for recycling or sorry, not recycling, but for remanufacturing is like it's it's time consuming and the person who's doing it for them to actually figure out what kind of make and model, what kind of repair it gets do I need. What is the sort of- How much repair does this does is actually need is it profitable for me to actually spend the time repairing it? So if you had something like this, you know, could could it technology that could help you kind of this or put you in the correct direction? I think that could that could be an opportunity.

Johanna Reimers [00:14:30] Definitely. And I think that's the actual application that this would be used for. So I said automation in general will make things will enable more circular economy. And our technology is specific. Well, we are good at recognizing things automatically and an initial task for everyone who works with used goods or use products. It's the visual inspection. You start with a visual inspection and then you test it. Is it- Well, does it look okay? Does it work? Does it power up? Or if it's a no on some of those questions, you may look for how are the spare parts?

[00:15:15] And for all of those tasks, there is the visual inspection comes first. You could deduct a lot of information from a visual inspection, as you mentioned, you can say. Well, one thing is the make and model. Another thing is the aestetical looks of it, a kind of grading so to say and then for the functionality. Well, then usually you have to do some hands on work, like you actually plug in or press. If the screen works and so on. I think those steps could be automated as well. But they don't necessarily need to be machine learning tests. Those could be more simple or more advanced. It depends on what they are. But definitely I think our are the technology that we have today could be used for the visual inspection. And if you take a look at that, part of that could give something that today people are the companies that do not really afford it because it requires so much stuff. And but manual labor and you're not really sure what you will get out of it afterwards. You know that if you will send everything to a crusher and the smelter, you have the cost for it. That cost is quite low. I mean, the gain is higher as the material gains higher upwards. But that material gain is or profit is is usually lower than the cost it would require to have someone manually inspect each item.

[00:16:53] So you take a bigger risk if you are going to put a person to do to check everything. And then it is safe to have this to skip that part. It's easier to just crush it, smell it and get the materials out. So I definitely believe that our technology is really good to to enable well, that it's actually being done, that the visual inspection is being performed at all. But as with all of us with almost all automation, usually it requires a certain volume for it to make sense.

Katie Whalen [00:17:31] Oh, yeah. Yeah. That's something that we see being a challenge. I've been doing research and reuse and repair and it always comes back to volumes. And do we have enough of this second hand profit product to even have it make sense? Exactly. And I had noticed that you  introduced the first vendor reverse vending machine for batteries, which sounds really cool. Could you tell us about that.?

Johanna Reimers [00:18:00] Yes, it is. It is a cool project, and we love doing it. It was about two years ago we were contacted by Energizer who wanted to- I mean, they have very green awareness. Yeah, they have- they want to  make people aware that they are environmentally conscious or say they haven't made this eco advanced the label and they are reusing parts of batteries to make new batteries and so on. So this was as it was in one way, a marketing campaign for them to to launch a reverse vending machine and or so, of course, a way to do it, introduce a new way of collecting batteries and you incentive, which is. I think it was really interesting we had to build something that would recognize that it's a battery and not something else.

[00:19:01] We had to count how many batteries were put into the machine and then print a discount coupons based on the number of batteries inserted to the machine.

[00:19:15] And it had to work for, well, any kind of consumer that would show up at the machine, usually kids. So it had to be very self explanatory, absolutely foolproof. So no one got hurt or got stuck anywhere or did or would be able to do any damage to the machine itself. So we had to do some research. Well, what does it require from a if you're if you have two to manufacture something that will be standing? Well, in a more public area and also buildings by any kind of person or consumers, we have not done that before. We had built machines that would be installed in two factories and had very limited access to it to who could use them. So it was a challenge. And the way it had to work together with people of the human machine interaction is a new aspect for us. Technology wise, it was not super advanced. We had. It was quite a simple task to recognize that it was a battery at all. So that part was quite easy.

[00:20:37] But then all the practicalities around it that it had to be possible to move around this reverse vending machine because they were doing a campaign and they would have it tour around different locations. So yeah, I know. But once we hadn't built that machine with, we said, well, if we get to build a second machine, we will do a lot of changes to it just because to make it as you do to ship unto to you said so. But it was- I think it worked fine. It was really nice, really nice project. And it also it also had to be able to communicate with it so that we could be able to say, well, how how many batteries are inside it? Does it have to be emptied on someone?

Katie Whalen [00:21:31] Oh, yeah. So you could have the remote communication.

Johanna Reimers [00:21:34] Yeah. It had to have a little router on it.

Katie Whalen [00:21:38] Is that really the Internet of things like to get to the next level.

Johanna Reimers [00:21:43] Yes.

Katie Whalen [00:21:45] Yeah, I know that. I mean that's a really a really fascinating challenge. And also I really liked what you said in terms of thinking about it before you had been doing mostly, you know, in a in a sort of closed off business environment and now something that has to be more consumer facing. And you have these different, different challenges that go along with that.

Johanna Reimers [00:22:10] Yeah. But then just to everyone, I mean, their natural next question is, "Okay. So what happened? Have we solved this on one of them?" No, we haven't. Because I would say that the real challenge in a product like that is that, well, the business model around it is a bit more complicated than other types of reverse vending machines, because factories, they they have negative values. That means that if you are going to recycle a battery, I'm talking about small consumer batteries now. It will usually cost you more than you will get out of it so to say so, which is not the case if you're collecting aluminum cans or glass bottles or pet bottles. But then you have a very homogeneous material to begin with and you don't need a lot of people involved.

[00:23:07] There could be one machine builder and then whatever gets put into that machine, well, you can turn it into a commodity. You could turn it into some kind of raw materials that can be sold to a metal trader or something. But for the batteries, where there is no there is no deposit for the systems that say there is no there is the the producer has a responsibility to collect a certain amount of what he has put on the market. But he pays another company to do that and then pay that price, pay that company.

[00:23:49] So if it doesn't have the same financial flow as some other, I think other reimbursement machines usually have. So I would say it. But on the other hand, you need to well, in the European Union, we have to collect a certain amount of batteries each year. So what we have- I mean, that we launched this in Norway, in Norway, the collection rate is very high already.

[00:24:23] So there was no need to introduce a new incentive for them. But they come to us. We have heard and most interesting about this from our the South American countries where it's more of an environmental problem and they need to find ways to collect more or. I would say southern Europe, what about. That is a bit wrong. Eastern Europe is more correct.

[00:24:53] So in countries where they don't collect enough batteries there, they have been very interested in. But then as I said, it requires like a larger system. Someone has to pay for the machines to me.

[00:25:05] When someone else has to pay for that, the batteries are collected and recycled and then it has to be put somewhere, preferably a some sort of supermarket or somewhere where people are all going anyway.

Katie Whalen [00:25:25] Do you see- You mentioned I think that's so right that you pointed out with the value of the battery. It is negative, negative. There's no there's no really value. Value, therefore, for turning around and reselling some of those materials. And it's difficult to recycle. Do you see that as a challenge for you for implementing your technologies? And in that it's easier for companies to just make new products than to to have reverse logistics systems for collecting phones or or batteries?

Johanna Reimers [00:26:02] Yes and no. I stay for for batteries, yeah, it's a challenge. But I mean, in those countries where you have to collect batteries. Well, it has to be worked out somehow. And then our technology has been useful. It's it's quite cost efficient. And the reverse vending machine, it has to be built in higher volumes to become more cost efficient. But I don't know how these companies or producers collect their batteries today. It's probably not for free. So it has to cost somewhere pretty short. But when it comes to other types of products, there is a lot of value in phones.

[00:26:42] I mean, you wouldn't get paid for it to recycle your phone a minute basis or people would pay for those things, but not for every thing that is an electronic product. So if you're an electronics recycler, you will also end up with a not so valuable products. So I think that's the problem.

[00:27:07] But you cannot just pick only the value with the materials you have to kind of play. Also, the white good sand, the lawnmowers and the Christmas candles or whatever. All the other things that are not so valuable, but you still have to recycle them. So that is the challenge with the waste stream in itself. When we call it waste, it gets mingled with a lot of other things.

[00:27:44] It usually becomes waste just by being next to other things that are not the same thing as as itself. So say so when you thought maybe things had it, it becomes more expensive to separate it later on. I would say if you would just create a more closed loops from the beginning in different types of product streams then I think it would be much, much easier. So now you have to weigh in the fact that you will end up with a lot of other things, you don't know what you will get. So it's quite unpredictable what you are going to do. That is the biggest challenge for this cycle. You don't know what you may end up with. If you knew already, then you could be very you could find better and so on. But now you will end up with pretty much anything.

Katie Whalen [00:28:44] Maybe I should have asked you this earlier, but I'm just thinking now, like, so who are- Who are your- who are the main customers that you're working with? These batteries- like these recyclers, but also you have a lot of fun like the fish program. It's not just- it's not just, you know, people in the waste stream, for example.

Johanna Reimers [00:29:06] No. Well, most of our customers are battery sorting customers. So they are quite similar. It's usually a private owned companies. They either do only the collection and sorting of batteries or they do collection, sorting and recycling of old batteries. So they they are experts at what they're doing and that makes them usually good customers because they know what they're getting. They are also- although they're very widely spread around the world, which makes it difficult just geographically to reach all of them, they are at least the working in very similar ways. The other business, the fish business. We are not we are not that far ahead yet.

[00:29:58] So our main customer has been the Nature Conservancy located in Indonesia. They are an organization. So they they are not a normal customer either. But and we are such a small company. I think it's been a challenge just to understand the electronics recycling market or industry. So what we have done within the fish industry is that we have hooked up so say with a reseller company and that is what we continue doing. We're trying to do that for the battery market as well. And that's something you can do when you have a more mature product. I would say it's not easy to do when you're still on a pilot face or prototyping phase. Then you need to be a closer connection to your customer or the specie grade, which is the Fisher's business recognition equipment that we are launching now.

[00:31:09] Then we are working with a Norwegian company and he will or that company. They will sell the products. They will also maintain the products once it has been it's the one that that has to be our business model when it comes to battery sorting as well. So we are moving towards that. We're outsourcing more of the production and we also outsource more of.

Katie Whalen [00:31:35] You're the technology experts and provider and then-.

Johanna Reimers [00:31:39] Yes.

Katie Whalen [00:31:41] That's where the focus is. And it's not about maintaining a machine or doing with a-

Johanna Reimers [00:31:48] You know, to a certain extent, we, of course, need to maintain the machine. But once- it depends on which level. I mean, with the battery sorting equipment, we have been- well, we have been working a lot with that to improve it, to make it as easy as possible for our customers.

[00:32:09] And on the issue there is that when you're working with dirty things, that may not always be batteries, it could be it screws or or lids or other metal things that end up you don't know what will happen. So you have to say it's mechanical. Tell us just to move things around. And that is what we have been improving the last year. But I think now we're in a position where it works really well. We don't need to to improve it a lot from here. So we are now more secure and we can have other people sell it for us. But you can't do that if if if there are stops or things happening with a wall. Exactly. It will be there. Yeah.

Katie Whalen [00:33:01] So what does the future for for Refind technology, especially with this idea of, you know, this interest in AI and Circular economy. What what do you want to see you going forward?

Johanna Reimers [00:33:13] I see that we launch a third product soon. I see that we launch a more general sorter. We call it the Refind Sorter. It would not be focused on batteries or fish. It would be a more general sorting terminal. You could say it could be used for any kind of visual recognition and also sorting. And it comes in in different size models the sorting modules. So that's why we have some discussions going on. I hope this will be the next thing we launch. But then after that, I also hope that we will be able to to only license out the software as a recognition software just for image analysis, because a lot of there will be a lot of image analysis tasks that will not require any physical handling or movement at all. So I think that will be our third or maybe fourth area. Because we get a lot of interest there. It's not all about industrial applications, it could be any kind of image analysis. But I think that will be within a couple of years. We will have a product.

Katie Whalen [00:34:46] Yeah, well, I look forward to hearing hearing more about about this. And and I wanted to ask you the final question that I ask all of my interview guests, which is about that in the loop game that I created. It's about Circular economy and your your producer. And you're trying to collect materials to make your product. And then there's different events that happen in the game that change the market conditions. And they're often inspired by real world events. So they can be positive or they can be negative. And I always ask the guests if you could create an event based on your expertise. What kind of topic would you would you focus on? So, Johanna, if you have kind of an idea in mind.

Johanna Reimers [00:35:33] Yes, I have an idea. I have what I talked about a deposit system for banks, I think could be an interesting game changer is a big it is a two big word, but it would it would put a new focus on the collection and also the recycling of batteries. And I think it would if people were incentivized to to actually return their batteries. Well, then there would be a lot more batteries to take care of. And then there would be opportunities for more technology or more processing processes being developed around this. I think that could be him very well. I think it could have a very positive effect on the overall resource efficiency. Well, some say the entire industry will affected.

Katie Whalen [00:36:33] I think so. I do think it could be it could be a game changer as new as you said. Yeah. And thank you so much for coming on the podcast. Where could listeners go to learn more about you and Refind Technology?

[00:36:48] Well, our web page is a good place to start refind.se where we have information about who we are, what we do, and the latest and upcoming events as well. We are also active, not super active, but active now on Twitter and that is Refind tweets.

[00:37:12] Thank you for listening. I hope you enjoyed this episode. For our show notes and links, go to our web site at gettinginthelooppodcast.com. And while you're there, subscribe to our mailing list to have new episodes delivered to your inbox every Monday. See you next week.

 


circular economy presentation guide overview ppt pdf introduction


About the Show

Getting In the Loop is a weekly podcast dedicated to exploring how to transform to a more circular society. Join host Katie Whalen as she examines the challenges facing our current resource use and discovers alternatives to the ‘take, make, dispose’ way of doing things. Each week she interviews circular economy experts about what they’re doing and learning. Together we'll uncover what circular economy means in practice and find out what's being done to keep our resources in a loop rather than sent to waste.

Related Posts