Get Geeky with Cropster as they talk about how A.I. works with their innovative roasting software. Transcription of the interview is below for those that like to read, but I recommend to watch the video below or get it on our Interviews page and subscribe on our YouTube channel to get notified when we do more.
The Interview Transcript
Nick
So why don’t we start? So I’m going to be recording from this from this point, forwards. Why don’t we start with you doing an introduction into cropster there for those that maybe don’t know, I mean, you are very, very well known in the industry. But let’s let’s not assume anything. Give an interview about or give it an introduction as to who cropster there is and what do you do.
Martin
Sure thing. So. Well, I mean, it’s like there’s probably a very long answer. I mean, cropster exists like 13 years, like more than 13 years. And we started out in Columbia. That was where we actually founded our company. I mean, it’s an all service company, but we all worked in a research centre and we worked with farmers.
Martin
And the goal from the beginning was like helping everyone in the supply chain, especially farmers, because this is where we came from.
Martin
This was like this is our passion, but also like looking at this and how can we help everyone having access to tools that will allow them to make better decisions, more informed decisions.
Martin
This is like all all the bookkeeping, all that. I mean, more like this, all the quality information. They all existed, but in some kind of like paper wise way.
Martin
And it’s like all this the whole supply chain as it works is a lot of information is like gathered along the supply chain, but there is not so much coming back to the farmer. So this is where we actually jump in and say, like we want to build like a traceability and the more like coherent information system that will allow, like, farmers also getting access from information that roasters that didn’t have or like if you can go to the consumer, that’s even better.
Martin
But it’s like it was always like the roaster farmer import export to the whole relationship and like getting all the information and giving everyone access to the information, the information that they need. So it’s not like a free for all, but it’s like everyone should own their own data. A farmer should be in possession of his own data and he should make the choices that to whom do I give my information but also then have the feedback from roasters.
Martin
So for the roasters, it’s great because they see like the trends over the years and it’s like, OK, we found out you roasted your coffee, tastes a little bit over fermented, but that only I notice when roasting you can write to give this feedback back to the farmer. He can change something. And as well, farmers learn to understand why do people like my coffee to whom it’s like the most valuable?
Martin
Because it’s like if you like the coffee less than I do, I should sell it. I mean, to me, because it’s like I would give you a better price. And this is not something like it’s just like what what you like and how it tastes for you.
Martin
So it’s a lot about traceability, quality, raising the bar along the supply chain to help everyone to to produce and make better coffee.
Martin
So that’s that’s the general mission of our company.
We are now in more than 90 countries. Working with farmers, importers, exporters, roasters. Yeah, so, yeah,
Nick
I done my homework more because I was looking at you more from the roasting side, and now that I know you do the whole track and tracing, you’re going to have to come back and we’re going to have to talk about Blockchain. We’re going to have to talk about.
Nick
Oh, yeah. you’ve just made moremore workwork for yourself.
Martin
Yeah But but
Martin
But that’s where we’re coming from.
Martin
But we do a lot of work with roasters, so it’s like there’s a lot of work that we recently and over the years have done some research that we can probably do a whole like talk just with like information from from here and.
Nick
Well, it’s OK because we’re using Google cloud storage so we can just keep going.
Nick
This isn’t like the old days when you were you run out of memory cards, you know.
Nick
So I tell you what looked at the title of this thing, it’s obviously I’m going to give it a catchy headline grabbing kind of title of of artificial intelligence. And that is why we’re here today to talk about it and so I’m so you give it up. Thank you very much for that introduction. And actually now I’m going to go back and and try to learn more about what you guys do, because it sounds really sounds really interesting and a lot of ways as well for other things that I’m working on.
Nick
So but artificial intelligence, we’re going to bring the conversation here. I’m going to geek out a little bit. But the first thing we need to do is actually to say what artificial intelligence is.
Nick
And oh, boy, OK, let’s agree that nobody can actually agree.
Nick
And and and and everybody will have different terms. But there is a basic there’s a basic history to it that we can talk about and we can talk about the terms and define the terms in general concepts, and then we can take it from there. So and I actually meant to be a bit of reading up on it in the last couple of days, but I forgot I might get a few things wrong. And if I do, you’re going to correct me because it’s been a couple of years since I really dealt deep into it.
Nick
But I was reading books, so I was think I gave a book a few years ago at BAFTA with with with some guys there on cyber cyber warfare and cyber defence and artificial intelligence. So I was reading up on it a lot at the time. And there’s basically there’s there’s just in cayou’rese youre interested. There’s two camps in the world. There’s the there’s the dystopian believers and the utopian believers and the utopian believers with people like Ray Kurzweil, who’s the chief.
Nick
I don’t know if he is still the chief data scientist at Google. He used to be. And he gets really involved in life sciences things. He believes that artificial intelligence and what we’re going to be defining as general artificial intelligence is actually going to make life really fantastic for everybody. And we’re going to live these amazing lifestyles where we’re free to be creative and worry about day to day concerns. And then there’s the dystopian view. I read a book by I’m going to get his name probably wrong.
Nick
I think it was Nick Bosworth or something. It was called Superintelligence. And that was the view that as soon as something becomes sentient, then it won’t. The first thing is going to do is protect itself by not letting on that. It’s sentient until it’s got to a position where it’s strong enough that you can’t switch it off and then it’ll it’ll it’ll do whatever it thinks is the right thing to do, which may not be what it’s been programmed to do, because that’s the whole point about being sentient so that these two different these two different things.
Nick
So most people who’ve watched the Terminator movies and everything else who understand we’re going to come on to obviously making sure that cropster isn’t going to your story isn’t going to you’re going to give me some assurances that your software isn’t going to become self-aware and start to murder people.
Nick
But but most people who are thinking about artificial intelligence, they’ve got this thing in mind about the Terminator and that sort of thing.
Nick
But that’s the general AI, what they call generally AI, which we’re a long way away. I think everybody can agree. We’re actually quite a long way away from that. But when we look at AI as it’s actually being used in the world today, these are narrow AI’s. These are AI’s for a specific purpose, such as a self-driving car or an ability for Google and you type cat into Google or show Google a picture of a cat. It finds other cats.
Nick
And these are examples of an AI defined for a specific purpose your voice assistants are using pattern recognition technologies for this. So in that sense, there’s a few terms that are being thrown around that people get very confused about, which are often intelligence, machine learning and neural networks. And basically what we don’t get to too deep into the into the technical part of it. Basically, if you’re not trying to to have a system that is making. Intuitive leaps, that is to say, if you’re not trying to make a system that you teach it how to play chess and then it can say, well, I understand strategy now, so I’m going to go and I’m going to learn how to play go.
Nick
That’s an intuitive and imaginative leap. And and if you’re not trying to do that, then most of what people are actually doing is effectively like a machine learning, which is you’re giving large data sets and you’re creating deductions from those data sets. You’re basically processing that data in many, many different ways and you’re deriving conclusions from that and maybe finding new things that you can do with that particular data in that context and that we call machine learning. And there’s a few other sort of things like and there are a few other sort of areas of it and neural networks is the idea of or is the the method to implement artificial intelligence where you’re trying to basically re-establish a brain’s patterns.
Nick
So a brain works on neural neurons in neural networks. And when we repeat something as a human being, we we do something over and over again. We’re actually strengthening the path right in that neural network. And so people try to recreate that in software to create a neural network. And again, here that people disagree whether that’s the right way or the wrong way to do it. But that’s like the basic concepts you’ve got general AI, which we’re not talking about.
Nick
We’re not doing anything with that. You’ve got Narrow AI, which is artificial intelligence to a specific a specific one thing. Very well. And within that artificial A.I., generally, people are talking about either using software, using neural networks or other approaches, but using large datasets to be able to create inferences from that data or intuitive leaps in some respect that we ourselves as humans would find it hard to do because we’re not processing that kind of quantity of data.
Nick
How much of that did I get, right?
Martin
most of it. Yeah. Well, I mean, I always try to explain my life because, like, there’s this umbrella term of, like, artificial intelligence, which is like just encompasses everything, which is like actually we try to teach machines to create the things that we have as humans or like human like abilities.
Martin
And then you got this like whole like super intelligences.
Martin
And then further down you have like several like stuff like what you call like fields.
Martin
Like one of them is machine learning. And machine learning really tackles using data. I mean, it’s just like Big Data is one of those things that got thrown in. And but there’s like countless others. And then there you have further down more like subfields. I mean, when you mentioned is like neural networks and.
Martin
Yeah. So, so it’s like I’m, I’m also very, very much in line with like the not so dystopian at least I think we will not see that.
Martin
I mean it’s very getting. I think it’s a very far leap from what we do now where we see like this, just like really advances and like and like, OK, we can now do really good recognition of, like, faces, but jumping from this to like there will be a Terminator. It’s like a really I mean, this is this is just like magnitude jump.
Nick
I mean, I still can’t get Alexa that I’m going to say it now. She’s going to do something. I still can’t get my Amazon device to understand what it is I’m really trying to do, you know, and I don’t know how many how much money they’ve thrown at that, but that is a form of narrow AI. And if I if I’m trying to get my voice device just to give me basic instructions, but, you know, it I would say about six times out of ten, it fails.
Nick
So I think we’re still a long way away from from machines taking dominance. Right.
Martin
About what you said before, that if you have enough good data and the huge data set and I mean, it’s like it’s very well prepared and you can use it to learn. I mean, you can beat the best Go player, which is an amazing feat, which like you say, but this goes game. Will not roast coffee. Never, ever in my life, because it’s like it has no understanding of anything else and it’s very this is sometimes hard to understand for people.
Martin
It’s like you’re so like, oh, this the thing must be really good, but it’s like it’s really good and is one thing and it can. Yeah.
Nick
Yeah. We all know people like that. Right. So, so we people who are very, very good at one thing tend to be very bad at everything else. Some of the people I know simply could not walk into a room and say hello to somebody because that would terrify them. And so and so we’re quite a long way off. I think people have talked about 20 years. The only interesting dynamic in this is the and this is something that Ray Kurzweil.
Nick
What I think brought up was you take a look at the acceleration. So it’s the acceleration of technology is logarithmic in the sense that we are getting faster at learning things faster. So it’s almost like a compound effect. And that means that, OK, it’s taken us actually think about it. It wasn’t that long ago that we had the Industrial Revolution right in Victorian ages. And we’re talking about moving from we’re talking about the Luddites and the and the the the the looms for for knitting.
Nick
Oh God my my, my, my my lexicon in that particular field is not great. But the the machines that they made for for knitting or whatever, the looms that they had. And you had the Luddites, which were the people who who were anti the loom technology because they thought know this is crazy advanced stuff and you know, it’s going to be the end of the world. And from there, it’s only been like literally the blink of an eye that we are now walking around with computers in our pocket.
Nick
And so what’s going to happen if you talk about that acceleration, if you multiply that on top, what’s going to happen in the next hundred years? Right. Right.
Martin
But I do think, though, that we as humans, we are we are not really good and like knowing what happens short term and really like long term.
Martin
So it’s like it’s it’s really it’s always been sound looks like it’s so close and it’s like self-driving cars are around the corner and then you feel them and like multiple companies, except for Tesla who still believes they have it next year.
Martin
But no one believes that they could.
Nick
Yep. Look, Elon Musk believes that we’re all living in a simulation.
Martin
Well, yeah, that’s right. And no one can prove him wrong, I guess. But there is an acceleration.
Martin
I mean, there’s definitely things coming. And there’s there’s so much going on in the field that I think we will see big leaps.
Martin
But, yeah, that’s where it is, is when this happens.
Martin
We will we still see that. I don’t actually know. I still hope for self-driving cars. I, I could be a big fan of that, just like I would like to.
Nick
I would be too.
Nick
Let’s bring this back to Cropster and I and I have to ask because there was a study done. There was a study done. You know, the question that’s coming now. Right? There’s a study quite recently that said 40 percent of software companies are using the term artificial intelligence. Find out that they actually not using artificial intelligence, but it just looks good to the investors.
Nick
And and, you know, we know this is true, but that’s because, you know, you look at some of the software out there, people are saying, oh, yeah, you know, my my my, this is a calculator. Now, that calculator used to sell for three pounds. I can now sell it for thirty because it’s an artificial intelligent calculator. But you guys are actually doing the real thing, right? You’re you’re actually you’ve actually built some proper AI into into this software.
Martin
Yeah, that’s right. I mean, it’s like I mean, there’s so many algorithms that you can call, I mean, statistics, right.
Martin
I mean, it’s like this is a field that you can use and then say like, well, it’s kind of artificial intelligence and so you can mingle hurts into it. And we did not put anything on our website for very long time because we as someone who comes from technology, it always seems like stingy, if you like.
Martin
But could we call this like an artificial?
Martin
It doesn’t feel right. But what we have done over the past, like one and a half years, it now feels right that we can actually call it that.
Martin
So we really dug deep and looked at it like millions of records. We really tried to it’s like we started this, the project. And then we we worked with like companies on figuring out what information can be can be gleaned from from what’s what’s stored.
Martin
But what do we give up?
Martin
Like, we all know certain things and have to admit that they have an impact, but roasting is so complex and the interactions and chemicals that are in there for for someone to just, like, understand, it’s it’s very serious.
Nick
Let’s take that back then, because that takes us to a really good spot, which is actually explain what the problem is that we’re trying to solve. I say we because I’m actually not doing much work on this at all for you guys. But OK, what’s the problem that you were trying to solve with with this with software? So imagine that you’re talking to a child because you wouldn’t be that far, that far wrong with me. And you’re explaining the roasting process and and the traditional roasting process before you brought in this functionality.
Nick
And what the challenge is that you personally as a company had that you were then trying to solve.
Lisa
But, yeah, at the time, we did something. So basically one step back. We talked about how technology evolved over the last years. Same with us. Like how we evolved, like when we started 13 years ago, like roasting. When you talk about roasting, it’s basically the process of making the green beans in green green bean into the roasted state.
Lisa
And right. Initially, there was like no technology. So basically you put your beans into the roasting machine and then you put heat into the roasting machine and that heat is green beans on some point. Get to the specific stage of brown in as a roaster, you usually kind of have a knowledge like how brown should it be? How long should be roasted?
Lisa
And at the point where there was no software, you basically tried to take notes about everything that he note down. How much green coffee did that put in? Like how many kilos?
Lisa
Like how long have they been in the specific temperature range? You tried to, like, note down temperatures, for example, like every 30 seconds. You know, noted down how much gas, like how much energy did I put into the roasting machines every 30 seconds to kind of try to understand what is happening in this time frame? It it’s like, let’s say 15 minutes to get your roast to this desired end state. And then after that, it basically took out the coffee and then you you tasted it and then you found out you either liked it or you didn’t like it and you need to change something.
Lisa
So you repeated this process so long until you got to a point that you’re happy with your product and then you try to find out how you can repeat that. And ideally, you have made so many notes that you know exactly what you should do again.
Lisa
And that is where we started with Crospter, where we then developed this profile, a system where we track every second of what’s going on into your roasting machine.
Lisa
So you get an out a roasting curve where you see exactly how much temperature, how much gas and all of that many out of variables that you can fix so that when you got to this desired point, you can immediately reproduce it. That is like how we started.
Lisa
And then even already when we started, we developed this this metric that’s very commonly known in the roasting rate of rise (ROR). And you can think about it like seeing the speed of the car, like how fast are you, like, increasing your speed? For example, if you see an increasing like 80 kilometers per hour, you kind of know how fast you’re going. And the same is with the rate of price. It basically tells you how fast you’re roasting machines heating up.
Lisa
And and with this, we select like the first approach in, like, helping the roasters to see if they keep it like this.
Lisa
Despite that, they kind of could calculate in your hand, OK, but keep the speed. Most likely I’m getting to this important post, develop first crack in this specific time frame so it can kind of, with this metric, calculate in your head when you’re getting to specific events and then make adjustments based on that.
Lisa
So basically, what Cropster always tries to solve is like helping you to see how your evolves, get into basically your desired quality, and then making it repeatable because making it repeatable is also like one of the biggest problems in roasting.
Lisa
So we always wanted to do that. We always wanted to help with that. But now, again, as in technology, we saw that we actually can do more. We cannot just like show the rate of rise or show what’s going on at the moment. But we could, with all the knowledge we have and all the machine learning algorithm we created, actually show what’s happening already in two minutes. So when you know, as opposed to of what is happening in two minutes, you can actually react right now.
Lisa
So which is like actually pretty good. And one of our customers framed it like that until the artificial intelligence was there, they basically reacted to a roast more on defense.
Lisa
So they always kind of be like, how can I keep that roast where I want it to go? And now with the artificial intelligence that can better go on offense with their roast and be like, this is where I want you to go so
Nick
that this is a really, really good explanation.
Nick
Thank you. And it sounds to me like and I and I was reading up on this, I’ve been I know very little, very little about the roasting business. So I’m going to be that that guy that asks all the really stupid questions. One of the somebody told me once, one of your really great strong points is you’re not afraid to make a complete fool of yourself. So so I have no problems about asking the really dumb questions. And maybe they’ll be other people out there who watch this will be glad that I did.
Nick
But it seemed to me as I was reading up on the IRA, while the rate of rise that that actually that predictive proactive process would be, it is really lending itself very strongly to to the benefits of of of AI and big data and being able to make those predictions. Is that is that right, Martin?
Martin
I mean, the thing that’s like it is a time event and time events are something that are very good for I mean, as a as a technology, we understand. So it’s like there’s a lot of predictions. There’s like over time. I mean, that’s what Lisa said the fact that the simple fact of life you just like know where you are in two minutes, you can draw a straight line, right. And say, like, OK, I’m there in two minutes.
Martin
That will not happen. Because what we said before was the complexity of what is going on.
Martin
If I make a change in my system, the whole system changes, just like you are now, like trying to incorporate all those things that are already there at that point in time for so much things have happened. So but it is not some some simple math in your head. I mean, that’s what you can do as an abstraction.
Martin
But if it grows differently or if you take less speed in the beginning or speed, it changes the whole variable level.
Martin
And with artificial intelligence, we we were able to to learn actually what what the influences are and protect them correctly into the future.
Martin
So this is where I mean, it’s a complex system. It’s not the completely sealed system, a roasting machine. And depending on the machine, it depends. But but those are also variables that they can can put in there.
Martin
So it’s like depending on the type of machine you’re using, your development will be different.
Martin
And also, depending on by that, the speed that you took before is really different. So our system can use that and can give you an accurate prediction no matter which machine you’re using.
Nick
So, Lisa, can I ask, is it is that do you use the software when you’re doing the sample roasts? So here comes one of my stupid questions. OK, so I guess you do the sample roast first. You get a sample roaster and you’re basically trying to create a profile. And the profile for the beans that you’ve got, you’re making an assumption that all the other beans that batch are going to be of the similar profile. And then you can say, great, got the profile nailed.
Nick
Now I’m going to roast all my beans in the future like that.
Nick
So it’s the is the place is the place for the software, for the cropster to software during the sample roasting, or do you also use it, you know, when you’re roasting every batch?
Lisa
So actually that’s a very good question. And the software cropster is for both. So you use it for sample roasting, but also for product roasting. It is the same software, but it has a little bit of different like functionality you can use. So in production production, as you said, you already have kind of figured out what the program is and then you can basically put your profile into a background and try to follow the profile. If you have not done more or less semi automatic machine, if you have a semi automatic machine, we even have functionality that with cropster.
Lisa
So you can actually replay everything you did, like all the gas changes in a specific time frame. So it just basically means, okay, I figured this out. Now, please, cropster help me to reproduce exactly what I did. So that is like the production side of things. And on sample roasting, actually the roaster prediction even even helps you even more like on production. It’s like a case of trying to stick to my curve, basically, but in fact, roasting.
Lisa
You need to imagine that if you said you have a new coffee, right. So there’s a new sample. You kind of know from experience.
Lisa
Okay, maybe how how does a European coffee behave or how does a coffee behave so you can put some experience in you maybe even try to start with a similar profile that you had in a different coffee from last year, but you more or less like start from a blank slate. So basically, when you then start roasting with cropster, you imagine more like a blank slate.
Lisa
You have this prediction that shows you how this like fresh product never moves before, most likely will behave. So you can actually you will actually left that choice to roast less sample batches to roast to get to the desired outcome. Because if you know an example, roast immediately, like, OK, I want usually to get to a first crack around with eight minutes or have like X percentage of development time. And, you know, that already has roasted before. You have the prediction and you can adjust immediately without being like after the roast, like actually four seconds later than I expected.
Lisa
And it goes again.
Lisa
Again, so with the with the prediction next year, you see how a new coffee, you never know before will behave because we gathered this knowledge and this is actually quite.
Nick
Fantastic. So so are you using like a Martin? Are you using like a hive learning model where you’re taking presumably the feedback from the software? If you’ve got I don’t know, let’s say you got a thousand roasters all over the world roasting different beans and doing different things in different ways.
Nick
Is that data being sent back to a central place where you’re able to learn from all of those different roasters and then be able to apply that to another roaster? So, for example, if a roaster in New York is is is is doing something a certain way and you learn something from that, from what he’s doing, is that lesson then applied to the algorithm that goes out to everybody?
Martin
Yeah.
Martin
Let me just take one one step back on that. So because, I mean, it’s very important to say like that the data that belongs to customers are customer data, not our data. And when we started out, we started this set before the data project.
Martin
So where we have like a of like roasters that we work with, that we are really going and like with their permission and like this all that’s going to like research and doing research on the data and saying like, OK, but what can we find? What do they know? Blending like the the human part was like the data part and then going back and forth on those things.
Martin
And if there is something that that shows to have value for the community, like for for the bigger like that is applicable not only for a small part, because this is also important that we can always like give this to everyone and not just like the people that can really afford, like, very specialized, like hardware.
Martin
Then we try to to roll it out to more people.
Martin
And in this case, we did this we did this case studies with like our roasters. And now the system completely makes that data anonymous.
Martin
So it’s like we have this data, but we do not know anything. It’s like that.
Martin
So the whole system, it does not know anything about any particular it wouldn’t know that you are from New York.
Martin
It wouldn’t know that, like you’d use your like this and this person.
Martin
So it just for it it is a set of like the features that we have and that we train it on. And now that allows us to like add new datasets and keep training research, I mean, there is what it means is like it’s not 100 percent tailored to you. So it’s like only what you are doing. But I mean, we all we all have unique styles, but they are all in at a general level, very similar.
Martin
I mean, it’s like if you always look at the bigger the bigger topics, the bigger the high level, there is always the chance that it is applicable to something else. So, yes, in the sense we get information back into our system, but it’s like detached from all the the, like, general information. It’s just like it’s really a blob of paper. It learns from that.
Martin
And then this can
Nick
You don’t need to know, like location data or it’s John Smith that’s not relevant to you. It’s different. Funnily enough, some years ago, someone did a test on some antivirus. So, you know, the antivirus that you’re running, your PCs and there’s lots of different companies out there like McAfee on or trend or whatever. And somebody went and did a test and they tested. I think about 20 different antivirus run things on the computer to see what they did.
Nick
And they ran what’s called a network sniffer to see what track that they were sending back to the central. And they were sending your documents back. Right. So your personal data was being exfiltrated back to to data centers based in the US, totally in breach. Of all European regulations, in order for them to to to do the cloud analysis now for the antivirus companies. That makes commercial sense, it makes no it makes no branding or reputational sense, but it makes commercial sense as long as you don’t get caught.
Nick
But for for for for Cropster and what you’re doing, that’s irrelevant data. So I can imagine you can completely anonymize the data at the point of collection, anonymize it, and you’re just sending up like raw data that could come from anywhere from anybody. But you’re able to learn from that, because I think it is quite in looking all these AI models, the bigger the data set, the better the information and then that benefits everybody. Right.
Martin
So you can build a more comprehensive model. And you also have the ability to like if there is a new machine, like someone coming to the market and it is like you can incorporate like more sync. Sounds like if you were the first person who bought it and you’re the only one in your vicinity that has like a certain type of roaster or there’s not so many years, your you don’t have that information. But if you look globally, there’s more people.
Martin
And so you can harness like information from others that also have similar experiences.
Nick
So so so does this does this software work on every type of roaster or is it is it restricted to certain types of roasters?
Lisa
Well, we have a quite substantial list of roasters. I mean, we always say more or less we can connect to nearly any roaster, but we also work very closely with the roaster machine manufacturers. For example, the customer comes and has a new machine we have never heard before, which by now rarely happens. We try to figure out the way how to connect to them because we can either connect directly, like with the ethernet cable, like on the more modern machines we can connect directly, but also to all the machines.
Lisa
We have a specific connector that they can really connect to any machine. So you just need a computer, download our software and you can get started basically.
Nick
So if I have a very old or non sophisticated roaster, actually this is a way for me to substantially increase the value, the benefits or the capabilities of what I can do with that roaster without having to to buy a new roaster.
Lisa
Exactly.
Lisa
I mean, the roasting machine can last 20 years, right?
Lisa
We have customers that have roasting machines that are 15 years, 20 years old. And you don’t want to buy any roasting machine if it’s fully workable.
Lisa
Just for new software, if you can just edit on that is like one of our biggest.
Lisa
As Martin said before, the thing was, was that was most important to us is that we bring this machine learning algorithm to everybody that is using it doesn’t matter if you have a new machine on machine. Doesn’t matter. We will cover all of them.
Nick
So so I sorry to pin you down on it, but or tend to fly left field. But how many how many installations do you have of the software out there of them. And can you say or is that, is that commercial confidential.
Martin
And that’s a bit confidential. But but it’s there
Nick
I have to ask.
OK, so in that case, what’s the what’s the plan going forward? So you’ve got this how are you going to take this forward? I mean, you everyone’s got ideas. This sounds like a great idea to me. Sounds great. I wish I had a Roster so I could try it out. I have to go find something. But you have to tell me somebody locally in my area, I can go down and have a look at it.
Nick
No, seriously, I’d really love to do that. But what what are your plans for the future then?
Lisa
Well, I mean, Martin, the before I mean, there’s a lot of work already done. We started, I would say not one and a half years ago. We have no actual data science team that is just working on these kind of projects where we look into different ideas, different different ideas that we get from the people, from the data project where people are like, hey, guys, you must be able to do that or you must be able to do that.
Lisa
And we like actually good idea. Right. So our customers are actually also quite inspiring because they have a lot of ideas for us to look into.
Lisa
But also besides that’s our design team is working continuously, like every day on improvements and ideas. And for us, this was the first step. Like I can say, we’re really working on more projects, but this was the first step because first of all, like like we have a large number of customers that are using our roasting software.
Lisa
And it’s like this closed system where we can where we know which variables we can use, for example, like gas changes or machine types or these kind of things.
Lisa
So it was for us the logical first step to go because we can bring a benefit to many people. But at the moment we adding more projects for the roasting process directly because there are more things you can potentially predict than just the bean temperature, but also evaluating other things, because we have this online platform where we have like a lot of reporting on the inventory reports production report and these things., we are also looking into things where we can help there with machine learning and artificial intelligence to help you maybe predict something else in the future so that we can say for us cropster the start, a new era of features.
Martin
And just to quickly add on to that, I mean, these are covered this very well. I mean, our our goal with this is to your initial comments like that.
Martin
Generally, I, I mean, our our goal very much is to help the people operating.
Martin
It’s not like replacing. I’m a firm believer in it. Like there is a lot of knowledge that is very hard to capture by a machine.
Martin
But we do also a lot of unnecessary work. I like standing next to the roaster just to watch the curve go up is maybe not the best time that I can spend.
Martin
So it’s like we really want to approach this from the where can we take away just like this, like grunt work and replace it with something else that is like more suitable to the human. Like it’s like something some creative parts. I mean, it’s like there is goes a little thing to like figuring out how you roasted and but not necessarily only watching the curve. Right.
Nick
Right. So you’ll do well in in some countries and in some ex-communist countries where I was at recently and had to pull up and give my parking ticket to one person who I would he would then take my money and and then give me the ticket back, the same ticket back. I would then drive two feet and give the same ticket to another person who would then open the gate for me. And so, yeah, it’s a productivity improvement. Right.
Nick
So the idea is to get away from having lots of people just for the sake of having lots of people. You don’t you don’t need to be standing around watching the roaster. It’s it’s a case of having this people actually move up the value chain do a lot.
Martin
And also you can do so much. It’s not like getting I mean, not not having the people, but while you’re there, you can already prepare the next batch. I mean, most people do, right, that they run around.
Martin
I mean, it’s like it’s it’s amazing how how busy like a roasting floor is and how much those people have to do.
Martin
And in the same incident, we asked them to calculate like, well, year and a half.
Martin
I mean, it’s like let’s let the computers do what computers are good at, which is just like working with numbers.
Martin
But the other things free up the time to just like concentrate on this task and not just like be like constantly spinning and multitasking and
Nick
fantastic.
Nick
Well, listen, wblockchaine’re coming to the end of the session, but you have you’ve been fantastic. I really I feel like I’ve learned a lot as well. But but I’m going to have to come back and talk to you about about block chain and track and trace, because it’s another little thing of mine where I’ve been I’ve been involved in in and track and trace and track and trace applications and actually mostly on the in a different industry. But be fascinated to see what you guys are doing in that, because I’ve been reading about it for a long time, but I haven’t seen a lot of of it in action on the ground, not just in coffee, but also in cocoa.
Nick
So we cover the coca market as well. And traceability is hugely important. We actually just I’m going to give you like a little that’s a little a little punt for myself here. But we just released a a research paper on on the problems in Cote d’Ivoire with with deforestation and the problem. The Cote d’Ivoire has got to deforestation for the cocoa. It’s all it’s really mostly for for cocoa planting, cocoa farming that they’re cutting down the forests. But the problem is, is not as widespread in Ghana.
Nick
Why the difference? Because the track and trace in Ghana, they have a GPS system, some very, very basic not blockchain, nothing. It’s just GPS. And you can actually show the providences all the way through to the farm. You can show the provenance of where this cocoa beans come from, but they haven’t got it in Cote d’Ivoire. And the same thing can be in the coffee industry. Not so much from a deforestation issue, but more from a knowing actually what you’re buying and how much you’re paying for.
Nick
It is is really important. So I had no idea, actually, that you were involved in that. So I might I might call upon you again.
Martin
Yeah, maybe a real happy to talk again. It was very nice. You’ve been very generous with your time and I really appreciate it.
Thanks so much. It was great talking to you. Martin and Lisa
