Subscribe
Share
Share
Embed
This podcast is dedicated to changing the automotive industry for the better, one conversation at a time.
Whether you're a technician, vendor, business owner, or car enthusiast, we hope to inspire you to improve for your customers, your careers, your businesses, and your families.
Lucas Underwood [00:00:02]:
Fellas, introduce yourselves.
Henrik Rosseland [00:00:05]:
You start.
Christoffer Weber [00:00:05]:
Okay. My name is Christopher Weber. I am the CEO of a company called Wiretronic in Sweden.
Lucas Underwood [00:00:13]:
Okay.
Henrik Rosseland [00:00:14]:
And my name is Henrik Rossland and I'm sales responsible for the company.
Lucas Underwood [00:00:18]:
Okay. Last night, David and I were walking through the tool show here at etiquette, and I see a box of connectors. And connectors have always, you know, I am a tech at heart and so connectors have always had a special place in my heart. And you pick up that connector and you show me with your cell phone that you can scan that connector with your cell phone, and you can show me the part number and you can show me the wire number and you can show me the tool number. Tell me about this product, because it is insane.
Christoffer Weber [00:00:54]:
Well, actually we started off as producing breakout boxes for, in the beginning was only Volvo, okay. But we've done quite a few for Ferrari, McLaren sports car manufacturers.
Lucas Underwood [00:01:07]:
Well, now I feel kind of bad about talking bad about Volvo at lunch.
David Roman [00:01:12]:
I'm a Volvo fan, by the way. I love, I do, I am, I love Volvo cars. Everybody hates them. I love the crap out of them. They're so cool.
Lucas Underwood [00:01:20]:
That's because you're not the one having to work on them.
David Roman [00:01:22]:
I know, I let my techs handle that.
Henrik Rosseland [00:01:24]:
It's a swedish car.
Christoffer Weber [00:01:28]:
Well, and the first couple of years we only supply them with harnesses of those kind. But over the years, we turn out to develop also prototype harness for the new camming cars. And in that process we have one guy who is really an expert of connectors in the company. And we said we need something to help him out identifying connectors. And I myself have been very interested in AI for many, many years. And lately I've studied quite a lot of that. And we said that, why don't we try to develop something that we can use ourselves? And this turned out to, we used the local university to develop a prototype almost like this Google lens, but for connectors. So we trained them up on the connectors that we have ourselves.
Christoffer Weber [00:02:21]:
The problem then was that the precision was quite low, so we didn't get when the result was not that fantastic. And then we said, it looks good, but. So we employed a few guys just out of the university and they developed a system for us together with us. And now we have like ten people working on this. And that is actually the result. We try to do identification and do something, sort out the low hanging fruit for AI within automotive, right?
Lucas Underwood [00:02:57]:
And we just heard a lot about AI and we heard kind of the thought process behind AI when you hear a presentation like that, what is your first thought? Because we had some additional conversations at lunch about where we're headed. What is your first thought when you hear a presentation like you heard at lunch?
Christoffer Weber [00:03:15]:
Well, I think there's a lot of presentations from people who have identified AI as something cool and interesting to be in. And of course, it's like a revolution. But having been involved with it quite deeply for some time, a lot of the things will take much longer time than people who tell us about it and want to make people kind of scared about it. It's a feeling that people who talk about it, that don't know about it, want to make people kind of react about it. So I think there's some things that looks kind of dangerous or annoying or can be talk about that it will take people's jobs, but I think it's quite reverse. I think it will help a lot of people with it. So I'm not really having learned about it quite a lot. I don't feel scared about it at all.
Christoffer Weber [00:04:07]:
I think it's a lot of boring stuff will be sorted out by it.
Lucas Underwood [00:04:11]:
Right, right.
Christoffer Weber [00:04:12]:
In the same way, like Internet came like 2025 years ago, people said, yes, it's gonna be just a small bubble, and then turn and turn away. And now everyone just is pervasive.
Lucas Underwood [00:04:22]:
We live on it, right? It is life as you know it right now. And. Or you say, no, I want to.
David Roman [00:04:30]:
Know about the little connector. So you take a picture of it. Are you guys also producing the different connectors?
Christoffer Weber [00:04:40]:
No, we actually have a very close relationship with the connector suppliers, like Tyco Sumitumo je we were very close to in the pools, when we produce a harness, the OEM tells us what to do and how to create it. So we purchased it from the supplier. So that's why we know them quite well. We do produce extraction tools ourselves. We used to buy them a lot, but we said that a mechanic always have to be very delicate when they try to extract the cable out of the connector. So we said, why don't we develop something where actually it's very. The extraction tool fits perfectly to the connector, and then just don't have to be the expert of all kind of wiggling and getting things right. And that's why we did this.
Christoffer Weber [00:05:27]:
And we did the first one we did for Aston Martin, and they got a tool together with a set of connectors for their chassis harness. And then we. I think it was last year we got an order from Volvo to produce extraction tools for them. And we did this. We had to buy our own special machine to do this. This machine is very precise. It's down to 10 μm precision in how it cuts out the tools.
Lucas Underwood [00:05:57]:
That's pretty impressive.
David Roman [00:06:00]:
The tools have always been kind of just generic.
Lucas Underwood [00:06:04]:
Yeah.
Christoffer Weber [00:06:07]:
We have to develop these specifically for the connector. So we have a few engineers to really work on getting it precise for the connector. So they actually have to have microscopes and 3d models and really to see how they get the spring out for.
Lucas Underwood [00:06:21]:
Their terminal and then, so it'll actually push. You know, I had an argument with one of my technicians the other day, and he has been in the Audi space for a long time and in the new car dealership and we had this argument and he says connectors don't fail. And I'm like, they do. And he's like, no, the plastic fails, but the pin itself doesn't fail. I said, no. I mean, like I'm telling you, I've watched, I've had cars I can go over and I can touch the connector.
Christoffer Weber [00:06:45]:
Yeah.
Lucas Underwood [00:06:46]:
And he said, well, but, but in the dealership, the protocol is we never replace the connector. If you want to replace something, you replace the whole harness. You don't just replace the connector. That seems asinine to me.
David Roman [00:06:58]:
Well, you're in a dealership, it's a lot of cost.
Christoffer Weber [00:07:01]:
I mean, it's really costly to rip out all the things and put it in. I think this is kind of the logic behind what we're trying to do. We try to make sure that we have tools for the engine, for the engineer or technician, so he can repair it to a professional standard. And we also have a partner with another company who do crimp tools so that we can. They actually measure how they crimp during the crimp, so it becomes perfect. So it really clogs around the.
David Roman [00:07:29]:
Do you sell that tool as well?
Christoffer Weber [00:07:31]:
Yeah, we sell it as well, but we mostly sell things to the OEM. So the OEM buys the kit and we put a kit together for the OEM. So for Aston Martin, they asked us to do something for the chassis harness and we did that. And for Volvo, we took their connectors and did it. And the same thing with this machine learning system or the AI system is that we take the connectors and we have produced our own scanning robot. And so we can get pictures from all angles of the connector, or any other component for that matter, and then use that, do a lot of special processing around that. The special processing is quite important because in the beginning there were a lot of problems with light. It was too yellow or too white or too.
Christoffer Weber [00:08:20]:
The kind of difference between being inside and outside. So we had, based on that, became kind of poor recognition. But now we've done a lot of things really to make sure that we get high precision. And that's why we need to be in control over the whole, as they say, data set generation, which is in the AI terms, to control how to generate the things that you train, the machine learning and so on.
Lucas Underwood [00:08:46]:
What type of coverage do we see with this? How many brands would it detect that connector for?
Christoffer Weber [00:08:51]:
It's difficult because we have concentrated on, we have done for Volvo, I think we did also recently for some McLaren and some others. But we usually make a deal with the OEM to take care of theirs. The difficulty to do that for independent is that there's such a big flora of a variety. A lot of the connectors are the same. So they're very similar between the brands, but you never know. And they're also added on this as a big issue with some of them are restricted. Some of the connectors are only restricted. So, like Mercedes don't allow anyone else to use a connector from coastal or wherever.
Christoffer Weber [00:09:31]:
So that means that it's kind of difficult to make something for the independent unless you go abroad and work with a wholesaler who supplies you with the connectors and scans them. And we're talking to a few to do that as well.
Lucas Underwood [00:09:47]:
Well, that's what I was sitting here thinking about. Is it connector experts, the website? And so they have it all cataloged and all that data is already there, but they don't.
David Roman [00:09:58]:
It's not the same, though. Fine. Pigtails is the same thing, too. The price is the price. The prices are a little high. So when you buy their connectors, it's $80 plus shipping, plus the little wire kit or whatever. So it ends up being about 100, $3140 for a two wire connector that there is no way to get other than buy the entire harness because it's a Toyota and it's on a 2006 whatever, and the old one melted, so. But what they don't have is a way to identify them.
David Roman [00:10:31]:
You have to go, okay, is it a one pin, two pin, three pin, male, female? What color is it? And then you're going by picture.
Christoffer Weber [00:10:38]:
Yeah.
David Roman [00:10:38]:
And you're like, yeah, that looks right.
Lucas Underwood [00:10:42]:
And then it was a while back that Ford went to, I want to say they like reverted and went to a paper catalog like you call there.
David Roman [00:10:49]:
Was them that always been like that.
Lucas Underwood [00:10:51]:
Yeah, I know, but I want to.
David Roman [00:10:53]:
Say there was looking for.
Lucas Underwood [00:10:55]:
Wasn't there a change? They, they had, it was like they had the connectors on repair link and then all of a sudden they weren't on there anymore and you couldn't find them or something to that effect.
David Roman [00:11:04]:
If they sell the whole harness, it's, it's cataloged the whole harness. If it's an individual pigtail, you have to go into their motorcraft, which is their aftermarket stuff, and then they'll have every single one of their available pigtails cataloged in a PDF. And you just have to scroll through. But it's the same thing. One pin, two pin, three pin, four pin, whatever, male, female. And then you start flipping through looking for one and they could be two and same connector, but you can't tell from the picture. One's, you know, 7 mm wide and the other one is 5 mm wide. And you're going to order one thinking it's the right one.
David Roman [00:11:37]:
And then it shows up it's not the right one.
Christoffer Weber [00:11:38]:
Because when I remember, we've been approaching a lot of the german oems as well and tortured them. And they used to say, yeah, all our stuff are sorted out. We sell them small pigtails or harnesses and that's how it works. Then I went to talk to the headquarters in each, for each of them. And then I went to a local dealer and I said, yeah, harnesses and connectors, such a mess. We never get the right one. We get wrong and we don't never find them. And I think this is the kind of, the thing for us is that if you take a picture of something and immediately get it, it takes seconds or a fraction of seconds.
Christoffer Weber [00:12:17]:
But to search for something in the catalog, it takes hours or sometimes days. If you don't find it, it's not even the searching.
David Roman [00:12:25]:
The searching is fine. You can put up with the searching, but to get the wrong one, have it shipped in two days later, cars down, then you've got the wrong one. Now you start to search back over and you have to cross your fingers. I think Volkswagen, from my experience, Volkswagen has been the only one that will you call them and say, hey, I got an oil pressure switch. It's. It's failed. It leaked oil. It runs up the wire.
David Roman [00:12:47]:
Can you get me one? And they'll sell you the empty cavity, the, the individual wires and then the little, the little weather pack thing. But what you don't end up with is the. Is the tool. So you end up using a universal one fidget. Yeah, yeah.
Christoffer Weber [00:13:01]:
And then you can work on the plastic, starts getting bent or misshaped, and then you put it in a terminal and this backs out again. So, yeah, you're never going to.
Lucas Underwood [00:13:10]:
Yeah, well, and, and for us, like, Volvo connectors, that's a place we've had a ton of issues because we'll order the connector. And they're like, well, it has to come from Sweden. It's going to be three weeks and then it shows up and it's the wrong freaking connector. And you're like, but what am I. Like, what am I supposed to do with this? Well, they're like. And they. I should not say this, but I'm going to. Our Volvo dealer is not exactly the pristine example of a parts department.
David Roman [00:13:36]:
Okay, it's a guy, he's on lunch. Our guy will tell us, hey, I'm on lunch and we'll have to call you back.
Christoffer Weber [00:13:46]:
We bring the information back and the telegeist in Gothenburg about that for sure.
Lucas Underwood [00:13:51]:
Well, and that's the thing that we're up against is a. We're trying to blend that consumer satisfaction with us. Right. And at the end of the day, they don't care whose fault it is. The part is not right. They don't care whose fault it is. The parts not here. It's me.
Lucas Underwood [00:14:08]:
Right. You know, I look at the people in our circles and I say, these guys run the top 1% of shops in the nation. And if I can't get it right, and they can't get it right, then the system's broken on the other side. Right. But, but in the same respect, the people in that parts department, they don't.
Christoffer Weber [00:14:27]:
It doesn't matter to them for sure. The people we meet, they really love to make sure everyone, all workshops can repair the vehicles because they really want to make sure that no one complains about them.
Lucas Underwood [00:14:39]:
Yeah.
Christoffer Weber [00:14:39]:
They need to be repaired quickly and efficiently, for sure.
Lucas Underwood [00:14:43]:
And that's the. We've had that conversation about Volvo before, is that in the states they can be so hard to repair that consumers are like, I don't know, I don't want one. Right. And the ability to get parts and get things timely. And I know you may love them. And I'm not saying I dislike Volvo.
David Roman [00:15:01]:
I'm just saying it's the same five things that break.
Lucas Underwood [00:15:05]:
What, all the, all the insulation falls off the wire.
David Roman [00:15:08]:
No, that's old Volvo. That's like nineties Volvo. The newer ones are much better. They are. They're much better.
Lucas Underwood [00:15:16]:
They really are. They really are. But you're right, the crankcase ventilation, the Ultra.
David Roman [00:15:21]:
Yeah, all of that garbage needs to be replaced. The car comes in with coat, all of it gets ripped out and you put it all new. Yeah, because if you. That one hose that you didn't, you're like, that hose is probably okay. No, it's not.
Lucas Underwood [00:15:33]:
It just split from top to all the way to the bottom.
David Roman [00:15:35]:
No, it's been, it's. Yeah, it now is the one causing the code. And the car will leave for two weeks and then come back with a code. So is the, is the software available to anybody or. How's that work?
Christoffer Weber [00:15:47]:
No, actually we, it will, it will be. We actually, we kind of do it like a white label. So this means, okay, we develop, we have an app that is connected to backend system. And the backend is where the AI inference takes place. So we take a picture with the app and then send it to a system where the actual recognition happens. And then they send back the pod number and also the information about the pod. So which terminal to use? Document all the documentation and we can put anything in there. So if anyone wants to like schematics, we can bring that in as well.
Christoffer Weber [00:16:30]:
But the main thing is that it's not going to be wiretronic or wire vision, as we call this product. It's going to be the OEM or the wholesalers brand on the app.
Lucas Underwood [00:16:43]:
Okay.
David Roman [00:16:44]:
So other than the OES though, are you talking to anybody else?
Christoffer Weber [00:16:47]:
Yeah, it's like wholesalers who would be selling connectors or who would be selling spare parts like SMP.
David Roman [00:16:54]:
Standard motor products would be.
Christoffer Weber [00:16:56]:
We just installed a scanning robot at a big wholesale outside of Milan where they actually go to span, I think, six to 18 months now to scan every part. They have three model on, 3d model first for the catalogs. So they can have three models, all catalogs, but then also looking to train identification system on that.
Lucas Underwood [00:17:24]:
So the, the thing that spoke most to me about this technology is the chainsaw chain, because you showed me how we can scan a chainsaw chain.
Henrik Rosseland [00:17:33]:
Actually, we talking about automotive now all the time. But we're going for other customers, other outside automotive as well.
Lucas Underwood [00:17:41]:
It's really interesting because you were showing me that it could even tell down to the link.
Henrik Rosseland [00:17:45]:
Yeah.
Lucas Underwood [00:17:46]:
Right. Hey. To ensure that this is the right chain. This is the right chain that revolutionizes so much about not just the automotive industry. Right. It is impressive.
Henrik Rosseland [00:17:58]:
Yeah. If we can identify a chain, I mean, if this is old or rusty or something, and we get the right one, this is very difficult. So I think it's more difficult than connectors.
Lucas Underwood [00:18:09]:
Yes.
Henrik Rosseland [00:18:10]:
So that's what we.
Christoffer Weber [00:18:11]:
Connectors are colorful and they have a lot of visible signs that they look different.
David Roman [00:18:17]:
Right.
Henrik Rosseland [00:18:18]:
So this, this knowledge is the. Our IP or the value of us, right, to speak. Yeah.
Lucas Underwood [00:18:24]:
Well, I mean, I'm just thinking of all the different things that you could use that in.
David Roman [00:18:28]:
Could. Yeah. Could it be expanded out into under hood components? Yeah. Hoses?
Christoffer Weber [00:18:34]:
Yeah, yeah, sure.
Henrik Rosseland [00:18:35]:
Any part.
David Roman [00:18:37]:
Are you working on that? We are looking for customers are a nightmare, by the way.
Henrik Rosseland [00:18:41]:
We're looking for customers, all right. With a lot of spare parts, of.
Christoffer Weber [00:18:46]:
Course, when it comes to whole service issues, as long as it's distinguishable for us as humans to see if there's a difference in the way things look, AI can do a slightly better job in the identification.
Lucas Underwood [00:18:59]:
So we deal with a lot of. You take a BMW, right, at lunch, we were talking about the fact that BMW, when you replace a valve cover gasket, you're replacing the valve cover. And so I brought up the fact that there's plastic hoses all around the valve cover. And so a lot of times we'll pull them off and they'll snap. And, and we went through an ordeal with BMW a couple weeks back where we called and we said, hey, we need this part. And the parts guy said, I'm pretty sure, I think it's this part. And so they send us that part and it's like, nah, that's not it. Okay, well, here, I'm going to try this one.
Lucas Underwood [00:19:30]:
I'm going to send you this one. It could be it, too. And so we're looking, you know, it's on a screen and it's this tiny little picture. And you can zoom in, but the way it's laying on the. On the picture, it doesn't look like the part that you've got in your hands. And you can't see all the bins and you can't see the intricacies.
Henrik Rosseland [00:19:47]:
You don't see the part number on that or.
Lucas Underwood [00:19:50]:
No, no, it's. I wish that stickers don't. And so, like, the thing about a hose is, is right, if you pull it up in an image for us, well, in that image, you see it's got a bend here and it goes over here and it turns and it does this. But on that picture, it's flat. Right? So you can't see it. Or they'll use diagrams that it's just a draw.
David Roman [00:20:14]:
I'll give you one better. We had a Toyota with an EVap hose. And the part number we ordered a part number comes in. It was the wrong one. We order. I'm looking at the catalog, and we went into partsook, and we had the whole breakdown. And I'm like, I'm looking at it going, okay, there's only two hoses. EVap hoses that connect to this fuel tank, and we ordered this one.
David Roman [00:20:38]:
It's wrong. This one has to be it. There's nothing else. You take that part number, you google search it because you want to look at an image. It is a two connection hose. And what I have on the car is three connections. It has a split. So I call the dealer.
David Roman [00:20:52]:
I'm like, hey, I've got this part number. Is it matching to this car? Yeah, it's correct. To the car. Great. Why is it showing that it only has two? He's like, yeah, it looks like it has two. Okay, the one I have on the car has three. I don't know what to tell you. We're just gonna have to order it and see.
David Roman [00:21:06]:
We ordered it. It showed up. It had three. All of the pictures were wrong. All of them? All of them. The pictures were wrong. The descriptions were wrong. His catalog, my catalog, everybody's catalog showed two connections.
David Roman [00:21:20]:
It was a three connection hose. That's where we're coming in, and we're gonna 3d scan all these suckers.
Christoffer Weber [00:21:26]:
It's quite a typical issue. Another company asked us about, like, break discs and said they always send three break discs because they take one that they think it is, and then they take one on the other side and the other side of size or whatever it means, and then send it. And then you have to send them back. The wrong ones back. Yes. Otherwise, we'll be waste of time. And wasting the time when you have to.
David Roman [00:21:52]:
German manufacturer, they always have to put three different sizes of rotors. So 200 and 8300 and 2350.
Lucas Underwood [00:21:59]:
You're like, and I don't know if you fellas have figured this out yet. When you throw the metric system on a bunch of american rednecks, things go downhill quick. Okay, it's bad.
David Roman [00:22:09]:
But you're measuring it, and you're like. You're like, okay, my options. 280 or 298. You measure it, and you're like, this is 287. Well, which one do I choose?
Lucas Underwood [00:22:23]:
I'm not kidding.
David Roman [00:22:24]:
That happens. You're like, oh, you're measuring it in the slightly wrong spot, or there's rust.
Christoffer Weber [00:22:33]:
That's also what we kind of try to combine. I mean, with a chainsaw, we combine that with counting. So we take a picture and count the links, right?
David Roman [00:22:43]:
Yeah.
Christoffer Weber [00:22:43]:
The links and the other part with disks, we actually try to find a way where we actually can have a scalar behind. So the scalars we've been using are similar, almost like QR codes or could even have, what do you call them? A measurement stick.
David Roman [00:23:05]:
Yeah, a ruler.
Christoffer Weber [00:23:06]:
Exactly. To put behind. And then we use that to make sure that you can get it right. Because using tools in the telephones, like ar tools, they usually are not so precise. And I mean that you need to be precise, otherwise it will be the wrong one and then the wrong one is.
Lucas Underwood [00:23:25]:
So let's say a manufacturer comes on with you today. And in my mind I couldn't see why they wouldn't just instantly say, hey, we have to have this technology. Let's go. Right?
Christoffer Weber [00:23:35]:
Yeah. The mechanics, we tried out the system with the OEM in Scandinavia and they tried it out in Europe, US and China.
Lucas Underwood [00:23:46]:
Yeah.
Christoffer Weber [00:23:47]:
And they just love that. All the mechanics just loved it. They gave full points on everything. But the headquarters say, yeah, it's fantastic. But you know, it's the dealers that has the problem. We don't have a problem.
David Roman [00:24:03]:
Yeah, that's not their issue.
Christoffer Weber [00:24:04]:
But I think nowadays it's all about all courses sold as financed mostly, which means that the problem is usually it's financed by the OEM's financing company. So the problem isn't there any. So it's probably just a way of, they had just realized that they actually own the problem. But at the moment, most oems, they kind of struggle with cost, scared when going into new innovations and things like that.
David Roman [00:24:32]:
Wait, so is the opportunity then going to be entirely because this needs to trickle down to the retail aftermarket independent repair shop? That's a huge opportunity. Is the opportunity then with specific aftermarket manufacturers or even cataloging companies?
Lucas Underwood [00:24:57]:
That's what I was getting ready to say is, what kind of workload is that? I'm thinking of a GPC. I'm thinking of an O'Reilly's. I'm thinking of a, you know, a Toyota. Whoever comes in and says, hey, I want to implement this. Well, now they have to have that inventory to be able to scan.
Christoffer Weber [00:25:12]:
Yeah, yeah.
Lucas Underwood [00:25:13]:
And then you also, on top of that, have to be able to implement into your systems. There's, there's a lot of variables there.
David Roman [00:25:20]:
I don't know. I don't know that that would be that difficult. I'm just thinking who, who would want to, who'd want to jump in like you? You look like an epic or parts cataloging company. Yeah, a parts tech. I don't know. Parts tech has the bandwidth, though. You think they got the bandwidth, but.
Lucas Underwood [00:25:41]:
I mean, they, they don't have the parts. Right. You would have, you would have to set this up with a GPC. You'd have to set this up with an advanced auto parts. You'd have to set this up with the individual dealers.
David Roman [00:25:51]:
Repair link.
Lucas Underwood [00:25:52]:
Yeah, repair link is probably the, the best shot you've got at this, maybe.
David Roman [00:25:58]:
But they just grabbing the, they're grabbing the OE's cataloging.
Lucas Underwood [00:26:02]:
Yeah.
David Roman [00:26:02]:
And they're slapping it up there and then it is what it is. I called in to, I think it was Mercedes. We were looking for something. I think it was a hose too. It was a coolant hose. That's what it was, was cooling hose on an ML, a stupid thing was leaking. You pull up the catalog and everything's just floating in space. There's no context.
David Roman [00:26:23]:
Is this the front of the car, the back of the car? Like, this doesn't make any sense. And so I called the dealer up and I'm like, hey, this thing doesn't make any sense. This is the right hose. It goes from here to here. He's like, man, I don't know. I'll send you my picture. He sends me the picture. It's the same thing.
David Roman [00:26:38]:
I'm looking at a repair link and I'm like, the crap am I supposed to do with this thing? There's no way to know. So they have repair links. Just pulling the information from the OE. Like they're not going to be able to do anything with it. It makes sense to go to like a standard motor products SMP because they sell connectors. But there's more, like you're saying rotors, hoses, anything that you can get context with.
Christoffer Weber [00:27:08]:
If we only get hold of the connector itself, it would be the wholesaler. But then also getting back around information, like provide where it is, the schematics and all those things that easy to put together. And also we've been also playing around with the idea where is it located on the vehicle. So taking a picture on the vehicle and then identifying the pose of it, it's fine. Then we would know if it's in the front, the lights or backlights or wherever, and we would combine that information and then we would increase precision of actually what it is and where it should be.
David Roman [00:27:41]:
Yeah. And all that information would be incredibly valuable to the person looking up the parts.
Christoffer Weber [00:27:47]:
Yeah.
David Roman [00:27:48]:
Which would cut down on, on accurate on, on inaccuracies, turnaround time.
Lucas Underwood [00:27:54]:
I'm realizing who it is at the dealership. Who is complaining about it? The part guys. Yeah, right. Because the part guys are complaining about it is they recognize that, oh, I'm not gonna have a job tomorrow because this customer's gonna scan said part.
David Roman [00:28:05]:
Yeah. And click order, maybe. I don't know.
Christoffer Weber [00:28:08]:
Usually when we tried it out, it was usually the mechanics that used it, and then they created a list of the parts that needed and then just gave the list. Just generate an email, and the email turned up to the parse guy and he would order it. But, I mean, the easiest way was, this is the parts. I need that list, and then send out requests and get it lay on as well. And also, the other things we've been looking at is that if you have the picture and you take a picture, you can also log quality issues, not only logging it, you also would identify the parts that you are seeing in the picture, where they are, and then you get a kind of content about what issues are there that had to be reported.
Lucas Underwood [00:28:51]:
I mean, this literally shapes our industry in a completely different way when it comes to the parts industry. Right. Like, it shifts the way that we've been doing this. If this were to be widely adopted, we wouldn't be doing business the same way we've always done business in a matter of years. Right.
David Roman [00:29:09]:
Yeah, but there's, there's the, like he's saying, like you're saying it's the. It's the profit motive. Like, the. The oes are like, yeah, we don't care. And the dealer's like, we're not paying for it. So everybody's looking at each other going, who's gonna blink first?
Christoffer Weber [00:29:23]:
Usually that's exactly what happens. So. But since the technology is there and it's being developed.
David Roman [00:29:30]:
Yeah.
Christoffer Weber [00:29:30]:
And the ones who picked it up and realized it's not that many, it's not a difficult business case to realize, actually, there's value in doing things quicker than slower. It will come there, and then suddenly it will just roll out. But I think it's kind of similar to most technology where there's come in, people are hesitant, and then when they start using it, they just love it. I mean, for me, it's just so boring to try to find anything. When I do housework at home, trying to find the right kind of screw to put a toilet on the wall, and it's the wrong one, the wrong length, and you just try to find, where the hell do I find this? And then you just have to. But get it done quickly.
Henrik Rosseland [00:30:06]:
But also, when the user realizes how much time he saves. I mean, if you save 30 minutes, for example, by using this.
Lucas Underwood [00:30:14]:
That was my point on that panel today, because, you know, they brag. I shouldn't say they brag, but they're proud of the fact that you can access the Oe information and that, you know, for the most part, you can get to it. Right. It changes often. And the website changes and the way that you log in changes and the source of the data and where you go and what button you click and where you click that button and the color of the button changes all the time. And society has shifted from that being acceptable to, we will not tolerate that. Right.
David Roman [00:30:53]:
And so the way around. What are you talking about? Yeah, they push an update on your phone. Everything changes.
Lucas Underwood [00:31:00]:
Hold on.
David Roman [00:31:00]:
Just overnight, you look at, like, you get used to, like, your movements and then you open up, you're like, the hell is this?
Lucas Underwood [00:31:08]:
But for instance, they just changed the UI in the Google phones. And the UI is a little more similar to the iPhone, but the UI.
David Roman [00:31:18]:
They didn't push it to mine. That's freaking me out.
Lucas Underwood [00:31:23]:
Tomorrow you wake up and you can't use your phone anymore. But the UI is intuitive. Right? And that's what I'm getting at. So, for instance, Facebook, right? Facebook will change things and they will break Facebook. And it is. It is so frustrating because we are so used to technology that use Facebook and Sweden.
David Roman [00:31:45]:
What's in Sweden?
Christoffer Weber [00:31:48]:
Facebook used by the older guys.
Henrik Rosseland [00:31:51]:
That's us. But I don't. I don't use.
Lucas Underwood [00:31:54]:
He doesn't have TikTok.
Henrik Rosseland [00:31:55]:
I have Instagram. If that's okay.
Lucas Underwood [00:31:58]:
I guess so. But, you know, they'll break something in Facebook. And it is infuriating when it doesn't work like it's supposed to or it doesn't work like it used to. It does upset me. And so my point in saying that is, is that, that they continually to run into these issues. The modern consumer does not tolerate that.
Henrik Rosseland [00:32:21]:
But you said something about age, young people, middle people or whatever you say that the different tolerances are differently.
Lucas Underwood [00:32:28]:
Yeah. I read an article a while back that said that you will literally lose a young person's focus past 14 clicks. If it takes 14 clicks to get somewhere, the chance of them actually following through to get there is very low.
David Roman [00:32:44]:
So I think the demand then is that it has to be smooth, intuitive.
Christoffer Weber [00:32:53]:
When I grew up, I started off with Lisa Mac and those kind of things. And when Lisa came, it was fantastic. So I started to make my own user interface. I don't know really what I did, but I tried to do the same kind of things, but now it seems like this is not so intuitive and UX and user friendly. It's different now because it's something people, the designer perceives how he thinks that someone will take it and then they stop thinking actually the user uses something in a way that he's been used to do and suddenly they change that and just thinks that something that is much more clever and it can be clever, but for the user it's different. So he just stops with. And I discovered that I went from Android to iPhone only half a year ago, and I thought it was so crap. And my belief was Apple was best.
Christoffer Weber [00:33:44]:
They had always been doing it, but now I went to, they actually have no idea. And the thing is just behind because I'm used to Android and it's different. And I think this is something a user interface developers have to think about is actually, it's not so much about making it user friendly. It makes me consistent.
Lucas Underwood [00:34:01]:
Yes, 100%. It has to be consistent. And we've talked a lot this weekend or this week about the user interface with CarPlay and Android Auto and the interface in the car. And I think manufacturers are all out here trying to say, oh, we want to go our own way. We want to make this all our thing.
David Roman [00:34:21]:
They're going to figure out that's a bad idea.
Lucas Underwood [00:34:23]:
Well, but even the way that you scroll, I've got this Jaguar rental car. I don't know who in the world decided that I would. I mean, is a rental car company what possesses you to buy a Jaguar?
David Roman [00:34:38]:
Some people, I don't know, they flip it quickly.
Lucas Underwood [00:34:42]:
Do they also own the token? But so, you know, even the way that it scrolls across the screen and the dots at the bottom of the screen are consistent with Android and Apple. Right. The way that we interact is often shaped by these little devices, and so they're forced to try and create that environment that is around the technology we're used to. I just don't know how they're going to get away from that. I don't think it's possible.
David Roman [00:35:15]:
Well, even if they do, I mean, you get into a car now and before it connects the Android auto, you've got the OE menu, it's trash. You're clicking around, trying to find the maintenance. We had a, this was like a 19, maybe it was a 21, it was a newer Audi or Q five. And it had some maintenance features were on the screen, but it depends on the model. So if you've got a certain trim level. You can adjust the maintenance stuff on the MMI thing, right? Yeah, but if you have too low of a maintenance model, it kind of looks like you can do something with the maintenance minder and all, but it's not there. You gotta use the scan tool. And by the way, sometimes you have to try three scan tools before you get one, usually.
David Roman [00:36:12]:
But they put that stuff in there and it's not intuitive. It doesn't make any sense. There's no consistency to it because they'll change it from year to year. 17 works this way, 19 works a completely different way. It's trash. So I don't know why they would try to go, hey, we're going to be software developers or UI developers.
Lucas Underwood [00:36:32]:
I read something once and it was around how much Google had invested in the UX and just the research and watching the clicks and watching where your hand sat on the phone and if you were idle.
David Roman [00:36:49]:
That's creepy.
Lucas Underwood [00:36:50]:
Where did your thumb sit on the phone? That's creepy. The level of detail that they went into and the number of psychologists they hired and all of this information that got poured into, how do we make it to where a human being uses this? And in my mind, I'm sitting here thinking, you've engineered this to such a degree that we can't put it down, right? Like, I mean, you look out here, everybody's walking around like this, right? And so they've engineered it so well that it has become part of who we are.
David Roman [00:37:24]:
David, maybe you. So are you now a fan of Apple or.
Christoffer Weber [00:37:28]:
No, I think I accept it.
Lucas Underwood [00:37:31]:
You've accepted that you're just gonna live.
Christoffer Weber [00:37:33]:
A mediocre life now I will see.
David Roman [00:37:37]:
Well, everybody that tells me, like the Apple's the superior choice has to buy into the entire ecosystem. So it's like, I've got to have an iPad, I've got to have a MacBook, I've got to be hooked up with the cloud, I've got to have iMessage. If I've got all of that, then it makes sense to have the iPhone. Cause it's all interconnected.
Christoffer Weber [00:38:01]:
The main reason for me getting is actually their strongest platform so that we can run a local neural network on it and have a big enough network so that it can. And that's compared to the other models. They're not really there yet, but I think they're kind of.
David Roman [00:38:20]:
Is it too segmented or.
Christoffer Weber [00:38:22]:
No, we have our neural networks on the cloud, on a fairly big computer that can take. Crunch the data quickly, but for kind of like military purposes. You need to have it compartmentalized and then you need to run on the edge device, on the end device. And that's actually where Apple right now is. Better. Yeah.
David Roman [00:38:50]:
So the computer that's crunching the numbers, is that your computer? Are you buying it from somebody right now?
Christoffer Weber [00:38:58]:
We actually, when we do the training, it's our computer. So we have a lot of, as they say, GPU's, Nvidia GPu's. But when we do the inference, which is when you have trained the system and that's where you actually get the identification on, that's done on the cloud. And right now we're using Microsoft Azure and we started off using Google, but since we did some tests in China, Google is, it doesn't exist really in China for those. The chinese government don't let them work there. So we just flipped it and went over to Microsoft Azure and we just buy the services there. And for this one, we don't need some really heavy machines. It's a big network, but we don't need so really big machines for it.
David Roman [00:39:47]:
How many Nvidia GPU's did you snatch up?
Christoffer Weber [00:39:50]:
Yeah, well, actually there was like a half year we couldn't buy them, actually.
Lucas Underwood [00:39:56]:
Yeah, yeah.
Christoffer Weber [00:39:56]:
So when I think you could, you.
David Roman [00:39:59]:
Were just paying three times the price.
Lucas Underwood [00:40:02]:
David went ahead and bought a couple.
David Roman [00:40:03]:
Just so I bought one. I paid way too much for it.
Christoffer Weber [00:40:06]:
Yeah. Everything for bitcoin mining.
David Roman [00:40:10]:
Well, yeah, well, I didn't get an Nvidia, I got an AMD, but they were still triple the price.
Christoffer Weber [00:40:14]:
The issue is actually with AMD, they don't have the supporting software infrastructure bind, where Nvidia really have put a lot of effort into CuDA framework and those things. So that's why everyone in machine learning more or less use that, but everyone trying to do catch up there as well. I mean, even intel is going into the game as well.
David Roman [00:40:34]:
Yeah, but Nvidia is hedging all of their bets on AI, where everybody else is like, we just want to make GPU's fantastic.
Christoffer Weber [00:40:44]:
A lot of the tools and the tools that AI developers use, fantastic tools develop in Nvidia. They really understand to help out there and they really seem to be progressive. And also thinking out the software structure around Pytorch, which is the main kind of development platform. It's really good.
David Roman [00:41:04]:
Interesting.
Christoffer Weber [00:41:04]:
And that's why I think we're really impressed with our engineers. They really catch that up early and they find it really easy to accept and use these tools and really be aware.
David Roman [00:41:16]:
So you said you had connected with the university to start the ball roll. How does that even work?
Christoffer Weber [00:41:22]:
Yeah, we actually, we discovered that it's impossible to get any experienced AI developer engineers.
David Roman [00:41:29]:
Okay.
Christoffer Weber [00:41:30]:
So what we did was actually got in touch, I think by accident, actually with a few who just got an exam and had done their thesis projects on AI. On AI with the different companies. And I think what they told me is actually they liked that I actually knew AI quite well. So when I liked to work with a team who actually knew what they were talking about. Because in Sweden it's mostly kind of a consulting company. They do things around AI, not so much big companies. Spotify is of course really big in AI, but except for that, there's a lot of consulting companies and they really don't go in deep. They actually do the kind of this.
Lucas Underwood [00:42:16]:
Kind of high level presentation on it.
Christoffer Weber [00:42:19]:
And that's kind of what they got on board. And after one year we got a few, did set up some really cool thesis projects and that's what they really liked. They were kind of really on leading edge and so that's kind of the way we recruiting, getting them in on thesis projects and doing some really cool stuff there that's beneficial to us. But also kind of is somewhat academic. Interesting, but also.
Lucas Underwood [00:42:46]:
Yeah, for sure.
Henrik Rosseland [00:42:48]:
How many years, if five years, you can choose AI in the last year.
Christoffer Weber [00:42:53]:
Yeah, we've been having five years been working with. But I myself been.
Henrik Rosseland [00:42:57]:
But I mean, the engineers, they chose this last year.
Christoffer Weber [00:43:01]:
The last year. So they are in Sweden. I think they use what they call bologna model, which is three years to get a bachelor and two years to get a master. And all of them that works for us, they have a master in AI, in artificial intelligence.
Lucas Underwood [00:43:19]:
I didn't even know that was a thing.
David Roman [00:43:22]:
What?
Lucas Underwood [00:43:22]:
That there was a master in AI.
David Roman [00:43:25]:
And universities now have degrees in absolutely anything.
Christoffer Weber [00:43:30]:
Absolutely. The leading ones are the Americans, Stanford and Berkeley, MIT. I mean, they're the one who have the most momentum. And there's some fantastic courses on the web, mostly by Coursera.
David Roman [00:43:44]:
Coursera, sure.
Christoffer Weber [00:43:45]:
And I'm signed up on them. Deep learning AI by Andre. Fantastic. They're really fantastic lecture as well. It's really enjoyable to my wife. Been complaining because I spent a lot of vacations just sitting and doing all these things, but I couldn't help it was so fun.
Lucas Underwood [00:44:03]:
Right. Well, I think that the majority of our listeners are going to take great reprieve in the fact that you don't think AI is going to take over tomorrow. It's not going to take their jobs and they're going to be okay.
David Roman [00:44:17]:
We heard that the robots were going to put the tires on and off.
Christoffer Weber [00:44:20]:
I think this is going to be the thing. I think journalists and sometimes even politicians, they want to create this kind of hyped up, stressful situation. But in actuality, I think things takes much longer time than expected. And on the other side, suddenly it has happened. So in some ways things are quicker. But I think mostly it's like my grandmother, like 30 years ago, said that all the robots will take over. So what will all the humans do? And now in Sweden, the unemployment is much, much lower than that time. There's some robots in factories and stuff like that.
Christoffer Weber [00:44:57]:
And I think it's going to be the same. It's going to be. The big problem for the future is actually that it will be not enough people to do all things that we want to do. And I think also, which is also positive, that the overpopulation will start to decrease when people are getting better. And that's also who will do all the jobs for us, clean the apartment or do boring stuff. And that probably is where we should focus on getting automation and robots to do that.
Lucas Underwood [00:45:26]:
And that's one of the things that I took away from this morning's presentation as he was talking about that and looking at the different avenues that it could take, and the fact that it's more focused right now on the development that we're seeing is about improving the human element, not taking the human out of the element. How do we make them faster? How do we make them more efficient? How do we get them the information faster? You know, and the JetBlue thing came up because he's talking about, if you call them and they give you this information, and I know that American Airlines is using something similar to that. They called me on, I guess they called me yesterday, first thing in the morning, it was 06:00 a.m. and they called and they said there's a 90% probability that your flight out of Phoenix on Saturday is getting canceled. And I'm like, well, somebody had to tell them that, right? They're not just going to automatically call me up to tell me that. They usually wait till I'm standing at the gate ten minutes before I need to go, you know? And so that, that is interesting to me because it, it paints a very different picture than what mainstream media and everything else is talking about.
Christoffer Weber [00:46:34]:
I think this is exactly where it's going. You actually provide the right information at the right time.
Lucas Underwood [00:46:41]:
Yeah. And that's intriguing to me. Right. Because the human mind is a very complex thing, and we think in very unique patterns, and we all think differently. I've read a lot into psychology and how we think and the different thought tracks or the ways that we're raised. And we think differently than Bob over here, whoever it may be. We think and we process information differently. But the ability to get the information faster and more accurately and know where to go for the information and be able to process very, very quickly takes a human being that's 50% efficient and makes them 100% efficient just like that.
Lucas Underwood [00:47:22]:
Now I'm in the same respect. It can make you lazy. Right. They redid Gmail on Google phones and it took away, like, when I go to type somebody's email address, unless it's someone I sent an email to like super recently, it doesn't auto populate unless I've saved them. And so now I'm like, oh, man. And I've been doing it like this for years.
David Roman [00:47:50]:
Right?
Lucas Underwood [00:47:51]:
I don't ever save anybody's email address. In case you email me, you know what's up. But, you know, I think those are the things that we become lazy to. And that technology comes in and we kind of check out. When they take that technology away, it's almost like that little bit of a wake up call. Oh, you don't know how to do this anymore. Whoops.
David Roman [00:48:10]:
It's not a bad thing, though, because like, you're saying it's the boring stuff. Why do I need to know what your email address so like, let the stupid thing. Well, that's what I'm saying.
Lucas Underwood [00:48:21]:
They took it away. Yeah.
Christoffer Weber [00:48:22]:
I wonder why they do that.
Lucas Underwood [00:48:25]:
I know. That's got to be what it was. Kind of just like an awareness. We're in control here.
David Roman [00:48:31]:
That's too clunky. I don't like it.
Lucas Underwood [00:48:33]:
We use, we use Google chat for the shot.
David Roman [00:48:35]:
Yeah, I can't use any of that stuff. I'm a Microsoft person. I like my outlook.
Lucas Underwood [00:48:41]:
I just, I want to know. I started to ask the guy and I didn't think it would be appropriate in the middle of his presentation. You know, the video that they did as he's asked it to make a video of him talking about breaks and, and how to convey to the client they need breaks. And then it translates it to Spanish directly. After that, he said you could spend between 1000 for a video and 10,000 for a video. I'm curious how much he paid for that video, for that presentation.
David Roman [00:49:07]:
That's expensive.
Christoffer Weber [00:49:08]:
Yeah, yeah. It's kind of expensive to generate that kind of picture.
Lucas Underwood [00:49:12]:
Yeah, the detail. Right. Because I mean, that was, well, that's.
David Roman [00:49:16]:
A, so we post this stuff on YouTube and one of the questions that now asks this is new, is anything that you're about to put up AI generated? That's what I ask now.
Christoffer Weber [00:49:28]:
Okay.
David Roman [00:49:29]:
Because it wants to be able to tell somebody, hey, just so you know, this is fake. This isn't, this isn't the president speaking. It's, you know, been AI generated. If it can't be that, it can't really be that expensive because you see some goofy stuff go out there.
Christoffer Weber [00:49:47]:
Well, it depends. One of the ways you just changed how the faces changing and behaving and speaking can do. You can just change how the mouth is changing when you speak. So I think that could be not so expensive, but I haven't tried. I mean, I'm not so really interested in that space. But there's a lot of money to be made around these generative things. There's a company called Stability AI that seems to be doing a lot of that. And I mean, I'm not sure if it's, I mean, probably something for the film industry seems to be able to do a lot of that.
David Roman [00:50:27]:
Yeah, maybe. I don't. It's still, the technology is going to have to advance quite a bit. It's still a little clunky. We had, there was a gentleman, he was posting some copywriting. He was trying to sell something. In one of our comments, in one of the videos, he posted the video up and in the comments section, the guy decides he's going to try to sell, pitch his product. The problem was he had, when you ask AI to write you something, AI tends to use words that are not generally used.
David Roman [00:51:03]:
The vernacular is not common. It's not, it's an english word, just nobody uses it, no human uses it. And so you can always tell that, hey, that's been AI generated because of the usage of specific words. And I called them out and I said, hey, if you're going to AI generate this, it's fine. AI generate a lot of stuff. But I have to go back and I got to reread it and start changing words and put this word in front of that one and changing it around so it sounds like a human wrote it and not, and not AI. And I think the film industry and just stuff like that, that's what's preventing, I think, it from progressing any further. It's just, it's not far enough advanced.
Lucas Underwood [00:51:54]:
You need to watch from 3.0 or 3.5 to 4.0 if you're gonna do that, because it definitely makes it much more.
David Roman [00:52:01]:
It's 4.0, smoother, much clutch. Is it really? Yeah, I don't know.
Lucas Underwood [00:52:06]:
It's a pretty dramatic difference.
Christoffer Weber [00:52:08]:
I think it is. It's. Right now, one of the most interesting things that happened yesterday was actually meta released loma three as an open source software model. It's not. And they're going to release also this 400 billion parameter version of that, also open source. So that's going to really be something really interesting to see.
Lucas Underwood [00:52:36]:
What open ended think it will shift? I think it will shift the developer. How far cheap or free.
Henrik Rosseland [00:52:42]:
Is that better?
Christoffer Weber [00:52:44]:
Well, I think it's really good because at least. Yeah, it looks like it's going to be really good. I think the ones who tested looks really good. It looks really like.
David Roman [00:52:54]:
What's the difference between that and what Google tried to do?
Christoffer Weber [00:52:57]:
Well, I mean, they say they want to pay everyone who uses it, have to set up a schema and pay for it. And now it's open source. We mean that someone who can take the model and use it for their purposes there will see something in the licensing agreement that I'm not sure about the details there.
David Roman [00:53:14]:
Oh, I'm sure there's something in the background. It's like, I just let you know, we can see everything you're doing and we own it.
Christoffer Weber [00:53:21]:
Yeah, but that's the usual thing. But I think this kind of interesting with mehtas, they are not the one who were first with Chap GPT or not Gemini or Bart with Google. They're not really the ones who are on the leading edge there. The only thing they needed is a lot of compute. And then they trained a model that was slightly bigger or bigger than the one they run. And how do they go to mark, do the same as the other ones? No, probably not. They worked so much on Pytorch and made that also open source available for everyone. And that, I think changes the game for everyone.
Christoffer Weber [00:53:57]:
So suddenly, I think what they are into is to see how they can at least get the legs away from all the competitors somehow.
Lucas Underwood [00:54:06]:
Yeah.
David Roman [00:54:09]:
Where are they going to make their money?
Christoffer Weber [00:54:11]:
Yeah, that's interesting.
David Roman [00:54:13]:
Like Google Android's open source. So if you want to create the user interface or whatever for your phone, anybody can get it. Yeah, it's free and you just have to develop it. Where Google made their money, everybody's gonna buy the apps from us because everybody's gonna want to buy the apps from us. And we get a little cut of every single app that gets sold on our play store, but you can do whatever you want with this user. And pretty genius Apple decided to go, hey, we're gonna log everything down. Everything's through us. I just don't.
David Roman [00:54:46]:
Where does Facebook make their cash on this?
Christoffer Weber [00:54:49]:
Well, they make it on Facebook. So they advertise me that this, I'm.
David Roman [00:54:53]:
Saying the AI stuff.
Christoffer Weber [00:54:54]:
Yeah, man, I think they provide more services, different things, and so they kind of add to their kind of you're more, more functioning functions and all that stuff.
Lucas Underwood [00:55:05]:
I think they're always chasing the bullet, right? No matter what, they're going to be chasing the bullet. And all of them are trying to balance what we're going to give away to keep people engaged enough to pay us for. And I think you're right. I think advertisement is the leading thing for Facebook. If they can keep their platform relevant enough that now they can sell advertisements on said platform and they can keep creators engaged to where they'll be there posting their stuff. That's, that's the model. But they don't want to be MySpace, right? They don't want to fade back out of this. And, and I could not imagine being in the C suite of an organization like that trying to balance this right now, right.
Lucas Underwood [00:55:52]:
Because the technology is moving so fast and it's changing and the mind of the consumer is changing so fast and what they want and what they expect is changing.
David Roman [00:56:02]:
What's the difference between the, where's the big shift? Because Elon Musk came out with an AI open source something or another.
Christoffer Weber [00:56:14]:
I think that they are more on par competitor to the Lama three model. So I think that's interesting to see. Probably there's going to be a competition between them and so it probably might be the one who are fairest, clearest, or it doesn't provide so much kind of dreaming factor. We actually just invent things. Invent the answer. I can't remember the name of it, but that's the thing that probably the model that is best and more consistent that will be the winner.
David Roman [00:56:43]:
It doesn't start spitting out gibberish or doesn't have. What do you mean? Like, it doesn't start spitting out gibberish or doesn't have a bunch of caveats on the backend that like, hey, if you're going to use this, you've got to click to agree to open. AI seem to have the best or the most transparent model. It's like you can use our dumbed down version or you can get all the features you just got paid a little bit.
Lucas Underwood [00:57:06]:
The dumbed down version. Told one of my technicians the other day that he was the pinnacle of auto repair. And I said, I'm sorry, but it lied to you. It was like your input was off, buddy.
Christoffer Weber [00:57:20]:
But I think it's very interesting to see how these kind of things just turn out. There's many, so much money into it and so much compute means like football areas full of compute just to crunch it so well. It's kind of very easy for small companies to, to do it themselves. And that's why it's good for us as a small company to be able to get a already trained model and fine tune it. And so that's going to be interesting.
Lucas Underwood [00:57:48]:
There was a video documentary a while back. I'll see if I can find it. I can't remember the name of it. It was chip wars or something to that effect. And it was talking about Taiwan and it was talking about the US and the money that the US was investing to bring that technology here to make the. Yeah. And, and the, the difference between the four nm and the five nm chips and how much the US was willing to invest to be able to control that technology or at least be able to compete with Taiwan and China on said technology. And I think it indicates the, the fact they were willing to spend so much money and they were willing to do so much to get it.
Lucas Underwood [00:58:33]:
It shows how powerful this technology is going to be. Right. Because this is what's shifting and shaping the direction that we're headed, whether we like it or not. And so it's interesting to see they were willing to start world conflicts over making that happen. Right. And I mean, that's saying something, right? I don't know. I probably shouldn't say that.
David Roman [00:59:00]:
I'm pretty sure we are willing to start world conflicts forever for absolutely any reason. We don't care. Like, hey, the military industrial sector is a little lagging in their stock price. That country over there looks a little susceptible. It needs a little freedom. We need to go over there, give it a little freedom.
Lucas Underwood [00:59:22]:
Listen, they said Biden was willing to give up his paycheck to get them to move from Taiwan over here. So I mean, it must be pretty bad.
David Roman [00:59:31]:
All right, awesome. Guys, that was fascinating.
Lucas Underwood [00:59:34]:
We need to know how to a.
David Roman [00:59:36]:
Before we working with the OES. He's working on it.
Lucas Underwood [00:59:39]:
Well, I know.
David Roman [00:59:40]:
Can I buy that? Do that?
Christoffer Weber [00:59:41]:
Yeah. But my colleagues who have been helping produce these things don't tell anyone that you can buy this case. Yeah. We have to get into a project where we develop a case for you, and we would like to be in the case where we actually can say we have a. Like a case for some OEM. But the problem is, one OEM might have, like, 100 connectors or 200 or 300 connectors in it, a variety of that. And that's kind of difficult to make a more generic solution. So at least for the time being, we are doing a custom case, but the case or the app has to be then developed for each and every one.
Lucas Underwood [01:00:25]:
Okay.
Christoffer Weber [01:00:26]:
But progressing further, when we have trained.
David Roman [01:00:29]:
Like Tensorflow, we have one specific for our manufacturer. For example, because there's specialty shops out there, and there are Volvo and Saab specialty shops in where I live. That's all they work on is old Saabs and Volvos, mostly Saabs, because nobody else wants to work on them. Oh, yeah.
Christoffer Weber [01:00:47]:
Yeah.
David Roman [01:00:48]:
So you send it in Kansas City shops called Georgia Import Saab is very.
Christoffer Weber [01:00:52]:
Close to our headquarters. I used to work there as a welder at the time.
David Roman [01:00:56]:
Really? That's awesome. He will work on the old, savvy, grumpy old man. Nobody likes to deal with him, but he's the only one willing to work on that 93. 93. Like, he's the only one. And so if you have a kit that, hey, this is our Volvo kit. This is our VW kit. This is our.
David Roman [01:01:12]:
Whatever BMW kit. That would be cool.
Christoffer Weber [01:01:16]:
As it is right now, we actually do that for the respective oems.
David Roman [01:01:20]:
Yeah.
Christoffer Weber [01:01:21]:
So. And get us part numbers. It's a bit difficult for us to change the business model because it's kind of a different thing. We then would have to investigate to do the right. Do the right connectors. But it actually might be something we could consider.
David Roman [01:01:37]:
Well, if you can sell it to.
Lucas Underwood [01:01:38]:
The oems and we can buy it from the oems.
David Roman [01:01:41]:
Yeah, we need part numbers. So if we can buy, because we'll just order it from the oems so we can get at least a kit. If we see a lot of x manufacturer and it's something that we want to get into, it makes sense for us to have a small kit like that for emergency situations. Eight most common connectors like you have in there, but the tools is where the tools would be legit.
Christoffer Weber [01:02:05]:
And the other thing, what we're using, we also use the app so that we can make sure that we have the tools in the right place. So that, like an inventory thing.
David Roman [01:02:13]:
Yeah.
Christoffer Weber [01:02:13]:
So if there's something missing, you can buy.
David Roman [01:02:17]:
Oh, that's cool.
Christoffer Weber [01:02:17]:
We're working on that for a wrench. Manufacturer so that when they pull out the drawer, you can see which is missing. And so they can order the right one, because that usually seems to be a problem as well. Fully understood.
David Roman [01:02:31]:
About 10 mm lost.
Lucas Underwood [01:02:32]:
Well, just 10 mm. You ever seen a dealer tool room? It's bad news, bears. It's bad, buddy.
David Roman [01:02:39]:
Can't look worse than mine. All right, all right.
Lucas Underwood [01:02:44]:
That was all.