Loading summary
Transcript34 lines
- [00:01]
A
Welcome to the kitesurf 365 podcast.
- [00:04]
B
Welcome back to the show. I hope you all had a great weekend. Today I'm joined by DK for another techcast. DK is going to continue to answer the questions from the AMA we put out at the end of last year. We received so many questions, we're working our way through them. We are nearing the very end, so we'll call out for more questions. But today's episode, we talk about AI board calls and kite struts. Unfortunately, we didn't get get to catch up with Andrea yesterday. Unfortunately, with what's going on in the Middle east, we decided to push that back. So hopefully Andrea will be with us on this Thursday to talk about his jump from the A Dubai. Don't forget to follow me at kitesurf365 for all the latest episodes. Ladies and gentlemen, Dave K.
- [00:56]
A
Turned off, so I don't get any blings coming in. WeChat. That's cool. WeChat. There we go.
- [01:03]
B
WeChat. People use WeChat still?
- [01:05]
A
Yeah. That is like my number one comms for all the factories in China.
- [01:09]
B
The Chinese government's got all your information.
- [01:11]
A
Ah, totally. But the big thing with WeChat is it allows me to. I can do auto translate. A lot of the guys that I'm talking to directly on the factory floor, they had some English, but their written English is a little. Isn't that great? And if we're using WeChat, then it's just he translates one way, I translate the other. So no, it's like it's one of my most important tools.
- [01:30]
B
I mean, we use line here in Thailand and it's also the same. You can set up that automatically translate and that's very techie. And let's get to another TechCast before we get to the TechCast.
- [01:39]
A
TK number 27. We're on now.
- [01:43]
B
What did you think of Andrea's jump off that Aim to buy? I'm asking your opinion as a brand owner and designer and from a gear perspective, because I guess to the layman they probably think I just floated down, but I mean, it's pretty scary shit.
- [01:58]
A
I can see why Red Bull likes to fund this sort of project. It gets massive amount of eyeballs on them. Non kite surfers look at this and go, wow, that's just absolutely incredible. From a gear designer point of view, we aren't making paragliders. Paragliding industry is so much more internally regulated in terms of gear quality than what kite boarding is, simply because, like the consequences are just massively higher. But also, paragliders are also more inherently damage tolerant. Like you're not hanging off just two lines. Basically all the bridles kind of terminate in one place. So in terms of like one broken line doesn't, doesn't have you plummeting out of the sky on a paraglider. One broken line on a, on a kite and you are plummeting to your death. So spectacular and gets a lot of eyeballs on, on Andrea and his sponsors. But from a product designer's point of view, it kind of shows our sport in a light that we probably, we don't want that to be a thing. Like you shouldn't be jumping off stuff on your kite. It's not designed for it. 30 meter jumps over water without any obstacles around is dangerous enough. Like you don't need to be jumping off big, big structures to increase that danger, like, well beyond anything that we've actually designed the equipment to do.
- [03:23]
B
You know, normally when those jumps happen, DK like they bring the kite round and it lifts them up first. The kite creates some lift. He just slips off.
- [03:31]
A
Yeah, totally. As far as the kite is concerned, a lot of people ask about why we can't use like paragliding profiles in kites. And paragliders have got like, their lift drag ratio, their, their ability to fly forward is much, much higher than the profiles we use in kites. A paraglider almost cannot luff because it actually sees constant load. Pilot is hanging directly below the, directly below the kite or below the wing with the paraglider. Means the paraglider is basically always seeing wind. Because if there is no wind, if you're not moving forward, you're actually dropping down. And in dropping down, you're actually creating wind which will cause you to fly forward. It's the nature of it. As long as you are hanging off the bottom of a paraglider, or in fact, if you jump off a big thing with a kite, the gravity pulling you down means that the kite seeing tension is actually flying because it'll fly because of the wind of you falling. So once Andrea is actually up in the air, like once he's actually cleared all the structure, the kite's actually flying even if there is no wind. And that's why he can do everything that he's doing on the way down. The part of that entire jump that's incredibly dangerous. And just like there is so much that can possibly go wrong with the tiniest little mistake is all about that launch. It's all trying to get off the top of that structure. The behind the scenes video of Nick's jump off the Berger lab in Dubai. The behind the scenes video of that, like the, the long form showing them like how gusty the wind was and how difficult it was to just get the kite in the sky off the top of the, off that helicopter helipad.
- [05:20]
B
Yeah, yeah.
- [05:21]
A
And that was some flat like effectively wide open space. I don't know. Like the top of that, the top of that wheel is not flat and doesn't have wide open space. It's just, and there's like lots of structure that you can potentially get hooked up on and zero wind. Like that's everything about that's just like dodgy as all get out. So. But we made it happen.
- [05:45]
B
All right, let's get to some tech questions and I'm going to jump ahead in our list down to this question about AI models. And it's from a rink off the portrait discord. It says how will advanced AO models and fast paced innovation change kite design in the near mid and long term? Because a lot of people ask me about AI all the time, dk, especially with jump heights. It's the most common question I get asked now. They are. And honestly my answer is I don't know. But for sure AI is going to play a major role. How do you see AI impacting kiteboarding over the next well from now into the future?
- [06:21]
A
Ah, this is a, that's a, this is a really good question. And this is, this could end up being quite a long answer because there's a lot here and I have a lot of opinion here because this is right up, this is right up my street. So what I'm going to, I'm not going to answer that the kite thing directly to start. What I'm going to say is I like I. A large part of my day to day work life now is I'm, I'm doing software development. Whether that's developing our kite design software here that Lacuna or Sue uses to design the lacuna kites or I have third party clients who I'm developing sophisticated backends for them to use within their manufacturing workflows. All kite related. But I do a lot of software development and I also have a couple of commercial software products that I sell. I use AI large language models quite a bit when I'm doing software development these days because they're incredibly handy tools. They also, they can produce big piles of garbage. But if you, yeah, if, you know, if you have a good knowledge of the domain that you're working in, they can, they can be incredibly handy to develop software with. And the reason that they're in so incredibly Handy is for 20 years people have been publishing their open source projects onto GitHub. And for probably 25 years there's been Stack Overflow, one of the major software development forums available on, on the web, and every single one of those open source projects which are on GitHub, and every single forum post that's ever been in Stack Overflow, plus a massive amount of other data has been scraped by every one of the companies that have developed AI tools. All of these large language models have been trained on a massive amount of publicly available software. Just stuff that can be read and some of it's incredibly well documented, some of, and some of it has been built upon over years and years of development. And those models just go through and it's just all there. They can just, they can scrape, scrape it right to the bottom of the barrel and get all of that past human effort that can then be effectively. My AI model is like if I'm writing a piece of code with ChatGPT, I'm literally asking it to go and have a look at the history of GitHub to try and find a solution to the problem that I'm talking about. AI isn't really doing anything more than just consolidating well published human knowledge. I could just use Google to go and search through GitHub posts to find the solution. Large language models are a more efficient way of doing that. So AI works great. Computer modeling in terms of a design tool works absolutely fantastic under two rules. And the first rule is one, your problem is really well defined. And secondly, that there is a very large knowledge base of previous stuff that you can build upon. And writing software using large language models defined, both of those things. If you can define the problem that you're trying to solve with AI very well in your, your post and it's been solved in some way in the past that's been published somewhere that can be regurgitated by the large language model. Wicked. You're going to be very productive. Okay. Trying to apply that to the kite problem. Firstly, our kite problem is really ill defined. Like what is high performance in kiteboarding is typically light bar pressure, fast turning. That's very much a, that's a qualitative measure, not a quantitative measure. It's not something you can quantify when you look at a design of a kite. And firstly, our problem is ill defined. We don't have a well defined problem. And the second rule is that there is a large base of knowledge available to, to draw upon and there is no kite designer has ever published anything more than like, nothing. In fact, there is no publicly available data on kites from our industry. Like every single kite designer has what they've been working on, and they certainly do not discuss what they're working on with anyone else, let alone release that information publicly. So the AI's got nothing to train on. The AI's got nothing to train on. It's got nothing to, it's got nothing to give you back. So unless there's a law that changed at every single kite design that's ever been produced by any kite company has to go into some publicly available database so that everyone can, everyone can train their AI models on it, or out of luck, it's just not going to happen. So the AI needs data, that data is not available. No one's ever going to give that data out. So the AI model's never going to get anywhere. Maybe some companies will start reusing their old data inside machine learning models to see if that will improve their workflow, but I don't see it.
- [12:12]
B
Where can you see AI being used in kitesurfing outside of creating podcasts and media and content, which to be honest, in four or five years it's all going to be AI. AI, Right. But where else do you see AI in kiteboarding? Like, what about jump height? I mean, that's the one I always get asked about, whether there's going to be something in AI that's going to put a nail in the coffin of this debate of who's the most accurate.
- [12:39]
A
Look, the problem with the term AI is just very poorly defined. When modern AI, we're talking about large language models. Large language models, fundamentally what they do is, is their text prediction. So they've broken down every piece of text that's ever been that they can ever get their hands on. Given the words that we have now, what's the next one going to be? That's fundamentally what it does. It's a little more complicated than that because it's a tensor and it's a multi dimensional model and the weightings across that model change how well it can predict what the next thing to do. But it's a text prediction. Text prediction is going to do nothing to work out how high you jumped. Like say pictures and movie creation are not too dissimilar to that. They're just a much, much, much more data. But it's effectively just, it's remixing Stuff that's already existing that's going to not help you jump height. There are many other types of algorithms that can be dropped under the AI banner. Many of those have been worked on for like 40 years. So large language models are kind of like the new kid on the block. But there's the idea of using computers to intelligently predict data has been or help you work through problems that are unsolvable been around for a very long time. I could describe my, my kite tuning software, the kite tuning algorithm that I've written for Corora cad, our kite design software. AI based kite tuning module. A couple of years ago I used to describe as an AI tuning module. I've actually completely got away from that because it's like, it's almost like poison. My tuning module is not a large language module. It's actually a genetic based algorithm where it's taking a problem that's six point bridle on a kite is a tuning problem that can be modeled as a multi dimensional hypercube where that cube has nine sides and the actual the volume inside that cube. So all possible solutions of bridles available for that one set of parameters that you can put on that kite is something in the order of 10 to the 24. I can't calculate all of those bridles, but some of those bridles are possibly suitable to fly kite with. So I've written an entire algorithm that chops that problem down from uncalculatable to 30 minutes. And 30 minutes gives us a bridle that's absolutely amazing and that's a really, really good application of what the world would now describe as AI. But it's an algorithm that actually helps solve uncalculable problems. What's already sitting inside say a WOO device or the, the software on your watch for a surfer in this day and age could easily be described as AI. It's already taking a problem that is essentially uncalculatable like it is trying to measure jump height with or trying to measure distance. We're using accelerometer is really, really, really, really hard. And in the 10 years that the, that Wu's been going they've chopped that problem down from really, really, really, really hard to this is usable and it's great and we can agree that it does the job. Is it truly accurate to millimeters? No, it just can't be. But it's absolutely amazing that we're even got what we've got. So yeah, AI in the big picture is already doing that for us.
- [16:18]
B
Can AI be involved in scoring.
- [16:20]
A
That's actually quite an interesting one. What you'd have to do is basically take a model and feed it the video footage, the raw video footage from every camera from every heat that's ever been, that's ever been run in kiteboarding and give it some parameters it needs. Your model would need to know what each move is and then what you would like what it was scored. If you've got enough like raw data plus the results that you're looking for from that raw data, then you can produce good results. So yeah, maybe, maybe if somebody could consolidate all the video footage from every single kiteboarding heat that's ever been run and pile it into one server and then consolidate all of the results from every heat, like what each rider got for each trick and time it up. Yeah, you might actually be able to do something. But I'm sure at some point like a bird will fly across the screen at just the wrong time and somebody will get a 10 point trick. It's like it's. There's always. There's so much noise that can be introduced to that, to these type of, these type of things. You would, you would never rely on it. You'd still have to have somebody to go really, at the end of the day, could help judges and in being able to give them indications and work its way through. But it's. You got to have the data to train it on.
- [17:46]
B
That's where it will probably start. Right. It'll start off as a assist. This is what AI is giving it. Then the guys can sprinkle on the variety or the impression, you know, which is optics really, and working it that way. Anyway, let's move on dk because I feel like we could do a whole bunch on AI itself.
- [18:06]
A
Yeah, yeah, absolutely.
- [18:08]
B
This one's from Donny Hall, 14, via Kite Forum. Do many kite designers experiment with different pressure levels on struts? And how are strut placements, diameter and pressure pressure determined?
- [18:19]
A
All right, so the. My first note here is every modern kite's on a one pump system. So we can't actually put a different pressure in the struts compared to in the leading edge. A lot of, a lot of high performance wings actually have separate inflation for both the strut and the leading edge. Strut's smaller in diameter and can take more pressure than the leading edge. So some designers actually recommend that you put more pressure in the strut than you put in the. Than you put in the leading edge. They can do it. We can't. We get one pressure everywhere. The stiffness of a strut though, or in fact, even the stiffness of your leading edge. Small changes in diameter make massive differences. Basically double the size of the strut and it's eight times stiffer. Half the size of the strut, it's eight times more flexible. So designers will play around with strut diameters and a lot of designers will talk about micro diameter struts and flex struts. A big part of what they're trying to do is have the stability of the canopy with that strut, but trying to get weight out. So the struts are a significant part of the weight of the kite. The smaller you can make them, the less surface area they have, the more you can drop the, you can drop that weight off the final weight of the kite. From there, strut number, strut placement, it's down to the designer's preference, really like what they're trying to achieve with that kite. Where the struts are placed on the kite is determined because they actually end up on the seams between, between each of the panels. The number of panels in the kite dictate where you can place those. Where you can place those struts. Some designers will actually vary the width of your panels, which allows you to sort of move the strut up and down the kite a little bit. If you're just trying to get a little bit more support further up the kite rather than further down the kite, you can play around with that. But all comes down to the designer and their iteration process and what they're like. What's the design brief for that kite? What are we trying to do? What are the performance characteristics we're going to get out of it if we change the struts and struts are like just one part of like the thousands and thousands of different options you can change on a kite to affect its overall design and its overall performance with struts.
- [20:57]
B
Just getting slightly off the topic there, you know, the three strut, five strut in big air kites goes back and forth. All one to three, went to five, down to three, back to five again. Why are the guys going back and forth on, on, on struts when it comes to those big air platforms?
- [21:12]
A
The number of struts on a kite has been used to determine what style of kite it is. It's. And it's so zero. Strut kites, one struck kites, they're light. Wind, they're foiling kites. Three strut kites all round kites, beginners kites, five strut Kites are big air kites. And it's like a very, very blunt tool to determine what that kite's actually or what the design intent for that kite actually is. So in big air kites in particular, the debate between three struts and five struts is weight, flexibility, turning speed, kind of versus. Yeah. Versus high end stability. Basically, if you've got more struts on a kite, where it makes a difference, where more struts on the kite is considered better than less struts on the kite is generally to the higher end of the wind range of what that kite's in. So as you are using a kite in more and more wind range, you're using it more and more depowered. A lower angle of attack and there is more chance of the canopy blowing down. And if the canopy blows down at a low angle of attack, it starts to lose shape and you lose, basically you lose stability in that, in the profile. Putting more struts on the kite allows you to keep more stability to a lower angle of attack. The kite generally performs better at its high end, but it's generally a stiffer frame. So three strut kites and big air kites tend to turn faster in terms of the, the big difference between those two airframes. And that's why there's so much back and forth. So, but you can, if you put a flatter pro, a flatter entry profile on a three strut platform, it's that flatter profile tends to not blow down as easily, so it'll stay stable to the higher end. It just tends to not produce as much lift and glide. But maybe if you, your three strike kite's designed to turn really fast, that'll help there. But then you start like it's Pandora's box. There's a massive amount of other things that determine how a kite performs than just the number of struts that's sitting on it. And you can't just point at a kite go, oh, it's three struts. It's X. It doesn't, doesn't work like that.
- [23:39]
B
How were the newer materials like a Lula did that allow guys to use the three struts in big air more effectively and still get that speed.
- [23:49]
A
It's not so much the strength and the lightness that actually makes Alula really appealing to a kite designer. It's its stiffness. So you can, what a, what you can do with an Alula based airframe is you can reduce the diameter of everything and still retain structure. So if we come Back to what I mentioned a few minutes ago. If you, if you halve the size of a, of an inflated tube, it's eight times more flexible. So it means very, very small reductions in diameter can make a, can make a Dacron kite very, very flexible or very unstable very quickly. And what the stiffness of the more modern materials does is allow you to reduce the diameter of all of your inflated components and still retain good stiffness, good stability at the top end. That's really what the designers are sort of leveraging on when they're designing kites around, around these new materials. The lighter weight, that's the, the bonus on top of that extra stiffness.
- [24:57]
B
Here's a question, dk, that I know you'll be very, very happy to answer. What thickness plywood is used as a core inside kite boards? When is plywood core preferable to foam?
- [25:08]
A
Well, firstly, no current production kite boards that I know of use plywood as the, as its core. So that's a very easy one. None. Plywood is a horizontal laminate of veneer that by definition is a horizontal laminate of veneers. So let's say, let's say five to seven pieces of veneer and with the grain alternatively laid in 90 degrees. So each layer of the ply is rotated 90 degrees to the previous one and laminated together to get the thickness that you're looking for. And it's a great material for reinforced sheathing on your house. But isn't something we use for the cause of kite boards because it's not really suitable in terms of its characteristics for what we're looking for. So what we do use for wood cores in kite boards is a vertical lamination. It's strips of timber typically about 15 to 20 millimeters thick and say 50 millimeters wide. And if we have a core, it's a core, is a standard kiteboard core usually starts off at like 1.5 meters by half a meter. So it's 10 pieces of those 50 millimeter wide. 15 millimeter thick strips are all laminated edge to edge along the length of the, along the length of the core. And that's what's called a vertical lamination. And what that gives us is, and I often get asked why we don't use a single piece of wood. Well, firstly, it has to be a really big tree to get 500 millimeters wide. You would the tree, let's say the Tree is only 550 millimeters wide. You'd only get one strip out of the middle and then the rest of the trees around you, you'd never be able to cut a single piece. And that single piece would also not be anywhere near as strong or as consistent as the vertical lamination. By laminating up many small pieces, so many 50 millimeter wide strips, or sometimes even narrower 25, like we can go 20, 25 millimeter wide strips, but those very small strips, we can machine a lot of those from a single tree. And if there are any imperfections, let's say there's a knot or there is a crack in one particular piece of timber by, or there's different flex patterns across different parts of the tree by sort of mixing it up and gluing the pieces back together. Anywhere there is, say a problem in one board, that board is sandwiched on each side by a board that doesn't have that problem. And so the sister boards are the ones, each side will take the load where the one that may have a problem might not be able to. So you get a stronger, a much, much stronger final product if you do a vertical lamination. And you also get a much more consistent product because they're generally not made from the same tree or the same part of the tree, it's generally lots of timber gets sawn up and then it gets all mixed up and then it gets glued back together. Any bits that are heavier, any bits that are a bit lighter, any bits that are a bit stiffer, any bits that are a bit softer, they all get mixed up with their peers. And that panel that's glued back together is way more consistent in both weights and flex pattern than if it was made from a single piece of timber. So we use vertical laminated Polonia for very good reason. So why we use vertical laminated Polonia wood rather than foam is a very good question though. And it's the. If we go way, way back in the past, some of the original production kite boards were made in the same process as wakeboarding, and wakeboarding uses at least high volume. Wakeboard production uses a polyurethane foam core. And a polyurethane foam core is a very cost effective and a very fast way of making a board shaped product, because you have a mold for your core and you actually inject into that mold a two part foaming polyurethane. So it just literally two components mixed together and injected into that mould and it foams up and fills up the inside of that mold and you get a board shaped thing basically in seconds, Rather than machining processes, which we have to do with the wood core, which can take about an hour per board to actually produce the final shape. You get this thing in seconds. And then once you've got this board shaped core, you can wrap it in fiberglass and epoxy resin and put it in another mold and press it and voila, you've got your board thing coming out, your board shaped thing coming out the other end. But we pretty early in kiteboarding development worked out that we, we want thinner, more flexible structures than traditional wakeboards. And those thinner, more flexible structures pretty soon showed how poor choice polyurethane foam is as a core material. It is in a thick, stiff structure like a commercial wakeboard, it works perfectly fine. But in a thinner, more flexible structure, the foam itself can delaminate it off itself. So it has poor shear strength. And as you flex it, it starts to break down and the board will break. So we needed something that was better than that. We moved from using the industry pretty rapidly, moved from that traditional wakeboard process to using structural like marine structural foams, predominantly cross linked PVC foam. And you'll hear words like divinocell is a, is a, used to be used as a, that as the catchphrase for a good quality PVC foam core. There are many other brands out there now, but in the early days of kiteboarding, like DivinoCell was, I think there was actually patent protection. DivinoCell was the, the brand name that was making crosslink PVC foam. And this is a much, much stronger foam than polyurethane, but it can't be molded, so it has to, it comes in a sheet form and it has to be machined and then gets laminated into a board. And for a very long time it was the, it was the goat that the, at least for the first sort of five or so years of commercial kite boarding. So 2000 to 2005, DivinoCell Crosslink PVC foam core was the thing we, the thing we used. But then we started to kite loop. And once we started a kite loop, then we started the amount of extra reinforcement you would have to put, especially in the heel area of a divinicel board, to just stop the rider just eventually punching their heels into that foam and destroying it became a real problem. And we started to look for alternative materials. And Paulownia is one of the lighter weight timbers in the world, but is actually one of the strongest timbers for its weight. And as soon as we started experimenting with paulownia cores, we realized two things. One, incredibly strong compared to a foam core, but two, incredibly lively. So the big difference between riding a foam cord twin tip and riding a wood chord twin tip is how quickly that board returns to shape when you load it up. So what a rider would describe as pop, what a designer would describe as reflex polonia wood cause give you a much stronger board with a much more lively ride, a much better pop than if we use foam. And that's why it's become the dominant core material across the across the industry.
- [33:29]
B
Well dk, I think we'll stop that this week. We are going to have you back at the end of this month because this was supposed to be February. We're still working our way through these mountains. The questions I think after the next one we'll put out a call for another ama. But DK again, awesome. Amazing. A bit of a heavy one, I think this AI one, I think we're going to have to go back to that I over the coming months and years because I think that will be as AI becomes a bigger part of our lives and it's already, you know, making big plays into our lives. Well, I think we'll dive into that and it'll be interesting to see what happens with that over the the next couple years at least. But dk, thanks again buddy and we'll see you at the end of this month.
- [34:08]
A
Yeah, cheers Adrian. And Cheers all the listeners. We'll we'll catch up with you all soon.
- [34:12]
B
Hey guys, I hope you enjoyed that episode. Don't forget, if you want to support the show, the the easiest way is to do it for free. Rate me on Spotify. I'm loving those five star reviews guys. It literally takes one second to do and I really appreciate that. Share them on the local WhatsApp or kite surfing groups or just simply tell your friends. If you want to support the podcast more regularly, head over to portraitkite.com Portrait is an independent media company trying to tell the stories of kiteboarding the way we believe they should. These projects are funded by people just like you. And if you believe in what we do, head over to portraitkite.com and check out all the madness there. The podcast guys will always be free and if you want more episodes just like this one, use the search Button at Kite Surf365 to search your favorite writer or topic. And we'll be back this Thursday for the Megapod.