1 00:00:00,240 --> 00:00:02,360 Hi, and welcome to the Neil Ashton Podcast. 2 00:00:03,000 --> 00:00:05,880 In each episode, we explained some of the fascinating ways 3 00:00:05,880 --> 00:00:09,080 that science and engineering are changing the world around us. 4 00:00:09,760 --> 00:00:12,720 We talked to leading engineers from elite level sports like 5 00:00:12,800 --> 00:00:16,840 cycling in Formula One to some of the world's top academics to 6 00:00:16,840 --> 00:00:20,960 understand how fluid dynamics, machine learning, supercomputing 7 00:00:21,360 --> 00:00:22,960 are bringing in a new era discovery. 8 00:00:23,920 --> 00:00:27,040 We also hear some of their life stories, their career advice, 9 00:00:27,640 --> 00:00:30,040 the lessons they've learned on the way that I hope will be 10 00:00:30,040 --> 00:00:33,800 helpful to you too. So sit back and enjoy this 11 00:00:33,800 --> 00:00:40,600 episode. Hi, welcome back to the new 12 00:00:40,600 --> 00:00:43,240 Ashton podcast. I thought I'd do something 13 00:00:43,240 --> 00:00:46,240 interesting today, which I was. I was looking back at the 14 00:00:46,240 --> 00:00:50,720 previous episodes and I saw that almost a year ago today I did an 15 00:00:50,720 --> 00:00:56,160 episode called the Future of CFD Five key trends to watch. 16 00:00:56,160 --> 00:01:03,160 It was season 2 episode 2 and I thought I will try and do 17 00:01:03,160 --> 00:01:07,280 another version of that and basically see a year on, was I 18 00:01:07,280 --> 00:01:12,040 right or was I wrong? So what I had as the five were 19 00:01:12,800 --> 00:01:18,480 the rise of GP, us in CFDAI and machine learning, shift to cloud 20 00:01:18,480 --> 00:01:22,360 computing, digital certification, and I think I had 21 00:01:22,360 --> 00:01:25,360 the last one which was mergers and acquisitions. 22 00:01:26,720 --> 00:01:29,720 OK, Now this was before I joined NVIDIA. 23 00:01:29,760 --> 00:01:32,960 Oh, I should just say. So obviously Nvidia's you know, 24 00:01:33,680 --> 00:01:37,040 creates GPUs. So the rise of GPUs in CFD. 25 00:01:38,360 --> 00:01:42,720 Can I answer that with an being completely unbiased? 26 00:01:42,720 --> 00:01:47,200 Probably not, but it's true. Basically, there has been an 27 00:01:47,200 --> 00:01:49,560 unbelievable rise in GPS for CFDI. 28 00:01:49,560 --> 00:01:52,040 Can say that even if I wasn't working for NVIDIA, I think 29 00:01:52,040 --> 00:01:54,240 people in the community know that to be the case. 30 00:01:54,960 --> 00:01:56,840 Has that accelerated in the past year? 31 00:01:56,880 --> 00:02:00,640 Yes, most definitely it is. Yeah. 32 00:02:00,720 --> 00:02:03,440 Just the number one question. And again, I know I work for 33 00:02:03,440 --> 00:02:06,040 NVIDIA, so I'm going to say that, but it's true. 34 00:02:06,600 --> 00:02:12,240 I go to conferences and I see it without me even talking to them. 35 00:02:12,240 --> 00:02:15,200 You know, they're showing slides saying, OK, it's 10 times faster 36 00:02:15,200 --> 00:02:18,000 on a GPUI did not get involved in that study whatsoever. 37 00:02:18,000 --> 00:02:19,280 I'm just sat there listening to it. 38 00:02:19,760 --> 00:02:25,160 So it it is absolutely the case. And as anything there was, it 39 00:02:25,160 --> 00:02:27,280 was slower at the beginning. People were like, OK, is this 40 00:02:27,280 --> 00:02:29,240 really true? You know, is this marketing? 41 00:02:29,760 --> 00:02:32,560 I think now there's been so many independent studies showing it 42 00:02:32,560 --> 00:02:34,400 that people are fully getting behind. 43 00:02:34,880 --> 00:02:38,840 And I think what's also made the difference is 1, the hardware's 44 00:02:38,880 --> 00:02:40,600 obviously keeps getting better and better. 45 00:02:41,320 --> 00:02:44,480 The pace of GP us, because obviously AI has helped. 46 00:02:44,480 --> 00:02:47,640 This is far out pacing. CP US in terms of like 47 00:02:47,640 --> 00:02:50,400 generational generation, you know, like the improvement that 48 00:02:50,400 --> 00:02:54,040 you get each generation is, is like this compared to CP US, 49 00:02:54,120 --> 00:02:55,480 which is, you know, a bit slower. 50 00:02:55,880 --> 00:02:59,640 So you're, by moving to it, you keep getting better and better. 51 00:02:59,640 --> 00:03:01,880 So the reason to move has probably accelerated. 52 00:03:03,360 --> 00:03:08,720 I think the, there's been a healthy competition between 53 00:03:08,720 --> 00:03:12,680 different CFD companies each seeing GP used to be part of 54 00:03:12,680 --> 00:03:17,160 that differentiating thing. And that has created a, a sort 55 00:03:17,160 --> 00:03:20,080 of an acceleration, you know, because there's one vendor 56 00:03:20,080 --> 00:03:22,840 releases AGP version number one wants to make sure they also 57 00:03:22,840 --> 00:03:24,840 have one. So there's been a little bit of 58 00:03:24,840 --> 00:03:28,520 that, but for good reasons. It's not purely, you know, just 59 00:03:28,520 --> 00:03:30,120 for the sake of it. It's because people are 60 00:03:30,120 --> 00:03:33,080 genuinely seeing a speed up customers liking it and and 61 00:03:33,080 --> 00:03:36,680 therefore they're doing it. And I think most interestingly, 62 00:03:37,400 --> 00:03:40,280 which is I think something I commented on is there's a lot of 63 00:03:40,280 --> 00:03:45,280 start-ups now who are specifically focusing their on 64 00:03:46,160 --> 00:03:49,160 GPU native. And that has also definitely 65 00:03:49,160 --> 00:03:52,480 played a ship because no surprise if you write code from 66 00:03:52,480 --> 00:03:56,840 scratch with a sort of software, hardware code design, you're 67 00:03:56,840 --> 00:03:58,600 going to create something really good. 68 00:03:59,200 --> 00:04:04,520 And so because of those start-ups, they have shown in 69 00:04:04,520 --> 00:04:09,960 some ways the like most optimum solution and has really shown 70 00:04:09,960 --> 00:04:11,600 some very interesting performance. 71 00:04:11,800 --> 00:04:15,640 What I find more at a technical level interesting is where the 72 00:04:15,640 --> 00:04:20,680 whole like unstructured implicit finite volume, which I guess was 73 00:04:20,680 --> 00:04:26,400 the industry standard does that. Is that the most optimum 74 00:04:26,400 --> 00:04:34,840 numerical approach on GPS? It could be, but because they're 75 00:04:34,840 --> 00:04:38,720 so much more powerful and because of that, you can start 76 00:04:38,720 --> 00:04:42,600 to do more transient high fidelity LES, then people are 77 00:04:42,600 --> 00:04:46,000 going, well, if I'm going to do an LES and I'm going to have to 78 00:04:46,000 --> 00:04:50,400 run lots of time steps and I've got this GPU, should I maybe now 79 00:04:50,400 --> 00:04:54,760 go back to explicit that previously wasn't maybe the 80 00:04:54,760 --> 00:04:58,000 optimum way of doing it, but now with the GP it is. 81 00:04:58,840 --> 00:05:02,600 And then maybe could I do a, a war modelled approach if I'm 82 00:05:02,600 --> 00:05:05,120 doing a war model, LES, and does that mean I should maybe use the 83 00:05:05,120 --> 00:05:07,400 most boundary method because I don't need to resolve the wall? 84 00:05:07,400 --> 00:05:12,120 And so that it's this interplay between the hardware and the 85 00:05:12,120 --> 00:05:15,040 software is creating some really interesting changes. 86 00:05:15,480 --> 00:05:18,640 And and finally, one of the interesting was a high order 87 00:05:18,640 --> 00:05:22,920 side, You know, I think high order, without going into too 88 00:05:22,920 --> 00:05:24,960 much of the technical details, there's some interesting stuff 89 00:05:24,960 --> 00:05:26,600 around like tensor cords, etcetera. 90 00:05:27,200 --> 00:05:30,920 So the whole high order thing is coming back in to play. 91 00:05:31,480 --> 00:05:36,680 So yeah, was my prediction of key trends for CFT correct, GPUs 92 00:05:36,680 --> 00:05:41,840 and CFT, I'd say yes. And is it still for 20/25/26? 93 00:05:42,480 --> 00:05:46,480 Definitely. It's only accelerating and I 94 00:05:46,480 --> 00:05:51,200 think you'll find that it continues to accelerate whilst 95 00:05:51,200 --> 00:05:56,720 there is so much more learnings to be had in terms of optimal 96 00:05:56,720 --> 00:05:59,160 architectures. And there's like a crossover 97 00:05:59,160 --> 00:06:01,440 point when your compute gets higher, you can go to a 98 00:06:01,440 --> 00:06:07,160 different modeling paradigm. So for example, and not wanted 99 00:06:07,160 --> 00:06:10,560 to speak about this for too long, you've gone from, let's 100 00:06:10,560 --> 00:06:16,280 say a wall resolved RANS like a low Y plus RANS where you wanted 101 00:06:16,280 --> 00:06:18,400 to resolve the ball and where you need a certain gridding 102 00:06:18,800 --> 00:06:23,200 strategy to be able to resolve the wall to a wall modelled LES 103 00:06:23,200 --> 00:06:25,440 where you say, oh, well, I don't need to resolve the wall because 104 00:06:25,440 --> 00:06:27,320 I'm modelling it. So maybe I can go to a different 105 00:06:27,320 --> 00:06:29,360 gridding paradigm and I can go to explicit. 106 00:06:30,240 --> 00:06:33,240 But if you go to the next stage, which is the war resolved Elias, 107 00:06:33,600 --> 00:06:36,640 that maybe say, well, actually now my gridding paradigm that 108 00:06:36,640 --> 00:06:41,280 was good for war modeled LES may not be good for war resolved 109 00:06:41,280 --> 00:06:42,640 LES. Maybe I need to do something 110 00:06:42,640 --> 00:06:44,760 slightly different. So it's interesting how these 111 00:06:44,760 --> 00:06:47,320 waves happen and the, like I said, doesn't interplay. 112 00:06:47,320 --> 00:06:52,040 So what I find most interesting is a year ago or two years ago, 113 00:06:52,680 --> 00:06:56,240 the most bold predictions was more towards Hybridrans, Elias, 114 00:06:56,240 --> 00:06:59,480 war modeled and everyone said our war resolved is too far 115 00:06:59,480 --> 00:07:01,920 away. I'm actually seeing now people 116 00:07:01,920 --> 00:07:04,960 be a bit bolder and say, well, you know what with the latest GP 117 00:07:04,960 --> 00:07:09,120 us maybe a war resolved Elias, it started to become possible. 118 00:07:09,320 --> 00:07:13,120 And that's interesting. I think not for everyday 119 00:07:13,120 --> 00:07:16,880 applications perhaps, but now that that is coming up, some 120 00:07:16,880 --> 00:07:19,480 other stuff starts to get interesting like just the sheer 121 00:07:19,480 --> 00:07:21,040 number of time steps you need to do. 122 00:07:22,400 --> 00:07:26,760 And yeah, so still important. OK, AI, machine learning. 123 00:07:26,800 --> 00:07:29,920 Well, duh. Yeah, that's clearly still a key 124 00:07:29,920 --> 00:07:32,600 trend. Has things moved up in the last 125 00:07:32,600 --> 00:07:33,640 year? Absolutely. 126 00:07:33,640 --> 00:07:36,760 I'd say one of the big breakthroughs I personally seen 127 00:07:37,240 --> 00:07:41,240 is probably a year ago or a little bit four year ago, one of 128 00:07:41,240 --> 00:07:45,560 the bottlenecks seemed to be volume prediction using AI 129 00:07:45,560 --> 00:07:48,720 surrogates. Stuff was coming out around like 130 00:07:48,720 --> 00:07:51,440 graph neural Nets and being able to predict things, but all the 131 00:07:51,440 --> 00:07:53,720 examples were always on smaller cases. 132 00:07:54,560 --> 00:07:57,800 And one of the things that I did with some colleagues was to 133 00:07:57,800 --> 00:08:01,600 create that drive ML data set, which was really put out there 134 00:08:01,600 --> 00:08:07,760 as a data set to give realistically size meshes and 135 00:08:08,360 --> 00:08:10,120 hopefully encourage people to see, right. 136 00:08:10,120 --> 00:08:16,480 Can my method take the entire volume and a year or certainly 137 00:08:16,480 --> 00:08:19,040 two years ago, it looked like not many people could do it. 138 00:08:19,040 --> 00:08:22,920 Everyone was just doing the surface or maybe a few slices 139 00:08:22,920 --> 00:08:25,240 and every time you saw something, you know in marketing 140 00:08:25,240 --> 00:08:28,000 online, it was always like the surface of the car or a slice of 141 00:08:28,000 --> 00:08:30,440 the car. What I found was breakthrough in 142 00:08:30,440 --> 00:08:34,600 the past year is peep various methods have come out and I 143 00:08:34,600 --> 00:08:39,960 guess the most, most most recent will be like Transformers that 144 00:08:40,120 --> 00:08:42,200 seem to be the most computationally efficient. 145 00:08:42,559 --> 00:08:45,280 Efficient for doing full volume predictions as in training on a 146 00:08:45,280 --> 00:08:48,680 full volume and then doing an inference and being able to get 147 00:08:48,680 --> 00:08:52,040 a full volume out. I feel like that really 148 00:08:52,040 --> 00:08:55,600 accelerated the realization that some of these AI methods could 149 00:08:55,600 --> 00:08:59,520 give you back a flow field that was more similar to a 150 00:08:59,520 --> 00:09:02,880 traditional one. And yeah, that transition from 151 00:09:02,880 --> 00:09:07,240 like graphs to maybe neural operators and Transformers for 152 00:09:07,240 --> 00:09:08,920 me has been a really interesting 1. 153 00:09:08,920 --> 00:09:11,680 And I'm certainly excited to see what happens in the next year. 154 00:09:12,240 --> 00:09:15,880 I think we'll continue to see new architectures, new evolution 155 00:09:16,000 --> 00:09:18,840 of, of AI. So that is definitely a trend 156 00:09:18,840 --> 00:09:21,840 that's continuing. Yeah. 157 00:09:21,960 --> 00:09:23,240 And they're obviously linked, right. 158 00:09:23,240 --> 00:09:26,200 So if you have a faster solver, you can generate data more 159 00:09:26,200 --> 00:09:27,920 easily. You can then do the training. 160 00:09:28,200 --> 00:09:30,720 I think one of the big discussions now is on the right 161 00:09:30,720 --> 00:09:32,960 code architecture. Sorry, like programming 162 00:09:32,960 --> 00:09:36,000 environment, because online training is definitely a hot 163 00:09:36,000 --> 00:09:37,920 topic. You know, how can you take 164 00:09:37,920 --> 00:09:40,280 advantage of the transient nature of the flow? 165 00:09:41,120 --> 00:09:43,640 So at the moment, I'd say the state-of-the-art is normally 166 00:09:43,640 --> 00:09:47,680 that you would just generate those data and even if it was 167 00:09:47,680 --> 00:09:50,720 with a hybrid brand or well modelled, you'll just take a 168 00:09:50,720 --> 00:09:52,720 time average and dump that and train on that. 169 00:09:53,280 --> 00:09:56,400 But I think what's emerging now is this interesting, well, can I 170 00:09:56,400 --> 00:10:00,360 use some of the transient data that's being created anyway, you 171 00:10:00,360 --> 00:10:02,600 know, dump them out at checkpoint and train on that. 172 00:10:04,600 --> 00:10:08,160 So I think the whole AI will continue and continue. 173 00:10:08,440 --> 00:10:11,800 The one thing that I would probably add to that AI bucket 174 00:10:11,800 --> 00:10:16,520 is the agentic AI side. So let's say you've got 175 00:10:16,520 --> 00:10:20,080 surrogates that's in one market and you've got LLMS for just 176 00:10:20,080 --> 00:10:22,720 general productivity gains and coding on the other. 177 00:10:23,400 --> 00:10:26,720 I put a gentic AI somewhere in the middle, which is how can you 178 00:10:26,720 --> 00:10:30,640 have multiple agents working autonomously through some sort 179 00:10:30,640 --> 00:10:34,640 of master agent that can go off and do tasks for you. 180 00:10:34,720 --> 00:10:37,640 So instead of you having to manually click run this 181 00:10:37,640 --> 00:10:41,160 simulation or write a Python code that goes and runs, you'd 182 00:10:41,160 --> 00:10:44,880 said, give a text prompt saying I would like to go and run this 183 00:10:44,880 --> 00:10:47,520 simulation with these parameters and please create me a report. 184 00:10:48,200 --> 00:10:50,760 And you'll have a workflow where an agent will interpret that 185 00:10:50,760 --> 00:10:54,760 request will then go and let's say create a simulation setup 186 00:10:54,760 --> 00:10:57,000 for you. But crucially also then run the 187 00:10:57,000 --> 00:11:01,800 simulation, let's say an HPC cluster, use a separate, you 188 00:11:01,800 --> 00:11:04,920 know, agent could be a surrogate model essentially, you know, to 189 00:11:04,920 --> 00:11:07,840 go run it or to do the meshing, then another LLM, maybe do the 190 00:11:07,840 --> 00:11:10,720 analysis of the results. And the point being is that you 191 00:11:10,720 --> 00:11:12,480 have to kick that all off with a textbook. 192 00:11:13,360 --> 00:11:16,040 That is the most cutting edge example. 193 00:11:16,040 --> 00:11:19,040 And that's what I feel is probably in the next 12 months 194 00:11:19,040 --> 00:11:22,000 will be the next big jump. And some start-ups are already 195 00:11:22,000 --> 00:11:25,200 doing it as some companies, but I feel that's probably one of 196 00:11:25,200 --> 00:11:28,240 the big next steps. I should add, by the way, for GP 197 00:11:28,240 --> 00:11:31,800 US, I didn't mention, I think precision is a very interesting 198 00:11:31,800 --> 00:11:35,400 topic now. I think there's huge gains if 199 00:11:35,400 --> 00:11:39,880 CFD can use lower precision for for obvious reasons, you know, 200 00:11:39,880 --> 00:11:45,280 that the so much flops in in the lower precision because of AI 201 00:11:45,280 --> 00:11:48,160 generally only needs, you know, 48 or 16. 202 00:11:48,560 --> 00:11:52,400 So it doesn't mean that you can't do 64 or 32, but if you 203 00:11:52,400 --> 00:11:56,200 can find a way to harness that power, then you can really 204 00:11:56,200 --> 00:12:00,480 unlock the next big accelerate. So the first one I had was the 205 00:12:00,480 --> 00:12:04,640 shift to cloud computing. Now, you know, people who know 206 00:12:04,640 --> 00:12:07,640 me or listened to this before know that you know, I currently 207 00:12:07,640 --> 00:12:10,280 work from video, but I used to work for Amazon Web Services, 208 00:12:10,280 --> 00:12:12,600 which is obviously a massive cloud computing company. 209 00:12:13,280 --> 00:12:17,200 And there is no doubt whatsoever that cloud computing has totally 210 00:12:17,200 --> 00:12:20,880 transformed the IT sector and the HPC sector. 211 00:12:22,080 --> 00:12:27,560 When I left AWSI definitely saw an increasing momentum in people 212 00:12:27,560 --> 00:12:34,280 adopting the cloud cause of the fact that they could essentially 213 00:12:34,920 --> 00:12:38,480 focus on the stuff that mattered to them and not on the sort of 214 00:12:38,480 --> 00:12:44,960 manual data centre stuff. I still see that increasing, but 215 00:12:44,960 --> 00:12:49,080 probably the bit that maybe I was not brainwashed because I'm 216 00:12:49,080 --> 00:12:53,560 joking aside a little bit. But you I definitely, I'm seeing 217 00:12:54,520 --> 00:12:59,480 I'm still a more hybrid sense where on one hand people are 218 00:12:59,480 --> 00:13:03,600 absolutely seeing the value of the cloud, you know, through 219 00:13:03,600 --> 00:13:06,680 ADBS or GCP or Microsoft or you know, all the cloud plates. 220 00:13:07,120 --> 00:13:08,760 And by the way, there's an interesting angle now with all 221 00:13:08,760 --> 00:13:13,400 these neo clouds like core weave, etcetera, which is mainly 222 00:13:13,400 --> 00:13:18,400 I guess to the AI some GPU angle, but I'm still seeing a 223 00:13:19,280 --> 00:13:21,000 hybrid used. You know, there's still quite a 224 00:13:21,000 --> 00:13:24,480 lot of on Prem, there's cloud and and I still think there's a 225 00:13:24,480 --> 00:13:30,760 lot of room for improvements in the way to access cloud 226 00:13:30,760 --> 00:13:35,200 resources. You know, I think SAS from ACFD 227 00:13:35,240 --> 00:13:38,440 point of view now SAS products are still not fully there. 228 00:13:38,920 --> 00:13:42,880 You know, if you ask most people are they doing it software 229 00:13:42,880 --> 00:13:45,680 service? No, they may use the cloud to 230 00:13:45,680 --> 00:13:49,120 have a cluster, but are they doing it in a true SAS like 231 00:13:49,120 --> 00:13:52,880 manner? I think that it's slower than I 232 00:13:52,880 --> 00:13:57,720 thought people still seem to prefer and there's good reason. 233 00:13:57,720 --> 00:14:04,560 So we'd have to get into of having a local experience in 234 00:14:04,560 --> 00:14:06,640 that you're not just doing through a web browser. 235 00:14:06,640 --> 00:14:10,040 And I think that's probably just decades of engineers being used 236 00:14:10,040 --> 00:14:14,240 to just SSH ING and stuff and and having like, yeah, running 237 00:14:14,240 --> 00:14:16,840 scripts. And it's something that you 238 00:14:16,840 --> 00:14:20,880 think would change, but it, it does seem remarkably stubborn. 239 00:14:21,240 --> 00:14:27,840 And yeah, so I, I definitely see the cloud continuing, but I'm 240 00:14:27,840 --> 00:14:30,480 still waiting for some company to come out there and really 241 00:14:30,480 --> 00:14:34,040 make it easy from a workflow point of view in like a true 242 00:14:34,040 --> 00:14:38,120 hybrid way or to just simplify maybe the cloud experience. 243 00:14:38,120 --> 00:14:40,120 You still have to be quite an expert to do it. 244 00:14:40,360 --> 00:14:43,800 So I am, I'm not changing that prediction. 245 00:14:44,360 --> 00:14:49,760 I'm just seeing that it seems to be a bit slower than I fully 246 00:14:49,760 --> 00:14:51,840 appreciated. And maybe in hindsight, that's 247 00:14:51,840 --> 00:14:56,440 because I was seeing all of the positive and success stories at 248 00:14:56,440 --> 00:14:59,760 ADBS and maybe I was not hearing the bits where people weren't, 249 00:15:00,480 --> 00:15:07,640 you know, doing it. The 4th 1 I had was on digital 250 00:15:07,640 --> 00:15:12,440 certification to a higher fidelity methods that, as I 251 00:15:12,440 --> 00:15:15,480 mentioned on the 3rd of it is definitely the case. 252 00:15:15,520 --> 00:15:20,680 You know, you're having a huge increase of people going to 253 00:15:20,680 --> 00:15:24,120 higher fidelity methods because of #1 you know, the fact that 254 00:15:24,120 --> 00:15:26,160 you can access to compute more easily. 255 00:15:26,800 --> 00:15:30,480 And that is definitely a topic that I think will continue to 256 00:15:30,480 --> 00:15:34,840 increase sector by sector. So the automotive has definitely 257 00:15:34,840 --> 00:15:37,440 pioneered that and all have moved to higher fidelity 258 00:15:37,440 --> 00:15:39,120 methods. I think now, like I said, 259 00:15:39,120 --> 00:15:42,240 they're, they're looking how can I go from hybrid RANS to war 260 00:15:42,240 --> 00:15:44,760 resolved earlier. I think the aerospace sector is 261 00:15:44,760 --> 00:15:49,240 probably still trying to get from RANS, you know, to hybrid 262 00:15:49,240 --> 00:15:52,240 random warm up LES and the work that you know, we did in the 263 00:15:52,240 --> 00:15:55,840 last high lift pitch workshop has definitely helped to push 264 00:15:55,840 --> 00:15:58,400 that forward. And there's some interesting 265 00:15:58,400 --> 00:16:00,920 work going on now to try and push it even further again to 266 00:16:00,920 --> 00:16:04,160 like more resolved LES. But I see that continually, you 267 00:16:04,160 --> 00:16:09,840 know, increasing and now it's more around building up best 268 00:16:09,840 --> 00:16:13,320 practices and understanding of those methods, you know, 269 00:16:13,400 --> 00:16:16,000 robustness that everybody knows where they go wrong, where they 270 00:16:16,000 --> 00:16:18,640 go right. So it probably will take time 271 00:16:18,920 --> 00:16:22,840 for that and for people to see the value because it is 272 00:16:22,840 --> 00:16:25,480 undeniably more expensive to do with the low fidelity. 273 00:16:25,480 --> 00:16:27,680 So I think there needs to be success stories that have been 274 00:16:27,680 --> 00:16:30,000 created. But I think probably one of the 275 00:16:30,000 --> 00:16:31,640 biggest challenges is transition. 276 00:16:31,760 --> 00:16:34,880 Certainly I'm speaking more from ACFD and external aero point of 277 00:16:34,880 --> 00:16:38,760 view, but one of the challenges still with even going to war 278 00:16:38,760 --> 00:16:41,120 monthly yes to hybrid is transition modelling. 279 00:16:41,760 --> 00:16:45,160 And I, I feel that that's why the war resolved LES is still 280 00:16:45,160 --> 00:16:48,200 very tempting because it doesn't fully solve it, but it, it 281 00:16:48,200 --> 00:16:50,880 definitely helps to overcome some of those challenges. 282 00:16:51,920 --> 00:16:54,200 So I think that will continue, but I would look more at how do 283 00:16:54,200 --> 00:16:58,160 we make war resolved LES approaches more, more 284 00:16:58,160 --> 00:17:01,480 affordable. I am going to be doing an 285 00:17:01,480 --> 00:17:05,119 episode where we're going to talk more on the combustion side 286 00:17:05,119 --> 00:17:07,359 of things. I do realise that this podcast 287 00:17:07,359 --> 00:17:10,000 just tend to be a little bit focused on external arrow. 288 00:17:10,200 --> 00:17:14,119 That's basically my background. So I am purposely trying to, 289 00:17:14,119 --> 00:17:17,920 yeah, diversify a little bit and make sure that we're covering 290 00:17:17,920 --> 00:17:24,000 other areas. And the final one was mergers 291 00:17:24,160 --> 00:17:26,200 and acquisitions and innovations. 292 00:17:26,280 --> 00:17:29,880 And this has definitely kept going. 293 00:17:31,120 --> 00:17:37,480 We have seen ANSYS and Synopsys, so two big companies merge. 294 00:17:38,080 --> 00:17:45,680 We have seen, you know Siemens, Altair, Cadence, Beta, CAE. 295 00:17:46,000 --> 00:17:50,440 So this has kept going and was a trend that I predicted and is 296 00:17:50,440 --> 00:17:54,800 still true today that there is so much interest in the 297 00:17:55,400 --> 00:17:58,760 community aid in engineering and semiconductor businesses that, 298 00:17:58,760 --> 00:18:02,600 you know, it's a growing business, really important. 299 00:18:02,840 --> 00:18:06,400 And big companies are realizing the value of acquiring or 300 00:18:06,400 --> 00:18:08,640 merging with others to, you know, bring some of their 301 00:18:08,640 --> 00:18:12,360 technology in and speed up. You know, that that time. 302 00:18:13,120 --> 00:18:15,760 I think that will continue. I would say the prediction for 303 00:18:15,760 --> 00:18:18,800 next year, it's not there's not many other big companies that 304 00:18:18,800 --> 00:18:21,640 can merge now. You know, it has consolidated a 305 00:18:21,760 --> 00:18:27,280 little bit to Cadence, Siemens and, and synopsis, but where I 306 00:18:27,280 --> 00:18:30,560 do see that being definitely move is there's a lot of 307 00:18:30,560 --> 00:18:36,520 start-ups and I'm sure that the many of those start-ups will be 308 00:18:36,520 --> 00:18:38,440 acquired by some of these bigger companies. 309 00:18:38,600 --> 00:18:42,360 So they can start to really bring those to a more enterprise 310 00:18:42,360 --> 00:18:44,680 level and integrate it. So I, my prediction would be in 311 00:18:44,680 --> 00:18:47,480 the next 12 months that some of the start-ups that you, you 312 00:18:47,480 --> 00:18:51,880 know, you may know of, I won't mention all names, but the 313 00:18:51,880 --> 00:18:54,160 start-ups, you know, I think you'll see some of them being 314 00:18:54,160 --> 00:18:57,960 acquired by high profile companies just because that is, 315 00:18:58,840 --> 00:19:00,760 that's a tried and tested route. You know, when I spoke to the 316 00:19:00,760 --> 00:19:03,560 CTO of Ansys and he openly admitted it, he said, that's 317 00:19:03,560 --> 00:19:05,520 what we do. We look for great start-ups. 318 00:19:05,520 --> 00:19:07,200 We may even, you know, encourage them. 319 00:19:07,200 --> 00:19:09,960 And once they reach a certain level, we might acquire it. 320 00:19:09,960 --> 00:19:12,560 And it, that's a great way of developing new technology. 321 00:19:12,560 --> 00:19:14,920 Sometimes it's better to let a start up do it. 322 00:19:15,480 --> 00:19:17,640 You have a, you know, faster way of doing than a, let's say, a 323 00:19:17,640 --> 00:19:19,640 big enterprise. But then they acquire them, they 324 00:19:19,640 --> 00:19:21,600 bring them, and that's how customers get to use them. 325 00:19:22,440 --> 00:19:23,560 So I think they also still continue. 326 00:19:24,600 --> 00:19:28,000 So where do I think things are going? 327 00:19:28,000 --> 00:19:31,040 Well, I think actually those five topics will still continue 328 00:19:31,360 --> 00:19:37,880 to be a thing for 25 and 26, and I'm excited to see how they 329 00:19:37,880 --> 00:19:40,360 accelerate. So I hope that was interesting. 330 00:19:40,360 --> 00:19:42,680 I thought I'd keep it short. I know these podcasts tend to go 331 00:19:42,680 --> 00:19:45,360 to an hour or two hours, so I thought I'll keep this one short 332 00:19:45,760 --> 00:19:49,200 and got a couple of really good guests coming up for the next 333 00:19:49,200 --> 00:19:52,560 two episodes. So yeah, look out for those. 334 00:19:52,600 --> 00:19:54,440 And for now, I hope you enjoyed this episode.