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 and 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 of 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 --> 00:00:41,160 episode. Hi and welcome back to the Neil 12 00:00:41,160 --> 00:00:43,840 Ashton Podcast. It's been a while, but I'm 13 00:00:43,840 --> 00:00:48,960 excited to start a new season and series and I'm going to 14 00:00:48,960 --> 00:00:54,480 start with a discussion on how I think agents and foundation 15 00:00:54,480 --> 00:00:59,440 models are are really a disruptor to the current CAE and 16 00:00:59,440 --> 00:01:03,040 EDA ecosystem. This, this is of course is just 17 00:01:03,040 --> 00:01:05,480 a personal opinion. This does not represent my 18 00:01:05,480 --> 00:01:08,960 current employer, NVIDIA, but I wanted to share some, some 19 00:01:08,960 --> 00:01:13,000 thoughts I've really had over the past six months of how I see 20 00:01:13,000 --> 00:01:16,280 this technology being, you know, obviously a massive opportunity, 21 00:01:16,600 --> 00:01:20,600 but also also a disruptor. And so we'll, we'll walk through 22 00:01:20,600 --> 00:01:24,160 that today from a technical point of view, from a commercial 23 00:01:24,160 --> 00:01:26,440 point of view, from a scientific point of view. 24 00:01:26,880 --> 00:01:29,400 And I'd certainly be interested to see whether you agree with 25 00:01:29,400 --> 00:01:34,760 my, my thinking on this. If we take a step back and you 26 00:01:34,760 --> 00:01:38,360 know, I'm going to look at as often I do in this, in this 27 00:01:38,360 --> 00:01:42,360 podcast through a sort of CFD lens, but I think much of what 28 00:01:42,360 --> 00:01:47,320 I'm saying is relevant across CAE and and also EDA for chip 29 00:01:47,320 --> 00:01:51,200 design, etcetera. If you look, where did things 30 00:01:51,200 --> 00:01:54,280 begin? I would argue in the 60s and the 31 00:01:54,280 --> 00:01:57,800 70s people were writing their own codes. 32 00:01:58,000 --> 00:02:02,920 Now, I've had episodes with Professor Jameson and others who 33 00:02:03,680 --> 00:02:08,240 who started writing codes themselves before modern 34 00:02:08,240 --> 00:02:13,320 supercomputing was around, and those often started at academic 35 00:02:13,320 --> 00:02:16,280 institutions. Those codes then obviously 36 00:02:16,280 --> 00:02:19,400 started to become useful and there was collaborations with 37 00:02:19,400 --> 00:02:22,360 industry, particularly in the aerospace initially. 38 00:02:23,000 --> 00:02:26,840 And that is why that then those aerospace companies started to 39 00:02:26,840 --> 00:02:30,880 work closely with those developers and eventually 40 00:02:31,000 --> 00:02:33,520 started to take them as their own codes. 41 00:02:34,040 --> 00:02:37,280 And that's why you see today even there's a legacy of 42 00:02:37,280 --> 00:02:40,200 aerospace companies having their own codes written from scratch, 43 00:02:40,760 --> 00:02:42,840 frankly, because there were no other options. 44 00:02:42,880 --> 00:02:47,200 And that was the way to do it. The automotive sector then took 45 00:02:47,320 --> 00:02:50,240 some things from the aerospace sector. 46 00:02:50,440 --> 00:02:52,200 And this depends on this a little bit, whether we're 47 00:02:52,200 --> 00:02:54,680 talking about structural mechanics or fluid mechanics, 48 00:02:55,760 --> 00:02:59,200 but it started off being you had to write it yourself with some 49 00:02:59,600 --> 00:03:04,160 expert developer. Then over the course of, you 50 00:03:04,160 --> 00:03:08,400 know, the 70s, eighties and early 90s, those codes started 51 00:03:08,400 --> 00:03:11,600 to become commercialized. Those professors out of places 52 00:03:11,600 --> 00:03:17,120 like Imperial or MIT or Stanford started to, to, to create 53 00:03:17,120 --> 00:03:20,240 companies. And, and ultimately those are 54 00:03:20,240 --> 00:03:24,880 the, the lineage, the lineage of where Abacus comes from, where 55 00:03:24,880 --> 00:03:27,520 Fluent comes from, where Openfront comes from. 56 00:03:28,920 --> 00:03:33,480 And in some ways those commercialization was because it 57 00:03:33,480 --> 00:03:38,040 was a natural thing as one computing started to become and 58 00:03:38,040 --> 00:03:42,360 make these tools more than just a, a niche application. 59 00:03:42,720 --> 00:03:46,040 HPC facilities, you know, Beowulf cluster decided to 60 00:03:46,040 --> 00:03:51,280 become commonplace and and of course, industry started to see 61 00:03:51,280 --> 00:03:53,200 the benefit of those simulation approaches. 62 00:03:53,720 --> 00:03:57,120 And frankly, then there was also a competitive nature that once 63 00:03:57,120 --> 00:04:00,560 one company starts, you start having competitions between, you 64 00:04:00,560 --> 00:04:04,200 know, the CD adapt Cos and the, and what what is now the, the 65 00:04:04,200 --> 00:04:06,720 Ansys and the later the Dassaults. 66 00:04:06,720 --> 00:04:11,800 There was that growing ecosystem of companies that that compete 67 00:04:11,800 --> 00:04:13,600 against each other, that accelerates it. 68 00:04:13,600 --> 00:04:17,959 The computing starts to get better and other than those main 69 00:04:17,959 --> 00:04:21,399 players, you also started to have the hardware companies get 70 00:04:21,399 --> 00:04:24,200 involved. So the Intel's, the AMD, the 71 00:04:24,200 --> 00:04:28,200 Dell, the Lenovo, the Cray in those days, the the main people 72 00:04:28,200 --> 00:04:31,640 building those those systems, those supercomputing systems. 73 00:04:32,400 --> 00:04:37,840 And if you look at how it happened, that was probably true 74 00:04:37,840 --> 00:04:42,120 for, for maybe 5 to 10 years or so where it was mainly a 75 00:04:42,360 --> 00:04:46,120 creating software to run on workstations, servers and 76 00:04:46,120 --> 00:04:49,320 supercomputers. And that, you know, spread 77 00:04:49,320 --> 00:04:52,440 across the world, frankly. And those tools were very widely 78 00:04:52,440 --> 00:04:56,360 used and started to create a real, you know, hundreds of 79 00:04:56,360 --> 00:04:58,800 millions or even billions by then industry. 80 00:05:00,200 --> 00:05:05,120 The the next technology that really came around was then 81 00:05:05,120 --> 00:05:09,720 around like cloud computing. So cloud computing maybe 5-6, 82 00:05:09,720 --> 00:05:14,800 seven years ago was, was a disruptor because it allowed 83 00:05:14,840 --> 00:05:17,480 and, and obviously I'm skipping huge other things that happened, 84 00:05:17,480 --> 00:05:20,080 but I don't want to spend 2 hours just giving the history 85 00:05:20,080 --> 00:05:24,480 of, of CFD. The reason I pick out cloud, and 86 00:05:24,480 --> 00:05:29,040 I saw it first time working for AWS, is it changed it from being 87 00:05:29,040 --> 00:05:35,400 a purely software license point of view to where these providers 88 00:05:35,400 --> 00:05:37,360 could also have a control of the hardware. 89 00:05:38,440 --> 00:05:41,880 Packaging software and hardware together is attractive for a 90 00:05:41,880 --> 00:05:47,040 company for an ISV because they can make profit off the hardware 91 00:05:47,400 --> 00:05:50,680 and they can lock it together both from a performance point of 92 00:05:50,680 --> 00:05:53,240 view, but also from a sort of stickiness point of view. 93 00:05:53,920 --> 00:05:57,640 And there's advantages for the customer in terms of making it 94 00:05:57,640 --> 00:06:02,280 more seamless because often if you just ship a binary, they 95 00:06:02,280 --> 00:06:04,480 have to be the one to then figuring out how to get the 96 00:06:04,480 --> 00:06:06,480 hardware. They are the ones who've got to 97 00:06:06,640 --> 00:06:09,200 think about, do they need to buy twice as much hardware? 98 00:06:09,200 --> 00:06:12,920 If they want to run twice as many simulations in the theory 99 00:06:12,920 --> 00:06:17,160 of the ultimate SAS, then it is fully scalable. 100 00:06:17,160 --> 00:06:21,520 So if you're running, you know, star CCM or fluent or power flow 101 00:06:21,520 --> 00:06:26,200 or whatever, you could essentially say, I want to run 102 00:06:26,200 --> 00:06:29,280 10 times more and through the, you know, infinite scale or 103 00:06:29,280 --> 00:06:31,560 inferior of the cloud you can do in the back end. 104 00:06:32,200 --> 00:06:37,760 So that sassification was, was something started maybe 6-7 105 00:06:37,760 --> 00:06:40,520 years ago. And, and obviously companies 106 00:06:40,520 --> 00:06:44,160 like Rescale were one of the first to do that at a cross 107 00:06:44,160 --> 00:06:46,760 software level. And there's others like Total, 108 00:06:46,760 --> 00:06:49,960 CAE, etcetera who tried to, to bring it together. 109 00:06:51,480 --> 00:06:56,400 That is where I would say AWS and Microsoft and Google started 110 00:06:56,400 --> 00:07:02,080 to enter into more seriously the CAE and EDA market as important 111 00:07:02,280 --> 00:07:05,120 partners. Because frankly, now it wasn't 112 00:07:05,120 --> 00:07:08,120 just The Dells and Lenovos, but you were deploying it on AWS or 113 00:07:08,120 --> 00:07:11,720 GCP or or Azure or or Oracle, for example. 114 00:07:12,320 --> 00:07:15,000 Now GPUs obviously had been around for quite a few years 115 00:07:15,000 --> 00:07:18,520 before that, but but arguably it's only been in the past sort 116 00:07:18,520 --> 00:07:23,440 of three, maybe four years where they have risen to become a 117 00:07:23,440 --> 00:07:25,560 really important part of the ecosystem. 118 00:07:26,160 --> 00:07:28,800 And frankly, that's because a lot of people have seen quite 119 00:07:28,800 --> 00:07:31,000 clear benefits and they're moving to the GPUs. 120 00:07:31,680 --> 00:07:35,280 And that's another technology point that's important because 121 00:07:36,000 --> 00:07:41,120 it added a new competitive angle to this because they needed to 122 00:07:41,200 --> 00:07:47,040 able to run things on a GPU, which meant that if you could 123 00:07:47,040 --> 00:07:49,400 optimize your code for GPU can get advantage. 124 00:07:49,400 --> 00:07:52,200 This is where flex compute, this is where luminary cloud started 125 00:07:52,200 --> 00:07:54,680 to come out. The rise of, you know, VC 126 00:07:54,680 --> 00:07:58,680 funding for these sort of companies led to a resurgence of 127 00:07:58,680 --> 00:08:01,240 competitiveness, I would say, in that market. 128 00:08:03,000 --> 00:08:05,840 And I the reason I point to the cloud before is that many of 129 00:08:05,840 --> 00:08:09,560 those companies lent on as the cloud being an additional sort 130 00:08:09,560 --> 00:08:12,040 of angle to their to their offering. 131 00:08:12,520 --> 00:08:19,480 So at that point another thing happens which is that the bigger 132 00:08:19,480 --> 00:08:24,800 companies start to look to consolidate and to start their 133 00:08:24,800 --> 00:08:28,040 sort of acquisition period. So you saw companies like 134 00:08:28,040 --> 00:08:31,520 Siemens and the timing of these are not all correct, but you 135 00:08:31,520 --> 00:08:37,679 know Siemens and Altair, Ansys and Synopsys, Cadence and Beta 136 00:08:37,679 --> 00:08:40,600 CAE and point wise are future facilities. 137 00:08:41,440 --> 00:08:44,680 Many of these companies started to acquire each other. 138 00:08:45,720 --> 00:08:51,040 And why was that done? Well actually so those companies 139 00:08:51,600 --> 00:08:55,600 then wanted to start quiet and of course this is where one of 140 00:08:55,600 --> 00:08:59,920 the key technology came in that starts to get to my point here, 141 00:09:00,760 --> 00:09:05,080 which is the AI side. It was, it has been around to be 142 00:09:05,080 --> 00:09:08,480 fair, for maybe six years or even 7 years. 143 00:09:09,640 --> 00:09:14,680 But really it was since the sort of ChatGPT moment in the early 144 00:09:14,720 --> 00:09:19,720 twenty 20s, 2021-2022 that brought this to life. 145 00:09:19,720 --> 00:09:24,920 And so, you know, they started some new startups to really get 146 00:09:24,920 --> 00:09:30,640 hold and the newer concept, the Physics X, the Navisto Extraity 147 00:09:31,720 --> 00:09:36,400 MEAI, these companies started to come about offering the 148 00:09:36,400 --> 00:09:40,520 surrogate modelling capability, the ability to do real time 149 00:09:40,520 --> 00:09:44,400 simulations. And that frankly was a threat 150 00:09:44,720 --> 00:09:48,120 and an opportunity. And I say that because that is 151 00:09:48,120 --> 00:09:50,400 what I've heard when in conversations. 152 00:09:50,560 --> 00:09:53,840 It's a threat because anything new is by definition a potential 153 00:09:53,840 --> 00:09:56,440 threat. These software companies who 154 00:09:56,440 --> 00:10:00,720 have, you know, obviously had quite a, a good position, see as 155 00:10:00,720 --> 00:10:01,840 you know, how do we deal with this? 156 00:10:02,120 --> 00:10:04,800 But clearly it's also a massive opportunity because it brings 157 00:10:04,800 --> 00:10:08,000 new technology, new capabilities to, to, to customers. 158 00:10:08,760 --> 00:10:12,560 And that it has been a very interesting thing to observe in 159 00:10:12,560 --> 00:10:16,240 the past, let's say 3 or 4 years of how do these larger companies 160 00:10:16,240 --> 00:10:20,080 deal with it? How do companies deal with it 161 00:10:20,080 --> 00:10:23,360 who have their own codes? How do they respond? 162 00:10:23,360 --> 00:10:26,240 Do they respond with hype? Do they respond with negativity, 163 00:10:26,240 --> 00:10:29,840 with positivity? And what happened was then some 164 00:10:29,840 --> 00:10:33,800 companies decided to start developing themselves, some 165 00:10:33,800 --> 00:10:36,920 started to buy. So Ansys bought Extrality, which 166 00:10:36,920 --> 00:10:40,080 is now SIM AI. Autodesk bought Navistone, 167 00:10:40,920 --> 00:10:44,880 Siemens, you could argue and some part of it was the Alta 168 00:10:44,960 --> 00:10:48,720 acquisition. And then some of these start off 169 00:10:48,760 --> 00:10:53,560 start-ups that started with a more GPU focus started to pivot 170 00:10:53,560 --> 00:10:55,760 quite heavily to the surrogate modelling. 171 00:10:55,760 --> 00:11:00,320 So Lumia Cloud or Lumia AI now and Flex compute, both of you 172 00:11:00,320 --> 00:11:02,520 know pivoted to, to, to that direction. 173 00:11:03,720 --> 00:11:08,680 And so we, so we're at that point now where the, the 174 00:11:09,080 --> 00:11:14,400 ecosystem is at a very influential point for a sort of 175 00:11:14,400 --> 00:11:17,840 inflection point. And there's one final bit that 176 00:11:17,840 --> 00:11:20,520 is important to that. And it's actually ultimately the 177 00:11:20,520 --> 00:11:24,080 main topic that I want to speak about today, which is agents. 178 00:11:24,600 --> 00:11:30,120 Because the missing piece to some of this has been if you 179 00:11:30,120 --> 00:11:34,560 create a surrogate model and you take it to its fullest extent 180 00:11:34,560 --> 00:11:37,680 and you have some sort of foundation model, you haven't 181 00:11:38,360 --> 00:11:41,360 really got rid of the need to do CAD. 182 00:11:41,480 --> 00:11:46,040 You haven't got rid of the need to do some sort of preprocessing 183 00:11:46,040 --> 00:11:48,880 or post processing. You could argue you've just got 184 00:11:49,080 --> 00:11:54,760 a faster plug in replacement for the for the bit of the solver or 185 00:11:54,760 --> 00:11:58,160 the solve part, but you still need all the bit around the 186 00:11:58,160 --> 00:12:01,840 rest. And that's why it's ultimately 187 00:12:01,840 --> 00:12:06,200 been more of an integration play through some of these AI 188 00:12:06,200 --> 00:12:12,000 start-ups because you're not going to replace CFD or FEA or 189 00:12:12,040 --> 00:12:14,640 Eva, you're just really creating an additional tool. 190 00:12:15,320 --> 00:12:18,800 So in that sense you can understand why the larger Isvs 191 00:12:18,800 --> 00:12:23,200 and others have not necessarily seen this surrogate modelling as 192 00:12:23,200 --> 00:12:26,520 a massive threat initially because they know that on it's 193 00:12:26,520 --> 00:12:32,800 own it's it's not a replacement that has led to of course 1 194 00:12:32,800 --> 00:12:36,000 angle, which is if you were to make a model that was so 195 00:12:36,000 --> 00:12:41,360 generalizable, then it could given the right plug and play, 196 00:12:41,640 --> 00:12:44,360 disrupt some solver only companies. 197 00:12:45,000 --> 00:12:49,600 If you only produce a solver and not the CAD and maybe not the 198 00:12:49,600 --> 00:12:54,400 preprocessing, then arguably you could, you know, replace some of 199 00:12:54,400 --> 00:12:58,720 it. But the technology that I think 200 00:12:58,760 --> 00:13:05,480 it adds a real new dimension to this is the agentic side. 201 00:13:06,000 --> 00:13:09,240 Because what are what are agents? 202 00:13:09,240 --> 00:13:14,240 Well, first of all, let's tie it back to the usefulness of Chachi 203 00:13:14,240 --> 00:13:17,720 PT or Gemini. We have all found them to be 204 00:13:17,720 --> 00:13:22,760 incredibly useful in our day-to-day tasks, but I have 205 00:13:22,760 --> 00:13:27,080 personally noticed a a huge shift in usefulness with the 206 00:13:27,080 --> 00:13:28,960 agentic side. And what do we mean by the 207 00:13:28,960 --> 00:13:31,520 agentic side? At least the way I define it. 208 00:13:31,840 --> 00:13:36,600 Well, I think of it is just simply now the LLM is able to 209 00:13:36,600 --> 00:13:38,520 call some tools. Now that's probably 210 00:13:38,520 --> 00:13:42,200 oversimplifying it, but from a practical point of view, instead 211 00:13:42,200 --> 00:13:47,600 of me just being able to ask some ask Gemini to do something 212 00:13:47,600 --> 00:13:49,920 for me or, or codecs or whatever. 213 00:13:50,280 --> 00:13:53,080 Although the use of codecs now is sort of murky in the waters a 214 00:13:53,080 --> 00:13:54,680 little bit. If I stick to like a chat 215 00:13:54,680 --> 00:13:58,040 prompt, it's only going to be able to really do things that 216 00:13:58,040 --> 00:14:03,720 rely just on text or to A to a limit where what I really want 217 00:14:03,720 --> 00:14:06,600 it to do is to be able to go and do things for me. 218 00:14:07,080 --> 00:14:11,880 So for example, if I ask, I want you to create a summary of all 219 00:14:11,880 --> 00:14:13,200 my emails. Great. 220 00:14:13,480 --> 00:14:17,360 But what if I want you to now look through all my emails, look 221 00:14:17,360 --> 00:14:22,680 through all my Slack messages, and then I want you to go in and 222 00:14:24,040 --> 00:14:27,000 also look at a journal paper that I'm writing. 223 00:14:27,560 --> 00:14:31,680 And I want you to look at the emails where I got some response 224 00:14:31,680 --> 00:14:34,200 from somebody. And I want you to take that, go 225 00:14:34,200 --> 00:14:37,320 into, go into that. Oh, and I also want you to go 226 00:14:37,320 --> 00:14:39,840 and create some graphs for me. So can you also create some 227 00:14:39,840 --> 00:14:42,680 Python graphs using, you know, map pot, Lib, etcetera. 228 00:14:43,240 --> 00:14:46,560 And I want you to do that all for me autonomously. 229 00:14:47,160 --> 00:14:51,480 What I'm describing to you now is absolutely achievable right 230 00:14:51,480 --> 00:14:53,880 this second using cloud code or codecs. 231 00:14:55,200 --> 00:15:00,920 It is also possible for the tools that that that that NLM is 232 00:15:00,920 --> 00:15:04,280 essentially calling or the agent is calling can be simulation 233 00:15:04,280 --> 00:15:07,880 tools. So if I extend that further and 234 00:15:07,880 --> 00:15:11,280 I say I want you to be helping me to do some journal paper, 235 00:15:12,040 --> 00:15:16,120 maybe I could say now, well, could you actually, could you 236 00:15:16,120 --> 00:15:19,160 run some simulations for me? I have an idea. 237 00:15:19,320 --> 00:15:24,440 You know, here's some e-mail correspondence, here is a thread 238 00:15:24,480 --> 00:15:26,760 on a Slack channel with some collaborators. 239 00:15:27,880 --> 00:15:31,840 Not just write the paper, but will you go and set up an open 240 00:15:31,840 --> 00:15:35,680 phone simulation that that tries out some of those ideas we 241 00:15:35,680 --> 00:15:37,240 discussed and, and can you kick it off? 242 00:15:37,640 --> 00:15:39,760 And then when it's done, can you create the graphs and can you do 243 00:15:39,760 --> 00:15:41,960 it for me? That's completely achievable 244 00:15:41,960 --> 00:15:43,640 today. That is not future thinking. 245 00:15:43,640 --> 00:15:49,760 That is achievable today. My point is in what part of that 246 00:15:49,760 --> 00:15:57,120 then is this CFD tool? The main thing it's not is it 247 00:15:57,640 --> 00:16:02,240 because actually codecs or clawed or vibe code now with 248 00:16:02,240 --> 00:16:05,600 mistrial entering as a new sort of player in this field from 249 00:16:05,600 --> 00:16:09,000 with a heavy sort of sovereign AI angle, European angle. 250 00:16:09,440 --> 00:16:13,680 So codecs clawed or or vibe, they can actually go and do that 251 00:16:13,680 --> 00:16:16,000 for you. And you don't have to be an 252 00:16:16,040 --> 00:16:21,840 expert user of Open phone because most likely those models 253 00:16:21,840 --> 00:16:25,000 have either been trained on sufficient smart data that 254 00:16:25,000 --> 00:16:28,000 included Open Phone documentation and various 255 00:16:28,000 --> 00:16:32,440 reports across the web. Or more likely that in a real 256 00:16:32,440 --> 00:16:35,200 life scenario, and maybe this is where the commercial bit comes 257 00:16:35,200 --> 00:16:39,480 in, one of those providers could have specifically added that 258 00:16:39,480 --> 00:16:41,680 capability. They could have looked at 259 00:16:41,680 --> 00:16:46,560 science and engineering as a key area they want to get into, and 260 00:16:46,560 --> 00:16:50,040 so they will actually try to make their models good at doing 261 00:16:50,040 --> 00:16:53,840 this. At that point, the stickiness or 262 00:16:53,840 --> 00:16:59,240 the the difficulty of using ACFD code is no longer there, which 263 00:16:59,240 --> 00:17:04,359 is arguably one of the reasons those codes exist or have such a 264 00:17:04,359 --> 00:17:06,880 strong Moat because they're not easy to use. 265 00:17:07,160 --> 00:17:10,480 They require expertise, they're very well validated. 266 00:17:11,400 --> 00:17:15,280 If agent starts to come in and very powerful LLMS come in that 267 00:17:15,280 --> 00:17:19,839 can use these tools very well, then actually the uplift of 268 00:17:19,839 --> 00:17:23,040 changing to a different tool is not as much. 269 00:17:23,880 --> 00:17:28,280 And the ability to orchestrate multiple tools is also the same. 270 00:17:29,080 --> 00:17:34,280 So it raises a very interesting question, which is are the 271 00:17:34,640 --> 00:17:38,520 software companies just going to reside to being just tools and 272 00:17:38,520 --> 00:17:42,440 skills and actually your relationship will be more with 273 00:17:43,120 --> 00:17:47,640 the open AIS, the mistrials, the anthropics, etcetera. 274 00:17:48,400 --> 00:17:53,360 Or, and it is the case today, will new companies arise that 275 00:17:53,360 --> 00:17:55,840 are trying to do this? Interestingly, and there's very 276 00:17:55,840 --> 00:17:58,040 little public information on this, but you have companies 277 00:17:58,040 --> 00:18:02,240 like Prometheus, I think Jeff Bezos put out some announcement 278 00:18:02,240 --> 00:18:06,400 about wanting to create a artificial engineer out of this. 279 00:18:06,400 --> 00:18:09,280 So read into that what you will. But it sounds pretty clear that 280 00:18:09,280 --> 00:18:11,880 they want to build some sort of capability like that. 281 00:18:13,440 --> 00:18:17,800 You see the recent acquisition of Meai by Mistral and you see a 282 00:18:17,800 --> 00:18:21,600 huge interest from open AI and Anthropic in general, AI for 283 00:18:21,600 --> 00:18:25,520 science, which makes you look at the future and start to think 284 00:18:26,560 --> 00:18:29,320 there could be a new wave now of companies coming around. 285 00:18:30,000 --> 00:18:34,560 And there are companies that are now base themselves purely to 286 00:18:34,560 --> 00:18:40,200 create agents to create artificial AI engineers. 287 00:18:41,680 --> 00:18:44,760 This in some ways has the opportunity to massively 288 00:18:44,760 --> 00:18:49,480 democratize and and reduce the need for you to be some 289 00:18:49,480 --> 00:18:52,440 specialist. An example I gave is I remember 290 00:18:52,440 --> 00:18:56,560 when I was an engineer that being a CAD expert was a 291 00:18:56,560 --> 00:18:58,800 challenge. You would have to learn how to 292 00:18:58,800 --> 00:19:02,760 use NX or CTIA or SolidWorks, and frankly, if you didn't know 293 00:19:02,760 --> 00:19:04,880 how to use those, you couldn't use it. 294 00:19:04,880 --> 00:19:06,960 You would have to go and find somebody, have to pay someone to 295 00:19:06,960 --> 00:19:09,280 do it. And I personally found it a 296 00:19:09,280 --> 00:19:12,480 blocker actually, because I knew what I wanted to do, I just 297 00:19:12,480 --> 00:19:17,560 couldn't do it. Now Fast forward and by the way, 298 00:19:17,560 --> 00:19:21,680 similar extent to some software tools where you know CFD, right? 299 00:19:21,680 --> 00:19:23,720 I understand the theory, I understand what I want to do. 300 00:19:23,720 --> 00:19:25,960 I just don't know how to use the code or I don't have to script 301 00:19:25,960 --> 00:19:29,040 it or whatever. As long as the Asian knows how 302 00:19:29,040 --> 00:19:32,600 to do it. I can say, you know, create this 303 00:19:32,680 --> 00:19:37,200 in this scan package, set up the simulation, this tool. 304 00:19:38,280 --> 00:19:41,640 It's kind of interesting then that it relies on a good skills 305 00:19:42,280 --> 00:19:45,560 capability. It relies on those tools being 306 00:19:45,560 --> 00:19:48,440 open to agents. And that is a very interesting 307 00:19:48,440 --> 00:19:53,440 point, which is will companies embrace this, you know, agent 308 00:19:53,440 --> 00:19:57,240 first approach or will they try to put more guard rails and and 309 00:19:57,240 --> 00:20:00,880 even block people to use those tools via agents? 310 00:20:00,880 --> 00:20:03,720 Will there be some terms and conditions and licenses? 311 00:20:04,720 --> 00:20:06,840 And and that brings up another point around licenses. 312 00:20:06,840 --> 00:20:10,360 A lot of licenses are based around a human having a seat on 313 00:20:10,360 --> 00:20:15,640 a on a machine, and just like cloud with the ability to scale 314 00:20:15,640 --> 00:20:18,440 disrupted the per core license model. 315 00:20:18,920 --> 00:20:22,640 And really wanted it more to be like a just a power session like 316 00:20:22,640 --> 00:20:24,960 like Star has with his POD license. 317 00:20:25,680 --> 00:20:29,200 Finally, and I know this is rambling on a little bit, but 318 00:20:29,200 --> 00:20:32,760 I'm just dumping all my thoughts down here, what about open code? 319 00:20:32,760 --> 00:20:36,080 Open source codes? Much of the challenge was around 320 00:20:36,080 --> 00:20:38,400 support and capability and validation. 321 00:20:40,040 --> 00:20:43,320 Is it not the case now with coding agents and the power of 322 00:20:43,320 --> 00:20:48,840 LLMS that these codes actually could become very capable and 323 00:20:48,840 --> 00:20:53,800 and could be used more easily in a workflow instead of a 324 00:20:53,800 --> 00:20:57,360 commercial tool because the agent can write and script and 325 00:20:57,360 --> 00:21:01,160 validate and go off and do test suites as much as possible? 326 00:21:02,200 --> 00:21:08,760 So I'm not saying at all that you will see a complete wipe out 327 00:21:08,760 --> 00:21:10,360 of all the current CAE companies. 328 00:21:10,360 --> 00:21:14,720 In fact, it definitely won't be like that, but I just put this 329 00:21:14,720 --> 00:21:19,880 out that there is a very interesting opening it because 330 00:21:19,880 --> 00:21:22,840 these two new technologies around sort of foundation 331 00:21:22,840 --> 00:21:24,680 models. This idea that you could have 332 00:21:24,680 --> 00:21:28,080 simulations that would run in seconds or minutes instead of 333 00:21:28,080 --> 00:21:31,520 many hours. The idea that if you have such 334 00:21:31,520 --> 00:21:35,200 surrogate models and foundation models tied with agents, you 335 00:21:35,200 --> 00:21:40,240 could finally realize the sort of reverse engineering or design 336 00:21:40,240 --> 00:21:43,360 optimization, the automatic generation of cases. 337 00:21:44,200 --> 00:21:48,960 If you have a pipeline that is able to do that, then you could 338 00:21:48,960 --> 00:21:52,400 simply have agents go and creating optimum geometries for 339 00:21:52,400 --> 00:21:56,720 you, running those simulations autonomously with you being 340 00:21:56,720 --> 00:21:58,400 essentially the the orchestrator. 341 00:21:58,400 --> 00:22:01,480 The potential for productivity is absolutely massive. 342 00:22:02,520 --> 00:22:05,480 It is. It could finally realize some of 343 00:22:05,480 --> 00:22:09,280 the dreams of people in the CAE and the EDA sector around 344 00:22:09,280 --> 00:22:14,720 productivity boosts, around finding designs around people 345 00:22:14,720 --> 00:22:17,880 with just domain knowledge driving the code rather than 346 00:22:17,880 --> 00:22:20,320 code knowledge. And that is important from an 347 00:22:20,320 --> 00:22:23,320 education point of view. I would say what matters more 348 00:22:23,400 --> 00:22:26,160 now is not are you really good at writing Python? 349 00:22:26,520 --> 00:22:28,280 No. Are you really good at using 350 00:22:28,280 --> 00:22:29,200 Openfox? No. 351 00:22:29,520 --> 00:22:33,360 Do you understand physics? Do you understand numerical? 352 00:22:33,360 --> 00:22:36,280 But do you understand how to pose the prompt? 353 00:22:36,280 --> 00:22:40,160 Do that is more important? Do you almost then it starts to 354 00:22:40,160 --> 00:22:41,560 become that you need to go higher up. 355 00:22:42,000 --> 00:22:45,760 If it's so easy to run ACFD code, why don't why don't I 356 00:22:45,760 --> 00:22:47,360 start to run an FEA code as well? 357 00:22:47,680 --> 00:22:51,720 Many companies are you are split if you are ACFD person or an FEA 358 00:22:51,720 --> 00:22:53,400 person. You're split if you're an 359 00:22:53,400 --> 00:22:56,360 acoustics person. Well, now you could become more 360 00:22:56,360 --> 00:23:00,320 than engineer who simply is just driving the tools. 361 00:23:00,520 --> 00:23:03,080 That position in a company is often the sort of chief 362 00:23:03,080 --> 00:23:06,040 engineer. Well, maybe now it's more 363 00:23:06,040 --> 00:23:09,120 important that you understand trade-offs and and larger 364 00:23:09,120 --> 00:23:13,640 engineering concepts and doing this sort of specialist deep 365 00:23:13,640 --> 00:23:17,920 down analysis can be more automated and clearly 366 00:23:17,920 --> 00:23:20,040 simplifying what it. There are many industries where 367 00:23:20,040 --> 00:23:22,360 this will happen for many years, but I'm I'm sort of trying to 368 00:23:22,360 --> 00:23:27,360 Fast forward and play a what if scenario to, to to all this 369 00:23:27,360 --> 00:23:29,160 point. It raises many interesting 370 00:23:29,160 --> 00:23:33,240 questions. Therefore, on the if the LLMS 371 00:23:33,240 --> 00:23:35,680 and the agents are key. It raises questions about 372 00:23:35,680 --> 00:23:39,880 traceability around sovereignty. Who owns these models? 373 00:23:39,880 --> 00:23:41,920 Where these models done? Should you be developing your 374 00:23:41,920 --> 00:23:44,480 own models? Can you take them off the off 375 00:23:44,480 --> 00:23:48,000 the shelf? So many, so many questions. 376 00:23:48,680 --> 00:23:54,320 It, it rises, but I, I find it a fascinating area that I think 377 00:23:54,320 --> 00:23:59,640 will be a key technology driver, an innovation area for, for, for 378 00:23:59,640 --> 00:24:04,000 years to come and something that I'd like to explore over the 379 00:24:04,000 --> 00:24:08,040 course of this, of this season. I want to look at speak to some 380 00:24:08,040 --> 00:24:12,080 people who can give insights into this on the the the core 381 00:24:12,080 --> 00:24:14,640 ideas around sort of foundation models. 382 00:24:14,640 --> 00:24:18,560 Because the reason that foundation models of surrogates 383 00:24:18,560 --> 00:24:22,960 are so important is that if a tool still takes a day to run 384 00:24:23,840 --> 00:24:28,800 the the the agentic sort of optimization workflows will 385 00:24:28,800 --> 00:24:34,200 ultimately be massively Hanford by the actual tool itself. 386 00:24:34,800 --> 00:24:39,720 If the tool can match the speed of the agent, the LM, then that 387 00:24:39,720 --> 00:24:41,680 is what will really have the breakthrough. 388 00:24:41,680 --> 00:24:45,720 So whilst kind of the advantage of the agent is that, and so as 389 00:24:45,720 --> 00:24:47,800 you don't need to use any of these surrogate model 390 00:24:47,800 --> 00:24:52,080 technologies, they're even more optimum when you when you do. 391 00:24:53,200 --> 00:24:56,960 And, and so this is where the the, the combination comes from. 392 00:24:56,960 --> 00:25:01,480 So, yeah, I don't know what the future will, will, will, will 393 00:25:01,480 --> 00:25:04,720 be, but I know that it is certainly an exciting one. 394 00:25:04,720 --> 00:25:07,680 I think this is a great time to be in the start up space. 395 00:25:07,960 --> 00:25:11,800 It's a great time to be researching, to be in academia, 396 00:25:11,800 --> 00:25:14,680 thinking of your PhD, thinking about what topics you want to 397 00:25:14,680 --> 00:25:17,320 do. It's a challenging time to be an 398 00:25:17,320 --> 00:25:19,880 end engineering customer. You don't know what to pick. 399 00:25:20,160 --> 00:25:23,240 It's a challenging time to be an established ISP, but it's also a 400 00:25:23,240 --> 00:25:28,080 great opportunity because you can potentially help drive your 401 00:25:28,080 --> 00:25:32,480 company to bring out the next, you know, industry 4 point O or 402 00:25:32,480 --> 00:25:35,160 five point O or whatever the the number is today. 403 00:25:35,880 --> 00:25:41,200 So I will, I'll leave it there. I'm sure I didn't cover many 404 00:25:41,200 --> 00:25:45,200 things and I hope you appreciate that this isn't me in any way 405 00:25:45,200 --> 00:25:50,440 diminishing what will be the classic example of any new 406 00:25:50,440 --> 00:25:54,440 technology taking years and years and years to them. 407 00:25:54,880 --> 00:25:57,960 So I'm not saying next year or even 5 years everyone will be 408 00:25:57,960 --> 00:26:01,360 doing this, just like some people are still running on 409 00:26:01,360 --> 00:26:04,800 Windows 3.1 or Windows 95 in some places around the world. 410 00:26:05,160 --> 00:26:10,600 But I am confident, I am definitely confident that if we 411 00:26:10,600 --> 00:26:16,160 look in three years time that the role of agents and, and and 412 00:26:16,160 --> 00:26:18,920 surrogate models on AI will be transformative. 413 00:26:19,040 --> 00:26:22,000 And so please, if you're listening, if you are one of 414 00:26:22,000 --> 00:26:24,880 these people who are still sceptic of AII would 415 00:26:25,360 --> 00:26:28,040 respectfully encourage you to to try it. 416 00:26:28,160 --> 00:26:31,840 And I would hope that you will believe me. 417 00:26:32,120 --> 00:26:35,360 The final call out, and I will say though, is in order to 418 00:26:35,360 --> 00:26:40,320 convince the community, we need transparency, we need openness 419 00:26:40,400 --> 00:26:42,680 as much as possible within a commercial mindset. 420 00:26:43,080 --> 00:26:45,680 People who know me know that I have always tried throughout my 421 00:26:45,680 --> 00:26:49,360 career to, to do things in a transparent and open way. 422 00:26:49,360 --> 00:26:51,480 It's part of the reason for doing this podcast is to sort of 423 00:26:51,480 --> 00:26:54,160 share the knowledge. The reason that I've created 424 00:26:54,160 --> 00:26:59,000 these sort of open source data sets is that desire to do it. 425 00:26:59,000 --> 00:27:02,000 The reason for doing these workshops or do CFD and others 426 00:27:02,000 --> 00:27:05,240 is to try and share it. I am fully appreciative of the 427 00:27:05,240 --> 00:27:07,680 need for commercialization. It's how the world works. 428 00:27:08,280 --> 00:27:11,520 But I think the one key thing if this agents and AI and 429 00:27:11,520 --> 00:27:13,480 everything is going to be a transformative. 430 00:27:13,920 --> 00:27:17,000 People will only get on board if they can believe it, if they can 431 00:27:17,000 --> 00:27:19,080 see it, if they can see transparency, they can see 432 00:27:19,080 --> 00:27:22,040 publications. I feel like academia and 433 00:27:22,200 --> 00:27:26,080 scientific rigor has a massive part to play in this 434 00:27:26,080 --> 00:27:28,480 transformation. And if it can be done with 435 00:27:28,480 --> 00:27:33,080 scientific rigor at transparency and good science, then I think 436 00:27:33,080 --> 00:27:36,800 the the world will move on. If it's done behind closed doors 437 00:27:36,800 --> 00:27:40,360 with no publications, with just sort of a, you know, black box 438 00:27:40,360 --> 00:27:43,600 solution that that will not convince. 439 00:27:43,600 --> 00:27:46,840 And in fact, that will in some ways be a barrier to this, what 440 00:27:46,840 --> 00:27:49,800 I feel is transformative technology helping. 441 00:27:50,360 --> 00:27:52,480 So with that, I will leave it there. 442 00:27:52,480 --> 00:27:55,160 I look forward to hearing your comments and I hope you'll join 443 00:27:55,160 --> 00:27:58,200 me on this new season. Please listen to the old 444 00:27:58,200 --> 00:28:00,240 episodes. There's quite a few, quite so 445 00:28:00,240 --> 00:28:03,200 many hours, but I hope you'll join for this season. 446 00:28:03,200 --> 00:28:04,720 So with that, thanks for listening.