WEBVTT 00:00.000 --> 00:10.000 All right, everyone. Thank you so much for being here. 00:10.000 --> 00:12.000 So, well, I'm Frédéric Devier. 00:12.000 --> 00:15.000 I'm an age embedded in IoT programs at the Eclipse Foundation. 00:15.000 --> 00:20.000 I was supposed to have a friend here, my flying flow is still around, 00:20.000 --> 00:24.000 but he's in Taiwan right now, celebrating Chinese New Year with his wife. 00:24.000 --> 00:29.000 So, that's just me. This is a talk about chips or crisps for our UK friends, 00:29.000 --> 00:34.000 whatever. So, please be kind with me because, well, 00:34.000 --> 00:38.000 how would I say that? I'm not the chip engineer of designer, 00:38.000 --> 00:44.000 but I will try to give you this taste of how Europe can gain back its competitive advantage 00:44.000 --> 00:49.000 using open source technologies around microelectronics and AI. 00:49.000 --> 00:53.000 All right, so, of course, we talk about AI. 00:53.000 --> 00:57.000 We talk about microelectronics. If I were to ask you this question, 00:57.000 --> 01:02.000 where are our AI success stories in Europe? 01:02.000 --> 01:06.000 And you would be silent, right? They are none. 01:06.000 --> 01:09.000 Or, if they are, we don't know them. 01:09.000 --> 01:11.000 And that's not your problem, okay? 01:11.000 --> 01:15.000 So, the thing is, we have bright engineers in Europe. 01:15.000 --> 01:20.000 We have world-class semiconductor factories all over the place, right? 01:20.000 --> 01:24.000 And significant financial support from the commission. 01:24.000 --> 01:31.000 So, in spite of all of that, we are late as Europeans to the AI revolution. 01:31.000 --> 01:36.000 But, not all is lost. So, the point of this presentation is to tell you, 01:36.000 --> 01:40.000 how we can catch up partly with open source hardware. 01:40.000 --> 01:45.000 And, of course, the type of work that we do at the open hardware foundation. 01:45.000 --> 01:50.000 By the way, maybe you were familiar with the previous name, open hardware group. 01:50.000 --> 01:53.000 That was an independent organization. 01:53.000 --> 01:59.000 And, since January of this year, open hardware group is now open hardware foundation, 01:59.000 --> 02:01.000 and a part of the eclipse foundation. 02:01.000 --> 02:05.000 So, that's why I'm here with this wonderful UD. 02:05.000 --> 02:08.000 Okay, so, three points in this short presentation. 02:08.000 --> 02:11.000 First, about microelectronics and sovereignty. 02:11.000 --> 02:17.000 A few AI use cases to show you how you could build a startup using open hardware technology 02:17.000 --> 02:22.000 around the AI, and finally a bit about the open hardware foundation in what we do. 02:22.000 --> 02:31.000 So, this gap in the AI, of course, we will fill it with good microelectronics. 02:31.000 --> 02:35.000 And, part of the solution is an open source solution. 02:35.000 --> 02:39.000 But, in order to get what you understand why this is the case, 02:39.000 --> 02:42.000 we need to understand the current gap that we've got. 02:42.000 --> 02:46.000 Oh, and my animations are in the wrong order. 02:46.000 --> 02:49.000 So, anyway, so first line. 02:49.000 --> 02:51.000 So, we start on the left. 02:51.000 --> 02:55.000 Essentially, when you look at the history of microelectronics, 02:55.000 --> 02:58.000 you know, this has been dominated first by the U.S. 02:58.000 --> 03:02.000 And, after that, you had Japan, Korea that inserted themselves in the supply chain. 03:02.000 --> 03:06.000 And, it stayed like that until today, because if you think about the big fondries, 03:06.000 --> 03:11.000 the SMC, all of that, all of that is either in Asia or in the U.S. 03:11.000 --> 03:16.000 And, well, of course, they are semiconductor faves in Europe, but that's not well. 03:16.000 --> 03:19.000 Anyway, general history. 03:19.000 --> 03:26.000 In recent years, of course, the growth of IoT embedded AI use cases drove a lot of consumption 03:26.000 --> 03:29.000 additional production in microelectronics in general. 03:29.000 --> 03:33.000 We have all chips in our pockets, in our phones, several of them, right? 03:33.000 --> 03:40.000 And, the EU, over time, you realize that, well, this is an important industry that we should develop. 03:40.000 --> 03:47.000 And, at the same time, well, didn't necessarily make the right choices in order to make that happen. 03:47.000 --> 03:53.000 Now, the good news is that when you look at compute capacities, 03:53.000 --> 04:00.000 okay, general compute, not specialized things like AI or whatever, this is a commodity now. 04:00.000 --> 04:05.000 It's very, very hard to innovate, which means if you pick up a laptop 10 year old, 04:05.000 --> 04:10.000 some most basic tasks, as well as the US machine, right? 04:10.000 --> 04:14.000 Now, they try to sell you new laptops, saying, AI, PC is an account of stuff, right? 04:14.000 --> 04:19.000 So, the focus is as well, which means you can feel the gap in general compute easily 04:19.000 --> 04:21.000 because this is a commodity. 04:21.000 --> 04:26.000 And, at the same time, tremendous value in custom chips for specialized applications. 04:26.000 --> 04:32.000 And, risk life, as an industry standard instruction set provides you a good building block 04:32.000 --> 04:39.000 to provide high-quality chips to your customers and consumers. 04:39.000 --> 04:43.000 But, the key there is, of course, risk life is just the instruction set. 04:43.000 --> 04:45.000 You don't have an open design around it. 04:45.000 --> 04:48.000 And, at open-out web foundation, we're trying to fix that. 04:48.000 --> 04:49.000 So, what's the plan? 04:49.000 --> 05:01.000 How you can make up is, you know, well, overtake even, you know, the US and its Asian partners. 05:01.000 --> 05:04.000 And, the solution is a bit in developing countries. 05:04.000 --> 05:05.000 Why is that? 05:05.000 --> 05:11.000 Well, developing countries are very advanced in their usage of wireless technologies, right? 05:11.000 --> 05:16.000 And, in order to get that connectivity into place, they didn't need to build the extensive 05:17.000 --> 05:20.000 copper networks that we have in more developed countries. 05:20.000 --> 05:22.000 So, the strategy here is the same. 05:22.000 --> 05:28.000 Essentially, we live frog, okay, our current gap by using cutting edge technology. 05:28.000 --> 05:33.000 But, we skip some of the steps in between. 05:33.000 --> 05:35.000 And, this is possible because of risk life. 05:35.000 --> 05:41.000 In other words, Europe doesn't need to produce a European specific micro architecture 05:41.000 --> 05:42.000 and that kind of stuff. 05:42.000 --> 05:46.000 Can you imagine, friendship designers and German, 05:46.000 --> 05:49.000 designers trying to agree on the number of registers in the chip. 05:49.000 --> 05:51.000 You will never get done. 05:51.000 --> 05:54.000 So, we live frog, that particular problem. 05:54.000 --> 05:58.000 And, because of risk five, we can go straight to the good stuff. 05:58.000 --> 06:01.000 The AI stuff, the more advanced two-scasers. 06:01.000 --> 06:05.000 Because risk five provides a foundational, busy block, okay? 06:05.000 --> 06:10.000 And, I will show you why you can do that very briefly in some use cases. 06:10.000 --> 06:15.000 So, first, before we get that, you will ask me how this is possible. 06:15.000 --> 06:18.000 Well, this is possible because of the power of open source. 06:18.000 --> 06:26.000 Open source provides you governance and a collaboration layer that enable competitors to work in common with things. 06:26.000 --> 06:29.000 And, you would tell me, who are this is not actually happening? 06:29.000 --> 06:32.000 Yes, this is actually happening, for example, in automotive. 06:32.000 --> 06:38.000 So, in automotive, you know, they started by sharing suppliers to reduce their costs. 06:38.000 --> 06:41.000 And, that's why you have big companies like Bosch, 06:41.000 --> 06:44.000 selling car parts to many, many builders. 06:44.000 --> 06:47.000 And, after that, they started sharing platforms. 06:47.000 --> 06:52.000 Oh, my Nissan is the same as this Renault when I get a stuff and whatever. 06:52.000 --> 06:59.000 And, now, they are sharing ecosystems because there are many intertwined suppliers and all of that. 06:59.000 --> 07:03.000 And, the next logical step is to have common open source platforms. 07:03.000 --> 07:08.000 And, that's what they are doing in our software-defined vehicle working group at the Eclipse Foundation. 07:08.000 --> 07:15.000 And, we think we can make the same kind of dynamic happen around open source hardware. 07:15.000 --> 07:19.000 So, the suggests the spatitially to leverage open source open standards. 07:19.000 --> 07:23.000 And, of course, European investments and risk life. 07:23.000 --> 07:29.000 And, by the way, risk life is a real thing with massive, you know, market footprint already. 07:29.000 --> 07:30.000 All right. 07:30.000 --> 07:32.000 So, case studies. 07:33.000 --> 07:35.000 Let's look at AI inference, for example. 07:35.000 --> 07:39.000 The kind of stuff that Facebook is doing in order to show you more ads. 07:39.000 --> 07:40.000 Oh, wonderful. 07:40.000 --> 07:43.000 The thing is, of course, this is about inference, 07:43.000 --> 07:46.000 because training happens once in this very expensive, 07:46.000 --> 07:53.000 but once you have your model, then you can milk the data of people, really, and make a profit. 07:53.000 --> 07:57.000 So, of course, this requires massive electricity, massive enterprise, 07:57.000 --> 08:02.000 it centers, massive, you know, the support teams around the, you know, those servers. 08:02.000 --> 08:06.000 But, when you look at the chip themselves, they are simple designs. 08:06.000 --> 08:10.000 Okay, you take a processing element, cotton paste, that's on myos, and buying. 08:10.000 --> 08:15.000 You have a chip that will be able to deliver the kind of stuff that Facebook is doing. 08:15.000 --> 08:18.000 Don't forget that machine learning is fancy statistics. 08:18.000 --> 08:19.000 Okay. 08:19.000 --> 08:23.000 So, the chips themselves are not that complicated. 08:24.000 --> 08:29.000 So, this is exists, and well, it's power by your spive, 08:29.000 --> 08:32.000 and serves search results to Europeans today. 08:32.000 --> 08:35.000 So, you can replicate the same. 08:35.000 --> 08:40.000 Of course, AI at the edge is where the real interesting thing is. 08:40.000 --> 08:41.000 Okay? 08:41.000 --> 08:45.000 Because then you can lower latency and have multi-authored advantages. 08:45.000 --> 08:47.000 So, there's an example here on the slide, 08:47.000 --> 08:52.000 and essentially what they did is to produce a low power type of deployment, 08:52.000 --> 08:54.000 and a very small form factor. 08:54.000 --> 08:57.000 Once again, leveraging risk five, and this is made in Europe. 08:57.000 --> 08:58.000 Okay? 08:58.000 --> 09:01.000 So, they did it, you can do it. 09:01.000 --> 09:06.000 Bosh also is working on something really interesting, 09:06.000 --> 09:12.000 and specifically, Bosh Automated AI project is leveraging open hardware foundation cores, 09:12.000 --> 09:16.000 and things like that, and really, this is supported by chips. 09:16.000 --> 09:19.000 JU with the Tristan project. 09:19.000 --> 09:24.000 And, well, some people are doing also telco infrastructure using risk five. 09:24.000 --> 09:27.000 It's in the UK, so technically in Europe, I guess. 09:27.000 --> 09:28.000 Anyway. 09:28.000 --> 09:33.000 And, of course, your pockets are full of risk five. 09:33.000 --> 09:34.000 Okay? 09:34.000 --> 09:37.000 Even if you have a fruity smartphone, right? 09:37.000 --> 09:41.000 There are tons of risk five-based microcontrollers in there 09:41.000 --> 09:45.000 that will manage the power, manage the eat, manage battery charging, 09:45.000 --> 09:47.000 and lots of stuff. 09:47.000 --> 09:49.000 And you can do the same. 09:49.000 --> 09:50.000 Okay? 09:50.000 --> 09:52.000 So, open hardware foundation briefly. 09:52.000 --> 09:53.000 What we are? 09:53.000 --> 09:54.000 Okay. 09:54.000 --> 09:55.000 One last minute. 09:55.000 --> 09:58.000 Well, we are the only non-profit organization in the world, 09:58.000 --> 10:02.000 working on verified, industrial-grade risk five cores. 10:02.000 --> 10:05.000 And they are open source, and they are on GitHub, so you can take them, 10:05.000 --> 10:10.000 but if I then put that on an FPGA, try the FPGA, whatever. 10:11.000 --> 10:13.000 We have a roadmap and relief. 10:13.000 --> 10:16.000 We have good coverage in terms of what we cover with those cores, 10:16.000 --> 10:20.000 so all the way from very, very, very small low power microcontrollers 10:20.000 --> 10:24.000 to data center type, eventually type of processors. 10:24.000 --> 10:25.000 Okay? 10:25.000 --> 10:30.000 So, that's in the far future, but already, what's in so it's green there, 10:30.000 --> 10:31.000 is quite robust. 10:31.000 --> 10:35.000 And, of course, we have significant EU-founded research projects 10:35.000 --> 10:39.000 that we are involved in, where essentially we bring together 10:39.000 --> 10:43.000 European, industrial, you know, semiconductor companies, 10:43.000 --> 10:47.000 and of course the open source community to work together. 10:47.000 --> 10:50.000 And so, in this case, they are leveraging several of our cores, 10:50.000 --> 10:54.000 and we have a series of open source support components, 10:54.000 --> 10:58.000 like the serial interface, GPIO, etc., that you can leverage, 10:58.000 --> 11:01.000 in order to do your own custom designs. 11:01.000 --> 11:02.000 Okay? 11:02.000 --> 11:05.000 And the goal there, why this is at the Eclipse Foundation, 11:05.000 --> 11:08.000 is really to have this comprehensive risk five stack, 11:08.000 --> 11:13.000 all the way to the chip, to the cloud platforms 11:13.000 --> 11:15.000 that we support your use cases in IoT, 11:15.000 --> 11:18.000 as DV, whatever you can dream of and everything in between, 11:18.000 --> 11:20.000 including developer tools. 11:20.000 --> 11:25.000 So, that's the plan, and that's where we think we can help. 11:25.000 --> 11:27.000 And essentially, the call to action there, 11:27.000 --> 11:31.000 please leverage this five, please fill this gap, 11:31.000 --> 11:36.000 and overtake the US as one European community 11:37.000 --> 11:40.000 by leveraging what we do at open hardware foundation. 11:40.000 --> 11:41.000 Thank you. 11:41.000 --> 11:45.000 Thank you very much.