Pat Gelsinger and Linus Torvalds talk Linux, open source, technology and more

Well, it is my pleasure to have somebody that I have admired for decades, Linus Torvalds, joining me on at the Innovation Conference for a conversation about history of Linux, technology, open source. So, you know, if you don't mind, Linus, maybe just start out telling us the early days, right? You know, when Linux was getting underway and what was that like? And, you know, take us back, right? You're now 30 years old, right? You know, a little bit over 30 years old for Linux. So this has been an incredible journey. So take us back to the beginning. Right. So 32 years ago, I got a new computer and I was used to home computers and I came from, for the geeks among you, I was used to 6502 CPUs originally and then upgraded to the 68k series. And the C80 camp and the 8086 camp was kind of strange to me. The ugly instructions. You didn't like that segmentation stuff? I really did not like that at all, right? And the odd addressing too, with no rules at all. But at the same time, the writing was on the wall and the 386 had come out, right? Yes, my baby. And the 386 really changed, took away, well, I mean, you could ignore, let's put it this way. You could ignore all the bad parts. And I really wanted to learn how it worked. And that's where your book comes in, because that was one of the basically two tomes I used to learn my new machine and start writing low level code to really understand. It wasn't just the CPU, it was all the PC peripheral components around it too, which I was completely unfamiliar with. And so I wrote a VGA character driver. I wrote the serial driver. I learned how to get into protected mode because that was kind of what I was interested in. And when you do that and you work on it for a few months, suddenly what you have is kind of the beginning of an operating system. And I realized that if I took this the next step, I could actually use this not just as a like teaching myself how the system worked, but also use it as a daily driver. And that's when I actually started actively developing as an operating system. So when it went from being, hey, this is just me to learning this new machine to it becoming an operating system, do you have any idea what you were unleashing? No, and I didn't really, I had a very low expectations because I was always expecting that, okay, this is one of these projects that I will do for myself. And then at some point, I will get bored and then I will get a real operating system eventually, which I couldn't afford at the time because I came from not blue collar, but like not a pretty humble beginning. Exactly, right. And so what happened was that I was still pretty proud of what I've started. So I put it out and it wasn't really initially about open source for me. Initially, it was more like, hey, look at what I've done, right? Kind of, this is kind of cool, this is interesting, maybe somebody else is interesting. And that completely changed everything. And that's, I went from being kind of knowing about open source to being completely convinced this is the only way I ever want to do it. That was your light bulb moment. That was my light bulb moment. Because it wasn't initially so much that it brought in all these other people, because that took a while. It really takes a while for the process to get to the point where it's self sustaining. But it meant that this project that was my small thing that I expected to go away, suddenly the open source part made it so much more motivational for myself, because I got questions, I got people involved that said, this doesn't work for me, can you do something about this, I ended up having to buy a math coprocessor. Because I didn't care originally. And I wrote a small software emulation layer. But there was this, I think, guy in Germany who said, I have a math coprocessor, and I can't use it, it only uses your emulation, and it's really slow. So I actually got myself hardware just to be able to support people who suddenly were interested in this project that I hadn't taken so seriously before. So this project is gaining momentum, you have people starting to contribute around the world, and you're taking a life of its own. And all of a sudden, you've unleashed the operating system wars. You have all these other Unix's going on, and FreeBSD, and Microsoft, and so on. And all of a sudden, it's not a little project anymore. So what's the dynamics here? Well, to me, I mean, you mentioned all these other operating systems, but to me, they didn't matter, right? I didn't care. To me, like... They cared about you. Not initially, but eventually they started noticing. But even when, like, we made jokes about Windows being the enemy, and we made jokes about world domination, and they turned out to not be so much fun after all, because the world domination kind of started happening. And the Windows jokes became pretty old, too. And for the people who actually were involved, it was never about the competition. It was really all about the technology and making Linux itself better. And it turns out, a lot of people asked, for example, so BSD had a lot of infrastructure already. And it's really hard to take other people's code and try to make that fit in your world, right? So we never even went down that path. And there was also the BSDI lawsuit, and everything ugly going on on that side, which would have made it a horribly bad idea. But it was not that we... Well, it was that we really ignored all the other operating systems, and there were people who were in both camps, but they're actually fairly rare. Now, you know, so one of the things... So we're in the middle, obviously your friend Dirk Hohendael worked for me at VMware, and we're in the middle of COVID, right? And all of a sudden, everybody has to work from home, right? And all of a sudden we have distributed development going on. It wasn't actually all of a sudden, because I'd been working from home. Precisely, precisely. And, you know, this idea of fully distributed development, and I think, you know, how often has the core Linux team actually even been together in person? I mean, you've always worked this way. So somewhat, you know, once again, you were like decades ahead of your time. We literally did not notice. In fact, somebody did the statistics, like in some of the early months, and people staying more at home meant that they were more productive when it came to kernel work. It did. We do actually have a yearly meeting where maintainers get together and discuss. Now, is that beer and pizza, or do you actually do work there? We do some work there. Okay, not much. But it is social, and it's a way to crutch about what's going on, and what's going wrong, and what's going right. And we missed that for a couple of years, but it actually happened two weeks ago. So we're back on track. We have met each other again face to face. But it's the real work. It's all done over email, and when we meet, it tends to really be about the social issues, not about code issues. Because code you can discuss over email, and you can actually quote the actual piece of code much more easily in email than you can in person. Yeah, and as we were in that period where now all of a sudden everybody had to do distributed work, it was like, man, you've been doing this for two and a half decades at that point. And it's wonderful. It is wonderful. I mean, you can tell a lot of tech companies have trouble getting people to come back to the office, I think. Yeah, it really has been much more productive. You didn't realize how far ahead of your time you were right there. Now, talk about open source in the context of cloud, right? You know, and cloud has taken advantage of open source in interesting ways, and just give us your perspective on that. You know what? There's no cloud is taking advantage of. There's more like some companies are not necessarily as good at really participating in the process. What do you mean participating in the process? So to be a really productive user, you don't have to just use the product. You have to interact with the whole process in doing bug reports, in helping support the project, and ultimately being part of the actual development community. And some companies are more open to that than others. And it's maybe, and it's not so much cloud or not cloud. Intel has been very good. I mean, I was... Thank you. Which actually surprised me, because when I moved to the US, we had a lot of problems with Intel engineers that could not participate because legal said, no, you cannot do that, right? That was a problem. And then something happened, and it was almost like a switch was turned. And it must have been something inside where the lawyers were told, make it work. And Intel became one of the more active participants. But we have cloud providers who are active participants and do a lot of development and do very well. And we have cloud providers who, I call them black holes. I will not name names. Maybe some of our listeners could guess. Yeah, some places you can probably guess. People go and become developers inside that company, and we never hear from them again. Yeah. So open source is really about not just benefiting, but it's participating and contributing to the ecosystem. And the participation doesn't necessarily have to be about actual development, because development is something... I mean, development is a part of it, but it's not necessarily even the major part always. We have... I concentrate obviously on development, but a lot of problems get found only once you have users. And having companies and individuals who just test and who participate in literally doing test cases, doing documentation, just sometimes teaching other companies how to participate has been a huge deal. So it's multifaceted, let's put it that way, that it's not necessarily... People think about open source as just the tech geeks who do the programming, but it actually is a lot more than that. Talk a little bit more. I mean, part of open source is also the licensing model, right? And how that has progressed as well. Just talk a little bit more on how you've participated in that and your views on where that is today. Well, I mean, we have multiple different licensing models, and I happen to like some of them, and I'm not a fan of others. But in the end, people choose the license that they are most comfortable with. And the open source initiative has a few rules for what a license has to do in order to be called open source. And they basically boil down to everybody has to be able to participate. And then you can have... The details may change, but you can't limit people's participation or people's use. And for me, personally, it has always been... The legal side hasn't been as important as the conceptual side. For me, open source is tit for tat. We all give each other something, and the tat is you give your improvements back. And it turns out, in game theory, tit for tat is considered to be a fairly good strategy in general, not just when it comes to source code, but just as a general approach to many problems. And I think one of the most important parts to open source for me has been that everybody needs to feel like they are in equal footing. We've had a lot of companies who do dual licensing, and it's perfectly legally fine, but it tends to result in a... In politics and internal strife about the project when one company or one entity has more rights than the other entities. So I've tried to... I've personally tried to always keep all my projects that very clearly... I'm not special. I'm the top level maintainer when it comes to the kernel. But if somebody else does a better job, and one day that will happen, they will take over, because I don't have any special rights except for the fact that people trust me, because I've been doing it for 30 plus years. It does seem like open source trust. I mean, this idea of contributing to the community, and having the exposure of the community, and then finally the trust of the community. Yes. But I mean, so you say, I don't actually like the notion of contributing to the community, because that implies a certain amount of altruism that I'm not a huge believer in. So... Okay. Tip for tat. I really think it's more like you're not contributing to the community. What you're doing is you're trying to make a project better for yourself. And the rules of the project end up then meaning that as you're making it better for yourself and trying to improve your own situation, you kind of indirectly contribute to the community. But you should always be motivated by your needs. And that's actually how a lot of these open source projects, including very much the kernel, have improved so much, is because I'm looking at all we've done in the last 25 years, and none of it was stuff that I needed, or I wanted. Because as far as I was concerned, 25 years ago, the system pretty much did what I needed. Yeah. And I've been very fascinated. When Linux was first starting, I was a young engineer, and watching this. And it was always curious to me how it would play in the enterprise space. Because enterprise, they need somebody to support it, somebody to maintain it. And some of the legal and security aspects associated with that. And that was somewhat surprising to me, how the whole enterprise licensed supportive version of open source emerged. A lot of it took... I mean, all the initial enterprise use tended to be grassroots use, where somebody in the IT... Yeah, some crazy guy, or often they were not in IT. Just decided I need to do a print server for my office, or something like that, and started using it internally. And it took years for... Especially legal does tend to be a bit nervous. And it's licensing, obviously, but it's in general, when you don't have anybody to blame, that makes some people very, very cautious. And having all the infrastructure, including having companies to blame when things go wrong, so that you have somebody you can call and say, hey, come and fix this. That took a long time. I mean, none of this has been exactly overnight. Actually, so Linux has been going on for a long time. Just to contrast that with the other project I started, Git. And that was almost overnight in comparison. It was like, that was a project that I kind of knew what I was getting into. And I've done this, I've started a project before, and now I'm starting a new project. And I was like, this is going to take years. And nobody's interested in source code control anyway. And CVS has been around for decades, and nobody has ever been able to really replace CVS. And initially, that seemed to be very true. And then that was overnight. The Ruby people, strange people started using Git, and suddenly just exploded. And that was very different from Linux, which was interesting to see. Yeah. Now, let's move forward a little bit. AI and machine learning, right? Obviously, lots of open source projects going on in that space. But in machine learning, the data is maybe more critical. The training sets, the data sets associated with that, right? Say a little bit more about your views on how that could and should evolve. And do we need more open data initiatives to complement what I'd say is more open source here? The code can be small, the data is big, right? Right. I mean, we've seen that in other projects, too. It's not just AI. I think AI is the one that people are now very aware of, just because it's such a big issue, both technically and from an economic standpoint. But if you look at a lot of other software and hardware issues, you do see some of the same things where game engines, for example, is something where, on the one hand, you have the code, and on the other hand, you have the data and the game environment, and the two are fairly clearly separated. And you can have one game engine that runs many different games, and a lot of the intellectual property ends up being on the pure data side. Yeah. Well, and even here, I mean, when we think about things like ethical AI, a lot of it has nothing to do with the algorithms. It has everything to do with the training sets, you know, for that algorithm. And if they're closed, do I know that this was a good model that was produced associated with it? Well, and we're seeing all the same issues that you see with people when you have data sets that are biased in ways you don't even realize. Yeah, no. I have to say, I only get involved with AI in the sense that we see it on the kernel side on the hardware space, where we provide drivers for the accelerators. And I've been so hyper-focused on the kernel that anything I say about AI is likely to be wrong anyway. And I love what you said to me in our stage interview. I was trying to get you to predict the future, right, you know, 20 years down the road. And how did you describe yourself? That I'm a plotting engineer, right? And I take pride in that, but I realize it also means that there are these people with visionaries that want to talk about the big picture, and I don't tend to be one of them. Yeah. Well, and I think there is something in that constancy of purpose in a consistent direction, right? You know, right, you just keep progressing. And you need all kinds, right? You need the kind of person who sticks to one project for 30 years, which, I mean, not everybody will do, right? Yeah, that's a different personality as well. So here we are, right? We're at the Intel Innovation Event, right? We say, you know, and as I like to say, you know, we used to have the Intel Developer Forum, and I am bringing back that spirit, engaging with developers, you know, rebuilding of, you know, our cred in the open community, open standards, you know, interfaces, Wi-Fi, USB, PCIe, and now we're doing chiplet standards, etc. And as you would look at us, how's Intel doing? Well, I think everybody knows that issue. The problems on the execution side have been pretty painful for everybody. And it's, I don't, I mean, I don't see any of the hardware and fab issues. What I see, as I've seen, are kind of the fallout from the Intel software development standpoint, where the last few years have not been quite as ebullient as they were before. And people have not been as happy with being inside of Intel, because I don't think Intel has necessarily been a very happy place for a few years. And I'm hoping that's changing. Well, that's certainly my job, right, to change it as well. And, you know, we're, we're getting on it, right, as we rebuild. And obviously, having events like this one are about rebuilding our commitment, our engagement, you know, with the community, you know, participating, driving standards as well. Other advice for me? So what I really liked about Intel is when you do standards based around hardware. I loved how NVMe turned out. PCIe is obviously another great thing. The x86 architecture in general. I just think, put good hardware out there and document it enough that it will be the standard. That's, that's all I ask for, because that makes my job so much easier. We, there are clearly other companies that don't do the documentation part, for example, and don't help. Well, that's why you liked me in the first place. I documented the 386 really well. Absolutely, right. No, documentation is important, even if I'm probably the wrong person to say that, since I never write anything myself. How dare you? Yeah, no, it's, it's, it's being a bit, yeah, asking others to do something I can't do. But it's also participation, yes, being, especially for a hardware company, when, when it takes years to get a process or a processor or, or other new hardware out into the market. And then it takes like a week when you announce it, and it's, well, it hopefully takes a week when you announce it, and it's on shelves. For us to support it, it usually takes months or years. So being proactive and actually interacting with the community before the product is out. And Intel has been good in that sense. Yeah, well, and tomorrow we'll be announcing, you know, early hardware on the Intel Dev Cloud, right, which, you know, starting to put up our products, you know, months, if not year or more, you know, that developers start to have access to them sooner and starting to really, you know, and to me, I'm just passionate about this, this re-engagement, you know, that people can start saying, oh, that's a cool idea. And we're ready to be tested on our ideas early in the process, get that feedback early, get the hardware out there in their hands. And since I have your ear, and this may be cut out, the thing I sometimes despise about different companies, and Intel has had that problem, is when you do market segmentation, where certain features only exist in certain markets that aren't necessarily as available as others. And I'm thinking of things like the transactional memory. I mean, it's gone now. Maybe it's never coming back. But that was very painful for us when only certain server CPUs had it, and the people who actually needed to develop for it couldn't access it easily on their own desktops. And that's, I think, an area where Intel can maybe improve and not let the market segmentation drive these kind of technical issues that make it hard for developers. Well, I look forward to always getting your feedback in the future, right? And, you know, maybe just one fun little way to finish our time together. What is this about penguins? There is no deeper truth behind the penguin. It was literally, I was in Australia many, many years ago, and there was a fairy penguin. They're literally, they're about a foot tall, and they weigh a pound or something like that. And it was in an enclosure with a low thing, and I tried to get it to come closer by making like a fish, and it bit my finger. And I thought that was very cute. So when we were looking for a mascot, it was like, penguins are kind of cool. And they're like, well known enough that people know what a penguin is, but they're not like a dog or a cat or something like so common that they're not really noticeable. So it turned out, it was funny, a lot of people thought that the penguin, especially the Linux kind of cartoony fat penguin, was a really bad logo, because this was late 90s. Like, Linux was not that corporate at the time, right? So a lot of people felt like having a cartoonish logo was a bad thing for a project that wants to appear. Exactly. Yes. And then the whole dot com thing happened. And it turns out that on the web, in particular, having a friendly and graphically interesting logo is actually a good thing. And I'm told that the Linux logo has actually been used as an example of great logos in marketing. Great technology, you know, the entire open source movement, and the pinnacle of branding as well. Yeah, no, it was completely unintentional. And I did not design the logo. I just gave a high level explanation of what I wanted. And I made it a kind of competition for who can come up with the best one. Well, and then others did the graphical part. My wife and I did a cruise to Antarctica. And we saw tens of thousands of penguins. So anyway, it's a pretty, you know, we like them too. Yeah, no, it's a nice, it's a nice animal. Although the Linux penguin is kind of an unholy mix of penguins and ducks, because penguins don't actually have like the orange feet that the Linux penguin has. Very good. Well, you know, I am again, so honored to have you join us at Intel Innovation. Thank you for taking time for a more in depth interview. Joining us on stage, you know, being here and of course, becoming the recipient of the Intel first ever Intel Innovation Award, I couldn't think of somebody, you know, more impactful for the evolution of technology, the development of entirely new software, you know, model and deployed in something that, you know, literally has become the operating system of choice for so many industries. Thank you so much. Thank you.