Open Source Is the Inevitable End State for AI
Open source is no longer just a software movement — it is the foundation the entire AI stack is built on, from CUDA to Kubernetes to the models themselves. Duane O’Brien, the new Executive Director of the Open Source Initiative (OSI), joins Alan Shimel on Techstrong TV to talk about the organization that has shepherded the Open Source Definition since 1998, why OSI-approved licenses are the Good Housekeeping Seal of the software world, and how OSI is taking on its biggest challenge yet — defining what open source actually means in the age of AI. Duane shares his journey from PlayStation play tester to leading open source program offices at PayPal, Indeed, and Capital One, and walks Alan through OSI’s growing role in policy, education, and advocacy. He and Alan dig into why the Open Source AI Definition (OSAID) matters, why AI systems don’t fit neatly into traditional software categories, and why — despite every attempt to build moats — open source will always win in the end.
Transcript
Hey everyone. Welcome back here to another Techstrong TV interview. I'm really happy to introduce my next guest here.
His name is Duane O'Brien. Duane is the executive director of the Open Source Initiative, and let's welcome him to Techstrong TV. Hi, Duane.
Nice to have you on. Thanks for having me. Super glad to be here today.
Absolutely. So Duane, I mentioned you're the ED over at Open Source. Is ED a good way to describe it?
I don't know. That's the most- You're the ED at Open Source Initiative. Mm-hmm.
But talk to us about your background. How did you get here? How did I get here?
I was always the computer kid in elementary school and cut my teeth writing code on Commodore 64s in the camera jewelry department at Kmart way back- Mm-hmm ... when mom to get off work. It wasn't what I went to school for, but I really started off in the late '90s as a play tester for Sony PlayStation games.
And so I started writing bugs for games, and then I started writing bugs for software. And if you look at some of those late PlayStation one titles like "Syphon Filter" or "Spyro the Dragon," my name is in very little print in some of those games. Really?
Very cool. So I started writing bugs for games, then started writing bugs for software, and then I had a variety of roles where I described myself as the Perl guy on the Java team. I was a build engineer, performance tester, for a number of different organizations.
It was really a little over 10 years ago that I shifted from that kind of work to either running operations or building open source program offices in the corporate world. So operations for the open source program office at PayPal. Built and led the open source program office at Indeed.
I was there for about five and a half years. Led an organization called Collaborative Engineering at Capital One. I was there for about two years and a bit.
And throughout all of those roles, the focus that I brought was how do we help these companies and these organizations show up as meaningful partners with the open source community. Excellent. It was just a couple of months ago that I accepted the role here as executive director, and I've been here with the OSI ever since.
You weren't at Capital One when Topo was there, were you? No, I don't think so. Topo Pa?
No? Because he initiated a lot of their open source initiatives, actually, back when it was kind of unheard of that a financial institution would actively say they were using open source, let alone writing open source and releasing it- Yeah ... to the community.
Anyway, great story, great background, and it's not an unusual path. A lot of us in tech started out as gamers. Uh-huh.
That's what drew us in, and then a little deeper, and here you are. Yeah. The Open Source Initiative, OSI, I think a lot of people have heard of.
" But for those who maybe are not even familiar with OSI, or really, they've heard it, but they don't know it, talk to us about the Open Source Initiative. Sure. So we were founded in 1998, as the shepherds of the open source definition, a set of 10 criteria that must apply for a software license to be considered open source.
And those 10 criteria are built around making sure that we respect the four freedoms that we understand from open source and free software. Use, study, modify, and redistribute. And that's the work that we're most well-known for.
Especially if you're in an organization, trying to figure out if you can use software if it has an OSI-approved license. That's shorthand for, we know that this is probably okay for us to use, and we might have certain obligations as well under the terms of that license. But it was instrumental in helping drive that common shared understanding of what we're all talking about when we use the words open source.
I love it. A little less well-known than the license work, but increasingly important, is our work around policy and education, and advocacy. So we also advocate for the usage, adoption, creation of open source, and in the last few years, we have been increasingly involved in helping to support legislators and policymakers who want to understand open source as they work to craft meaningful tech policy.
And then about four years ago is when we started down the journey to craft the open source AI definition, as we recognized that there was this emerging dead space in the industry where a lot of organizations were talking about their AI technologies as open source, but we did not have that same shared understanding of what that meant. Agreed. Before we jump into AI, because we're going to spend, as I often do on these things, it takes over every conversation.
I want to spend a little time talking about OSI, not just in regard to AI. Sure. You mentioned the licenses.
I don't want to fluff over that. It's an important piece, not just for the Open Source Initiative, but for all of us. As you mentioned, if you have a piece of open source software and you see it as an OSI-approved licensing, that's like the best Good Housekeeping Seal of Approval you can get.
Mm-hmm. When it comes to this stuff. I learned this the hard way A company I co-founded 2001, and I don't know, maybe around 2005, '04, we decided we're going to release a community project.
Uh-huh. Not quite open source- Mm-hmm ... but free.
You just couldn't get to the source code. Right. And we said, "Oh, what's the big deal?
" And we wrote our own license for it. Uh-huh. Man, did the stuff rain down upon us.
What a lesson. Sometimes you got to learn your lessons the hard way, Dwayne. I still got the scars to prove on that one.
But it is important, right, especially as we get into this age of AI. IP matters. Yeah.
Licensing matters. And using an OSI approved license for software, whether we're talking about AI software or not, is an important, as I said before, the best Good Housekeeping seal you can have. Yeah.
Are there still just seven approved licenses or eight? I forget. Oh, there's many more than that, and we- Yeah ...
see licenses submitted for approval on a regular basis. It's not quite as fast and furious as it might've been 10 or 15 years ago. In my opinion, there's only so many different levers you can throw with the language of- Right ...
licenses before you've hit every combination of them. I don't know off the top of my head the exact number, but there's more than that shortlist. Yeah.
And I would guess, when you said the stuff rained down on you, this is one of the things that you encountered. When you use a novel license that no lawyer has ever seen before, every legal department of every company and everyone who ever wants to adopt your piece of software has to have a series of meetings to figure out what it means, and do they have obligations, and do they understand what they are. And so in most- I mean, don't get me wrong, we had lawyers draw this license.
Oh, of course you did, yeah. Yeah, but- But the lawyers on the user end hadn't seen it yet, and so they hadn't- Yes ... talked about it.
And that, yeah. And- And look, lawyers are going to lawyer, right? Yeah.
That didn't bother us. " Yeah. "And this license isn't approved, and I'm not using it as a result.
" Yeah. And it was a painful lesson- Yeah ... that we learned and redid.
Dwayne, if you don't mind, I want to pivot into the AI stuff. Yeah. Right?
So in my mind, when we talk about the AI open source movement- Uh-huh ... it's kind of like an 8,000-pound baby or an 800-pound baby, right? There's a lot of open sourced tools around AI, everything from the models themselves, especially coming out of China and stuff.
A lot of that stuff's open source. But also the surrounding software, the stack, if you will. Uh-huh.
Right? Nvidia, the biggest company there is today. Mm-hmm.
Five trillion dollars. They've open sourced a lot of the CUDA stack. Mm-hmm.
A lot of people are building their AI apps on top of the cloud-native stack at a Linux Foundation, right, with Kubernetes and so forth. So open source is already a major factor in our AI software. Uh-huh.
However, I still feel it's really immature, and we're just scratching what's happening here. What's your take? It's an interesting question.
When I think about the systems that are built on top of all of these other open source technologies, and first, I don't want to drive past this too quickly. You've rightly pointed out that these AI systems are built on top of open source technologies that came from open research that was widely shared among other folks. And we got to this path through that collaboration to solve shared problems, to share knowledge, and to gain an understanding of the space.
And we should never lose sight of that while we're in the middle of this race to figure out who gets to control it all at the end, right? And so the systems that are built on top of these technologies, though, they don't conform easily to our traditional understandings of software in some cases, right? What I have understood from the AI practitioners that I've talked to is that you might have everything that someone had to put a model together and follow all of their steps and get something almost but not quite the same, because they're non-deterministic, right?
Yeah. Like having twins that are raised exactly the same but are slightly different, right? And so the complexity of those systems and the novelty, they're still only a few years old, means that we're still all trying to figure out what it all means for us, what the right guardrails are that we need to put in place, what the right processes are that we have to put in place, and it's moving very quickly.
So it's even more important because it's moving so quickly that we continue to share and grow our understanding of the space together. Agreed. You said something in the beginning of that segment along the lines of this will help us before we figure out who controls all this.
Huh. And really, you know what? It kind of comes down to that, Dwayne Right?
Who's going to control this? Can it even be controlled? And a lot of people, they look to open source like Obi-Wan, you're our last best hope, right?
And open source will prevent any one entity from-- This is civilization-changing, perhaps technology. Mm-hmm. And should any one entity, whether it be a government or corporation, a nation-state, whatever, should any one entity control it like that?
Mm-hmm. And is open source the ultimate hedge for freedom here? Yeah.
I think of it a little differently, but I don't think we're thinking about it too differently. It is the principles of openness that we have bundled together to call open source, that inevitably, historically, come out on top in the end. Right?
The people like to collaborate and build things together. We can't help it. You put a bunch of people together in a blank space, and they're going to figure out something to do.
And you put a bunch of people who like working with technology in a blank space, and they're going to work together on technology. And whether that be through code or something else. And inevitably, these closed systems gravitate toward an openness in the end because there's something about that in our human nature that we just can't resist, right?
And so when we look at organizations or entities that are trying to sort of get their hands around and... I do a lot of reading about AI companies, and they talk about sort of trying to build their moats or dig their moats or entrench their moats and everything else. And these technologies and the people behind them, they always will wind up in the open one way or the other, right?
And so, I don't think of it as a hedge. I think of it as an inevitable end state, right? Because of the thing that compels us to collaborate with each other.
Dwayne, I've been involved in open source software for a long time, 25 plus years. And my feeling always is we've gone through stages of open source software, really open source in general. Because there's open source hardware, too, right?
Yeah. With spec. But first was kind of very much the cathedral and the bazaar.
You're literally in a bazaar buying things you didn't show what you were getting. Mm-hmm. The Wild West, if you will, like from the book.
Then we had this big brother kind of era where a lot of open source projects were really managed, owned, shepherded by one corporate entity, basically for their own enrichment, though they gave lip service to helping the community. And then we had this foundational era. Mm-hmm.
We have the Linux Foundation, the Eclipse Foundation, whatever, where it allowed cooperators, coopetition, right? Right. Companies that would otherwise compete with each other to all contribute for the common good.
Mm-hmm. And it really has given rise to this golden era- Mm-hmm ... where open source has become the dominant form of software.
Mm-hmm. Open source, 99% of organizations running open source. With the AI, I think the job of the open source initiative and companies and organizations like you is to make sure it stays like that, right?
Mm-hmm. It can't change. We can't go backwards here.
Yeah. And I'd like to hear from you, what exactly is OSI doing to ensure that, to work with the community, and other foundations, other initiatives, right? The whole open source AI movement.
Yeah. How are you guys playing in there? We could do an entire second episode around the golden era- If we have to, we will.
and what AI means in that context. But I want to call to some of our more recent work. The work that we did helping the G7 sort of formulate their vision on AI openness and the work that we did to help the EU work on their strategy, their sovereignty package that includes open source kind of through and through it, right?
In both of those cases, we walked into conversations where there were policymakers or legislators really wanting to do the right thing, but not having the kind of technological background that you might need to craft good policy or craft a good vision, right? I went and looked some of this up last night, and we're almost 20 years to the day since the senator famously said in America that the internet was a series of tubes, right? Mm-hmm.
I remember that. You remember that, right? Uh-huh.
So legislators, they know how to craft policy, and they know sort of how that world works, but they need input from the community in order to make it good, right? And as we've seen in particular in the AI space, there was a period of time where open source as a term was up for grabs. We saw a lot of companies produce models that they claimed were open source but had fields of use restrictions or other kinds of restrictive licenses that inhibited the four freedoms.
And we had to, as a community, work to correct that. What we saw in the G7 vision With a couple of really important clarifications. One of them was that their open source AI definition hove very closely to ours.
They added sort of four points on this journey from weights available, which is proprietary licenses. And on the other end, you had open source with open data, which means everything soup to nuts was open source or using an OSI license. But the big win I thought was their clarification that open weights means weights that use an OSI-approved license.
It doesn't mean whatever we decide to call it. It doesn't mean whatever a company that produces a set of weights says is open, but you're not allowed to use it in certain countries or for certain purposes, right? That it's open, right?
Full stop. And it's that kind of work that we need to do and that the OSI is doing to continue to make sure that lawmakers, legislators have an understanding of what openness means, that we are following that same shared understanding so that when they walk into the next room and say, "We're talking about open source," everyone knows what we're talking about. Nobody has a different vision in their head of what that means.
And that work is constant. And a lot of it's behind the scenes, and I don't think the OSI has talked about it very effectively just as a group because we've been too busy doing it, right? So one of my goals as ED is to make sure that we're out there telling the story more effectively about how we are doing that kind of advocacy and education work behind the scenes to make sure that the next sets of tech policy that come out are more open source friendly.
And if you look at the sovereignty package, and we wrote a great blog post about this, we didn't get everything we wanted out of it, but we got a lot. And there's a lot there for the open source community to be excited about. And so, that's the footing that I want us to take and I want to everyone else to take.
When we see these policies come out, when we see these proposals come out, when there are wins in it for open source, throw your hands up and celebrate a little bit. " Right? This is a process.
Don't let perfect get in the way of good. Yeah. Right?
Exactly. It's classic. Yeah.
Dwayne, we're coming on the end of time. We're probably over time already, to tell you the truth. But for people who are saying, "You know what?
I believe in this. I want to get involved," how can people get involved? What could they do?
It's a great question, too. There are a number of working groups that we run with the OSI, that you can come and participate in. The lists for discussing licenses and understanding if they conform to the definition and so on are great places to just plug in to understand how that process works.
I think the best thing people can do today, who love and care about open source and want to see it continue through forever, is to work with the people in their immediate communities, their legislators, their policymakers, staffers in those organizations, whether it be anyone from your town or your county to your state or your federal or above legislators. Reach out to those offices. They hold public hearings.
They ask for input from the community. Show up. Carry the open source banner into those conversations.
Help the people who don't understand the specifics of how open source works or have misconceptions about how open source works. Help them understand how it does work and what we're fighting for. And help make sure that we're all pulling in the right direction.
Love it. Dwayne, we didn't even mention a website. I apologize.
The Open Source Initiative, what's the website for people to go get more info? We're easy to find. org.
Can't beat that, man. Dwayne, I want to wish you luck as ED here at the Open Source Initiative. It's an important time, an important position and important organization at an important time.
So did I say important enough? I feel pretty important if that helps. And Alan, thanks a lot.
I'm excited about what we have coming and it is a very critical time to be in this role, so I'm grateful to have the community support. Well, come back and keep us posted on how things are going and what- Will do ... our audience needs to know.
All right, Dwayne O'Brien, Executive Director for the Open Source Initiative here on Techstrong TV. We're going to take a break. We'll be back in a moment.