Open Source AI Challenges the Economics of Closed Models | The Open Current Ep 4

Hey, everyone. Welcome to another episode of the Open Current with Alan Shimel and Margaret Dawson. Hello.

Margaret, it is... Hello, and it's great to see you. I know you have been globe-trotting.

Lucky you. But welcome home. Thank you.

Yeah, I was in Shanghai for a week and Singapore for a week, and my learning is this is going to be one of those first-world problems where I got business class on Singapore Airlines, which lovely airlines. Which does not suck, my friend, yeah. It does not suck, except I will say the beds were different.

I don't know what plane that was, but there's a lot of different ways you can kind of put down a chair into more of a bed, and I was too tall. You'll remember- You were too tall? I was too tall, so I kept hitting the side.

So I had to sleep kind of sideways. So the first thing- Oh, goodness ... I had to do after I landed was call my chiropractor and I I would imagine.

I'll be honest with you, I have flown Singapore nonstop from Newark to Singapore. Yeah. Was it not Shanghai Airport, Chiange or whatever airport.

And, no, it is, it's C-H-I-A-N-G It is. Yeah. Yes.

But anyway, I flew, I think it was the Singapore Airlines 350. Okay. And so not the double-decker, but the really big Airbus.

Yeah, this one wasn't that really, really big one, yeah. Oh, okay, because those beds are quite comfortable. Were they nice?

But maybe you were in first class, not business class. No, I was in business. Okay.

But- Anyway, how were the beds? And I'm taller than... It's great.

Yeah, you're taller than I am, so I don't know. Yeah. No, the service is great.

The food is good. They are. So did you fly directly to Shanghai, or you switched?

I flew directly Seattle, Shanghai, and directly Singapore, Seattle. Oh, that's a pleasure. So it was a 15-hour flight, so it's magical.

I will never complain because as long as it's direct, I feel like you could do anything. Absolutely. I can sleep, as my mother tells me when I was a kid, I used to fall asleep in the middle of the sidewalk, so I'm fortunate that I'm like a baby in a car seat on an airplane.

But regardless. Good for you. Yeah.

I feel like I'm just now getting back to normal. It's been a very hard re-entry on many levels. It is.

Asia was amazing. It was just absolutely incredible. The energy at KubeCon Shanghai was incredible.

I was actually surprised, and there were so many people from CNCF and the Linux Foundation, PyTorch conference, first-time PyTorch conference has been part of that whole kind of the KubeCon, I can't remember all the different things. It's like four different conferences in one that they do now. So it was cool.

It was really cool. Yeah, no, Shanghai's an amazing city. I was there, believe it or not, I did a DevOps Days there.

Oh, interesting. Where Techstrong was one of the sponsors, and we brought a video crew with us. Yep.

And at the time, I was a co-founder of the DevOps Institute, and so we were a sponsor of that as well, and I had partners in China that we were working with. Yep. And what I was totally amazed at was the passion for open source.

Was the- 100%. That's the country. Yeah, let's go back.

It was. I was in Shanghai last, I want to say 2017, 2018, something like that, so it's been a while. But that was for an open source summit event, and it was the same thing.

It was just the energy then, and it's been a long time coming. And there's a lot of discussion about China, and I don't want to get into the political elements or- The politics of it. Yeah.

But I would say just the innovation. China, in the old days, if I think about Microsoft originally, there was a lot of concern that they were taking IP or there was a lot of noise in the system around that, I'll just say in the '90s. And I think what is the noise in the system now is can you trust open weight models from China or open source models from China?

And my answer is, well, if it's open source, then you can look at the code. Right. And so maybe open source in my optimistic Pollyanna view of the world that I sometimes get into is wouldn't that be the equalizer if we are all working from open source projects and communities and code bases, doesn't that automatically remove the threat of a bad actor because so many other actors from all over the world are able to inspect and audit?

So let me be the geek here a little bit. Okay. Oh, good, because usually you make me- So the difference between open weight and open source.

Mm-hmm. Right? Open weight, we know the 40 billion or one trillion parameter model and what it says, and we could run that model locally.

Yep. What we don't have insight into in the open weight models is what happened behind the black curtain? How is this concocted?

Right. What was the recipe? Right?

In other words, the source code, if you will, right? That was always one of the great frees in freedom, source code and open source. We don't necessarily have that piece of it.

That's right. What we're given is a finished product that's pretty transparent, and we can run locally. So it's almost free as in beer, if you will, but not free as in freedom.

But we don't know... And if you believe Dario Amodei or Amodei, as they say now, they're distilling that training by using Claude or Chat or our models. Right.

Which is pretty ironic when there's a gazillion lawsuits against Anthropic and Chat for doing exactly that, to my IP. But there's nothing stopping an open weight model deciding to be an open source code base. No.

So I think you're right, yeah. But here's the bigger issue, and I think we should be clear on it. Margaret, the Chinese people are, and I don't mean to stereotype or racist or anything like that, but the economy in China is powered by the resourcefulness of its people, and that resourcefulness, they're passionate about open source because that rising tide lifts all the boats and allows them to experiment from a common level, if you will.

I think if we just talk about the entrepreneurial culture, the underlying culture, and what I'll go back to is I spent most of the '90s in Taiwan, and the thing I always tell people is when I was there, I literally was able to take on anything that came in front of me without any blockers. I'd never experienced that in my entire life. I ended up starting a company, I ended up being a journalist.

I ended up doing all these things that if I'd been sitting in America, I never would've been able to do. And- Agreed ... the reaction I received from people in Taiwan and Hong Kong and China, even a little bit in Singapore is a little more corporate, but I would say in general, in those countries is okay.

Like it was just an automatic, like you're starting a company or you're doing this, you're leaping off the whatever it is. " And they go into supply chain mode. They go into support mode.

They go into, how can I help you? What are you missing? Or do I know someone who could help you?

Whereas when I would say the same thing, and I was a young woman, I was in my 20s, right? I'd come back to America and say, "Oh, I'm starting up this company, and I already have my first client," and blah, blah, blah, blah. And they'd go like: "Well, you can't do that.

You're not even Chinese. You've never started a company. You've never even done that.

" And they didn't care that I was a woman. They didn't care that I wasn't Chinese. They didn't care that I was doing something that they knew or didn't know anything about.

It's like the land of possibility, and I feel the same way- You know what? People used to say that about America, Margaret. This was the land of opportunity.

And maybe it is still that way. I will tell you, as a woman in tech, I don't always feel that way. There are times I do.

I agree. But I would say it's universal. It's a universal feeling, and that the For the most part, failure is looked at differently.

It's looked at more as an experiment or something. Well, that was the wrong way to do it, so try it this way and maybe you won't fail, right? As opposed to, oh my God, you failed, your life is over.

So there is this energy, there is a lot of energy around open source, but I always feel like more the art of the possible when I'm in China. First of all, the fact that you went out and did that in your 20s and did that, good on you, man. That is- Oh, it changed my life.

I don't think I'd do the same, but yeah. Wow. You're a tiger.

I think back to what I was doing. Well, because men mature later. I was a child in my 20s.

Are we going to start this? I think it's- No, let's not go there. But it's interesting- It's in your 50s, right, isn't it?

You're about 35. I always think men mature when they're about 55. Yeah.

Right around then we start understanding. Well, we'll do that on another podcast. I want to come back to PyTorch.

Okay. So the PyTorch conference, and I want to talk about that- Okay ... and this Open Secure AI Alliance- Alliance ...

that popped up. I think Jensen Huang and the NVIDIA folks were behind that originally. Yes, they were.

Two things. PyTorch, again, part of KubeCon and Linux Foundation. Yep.

Open Secure AI Alliance joined the Linux Foundation. Yep. I think we are seeing the coup of open source taking over AI infrastructure.

Oh. The question is who's going to own the AI stack? All of us, none of us, it's going to be open and we all can stake our claim.

We have to intersect this momentum. Oh, that was such a we are the world moment that you just said. But with that, then we have all this momentum around governance.

Do we slow down? Do we put the brakes on? Which is causing a lot of debate, I'll just say, in a healthy way, right?

I think everyone agrees governance is needed, but in a perfect world, as governance is needed, if open source is raising at the same time, does that become the ideal intersection? I would love to be as optimistic as you. I'm just not seeing that consistent energy around making sure all of it is open end to end.

We still are just seeing so many models that are being done in a black box, that you can't inspect, that are controlled by a singular company. And I think we're going to continue to do that because there's just always going to be this divergence of opinion and strategy that there's just always people that think complete control, which to them means black box even if it doesn't, is the only way to do something. So- And they think it's for the benefit of society, like, "Don't worry, we'll control it because if we open it up, people will-" I know, we'll tell you what's good for you.

Right. Yeah. Like the Pink Floyd song, "Mother" with- Exactly.

What do they call that? A compassionate dictatorship? No.

There's another word. Yeah. No, not the compassionate dictator.

Compassionate conservatism was George W. Bush. But I'm- Oh, I have compassionate capitalism, but that's not the right term.

It's like dictators that think they're taking care of their people. There's a word for that. It's like Uncle Joe Stalin.

Exactly. But- Exactly ... so let me tell you some of my good news on that.

Okay. I think what we are in now is what my grandmother would always tell me, it's just a phase. Don't worry, it's just a phase.

They'll grow out of it. And I think this is a phase where we need to come to terms because we haven't reached consensus. This is too important an issue for us not to, at some point, have consensus.

Agreed, but the problem is that politics is bumping into technology. Absolutely. And we know as soon as that happens, we lose sight of the technological purity of the debate, and it becomes mired in a whole lot of other b******t.

Yeah. Glad to say that word on a podcast. To put it bluntly.

But here's something. So for instance, just this week, both Anthropic and OpenAI, and let's face it, when we're talking about people who are doing black boxes, those are the two biggest ones. Mm-hmm.

Both Anthropic and OpenAI put out new releases. Mm-hmm. Fable, they're not Fable.

Opus 5-5. Yep. And then ChatGPT six, but not- Six?

5 or something? 6 I think is actually what it is. 6.

But it's not Armis. That was already out. They came out with a Sol and a Luna version of it, so at a 50% discount because here, economics, the market is ruthless.

It cuts with a guillotine. And the thing- Although token maxing is still going on. We could do a whole thing on the token.

Oh, yeah, token. It's nothing to token maxing it. Because I think that's going to blow up.

I think that's going to blow up on us. It's tokenomics. Tokenomics is the word, right?

The economics of token usage cuts like a knife. Sounds like a song. But it does.

And so as these open A really, really bad song. All right. Well, this is why I do this for a living.

But anyway, that is going to drive these companies because if I can do 80% of what I need to do without paying you those exorbitant token fees- Right ... I'm going to. That's the way the market works, and that's the way the open source- Remember what I've always said, so we talked about this in a previous thing, where it's like you always look at a company's business model to understand how much they care about your security, your privacy, and control of the data, or control of the code, right?

If their business model is they make money off that, then that will not be the thing that you get to play with, right? That's where you'll pay. That's in my book, actually.

That's the big point of it. Oh. 5, did submit it to some independent auditors.

I use the word auditor, that's not the right term. But to some independent reviewers to- Were they agentic or human? Human.

Using agentics, but humans- Okay ... about whether the right guardrails, the right kind of focus was put in place that these agents wouldn't run wild. And I believe OpenAI similarly did this.

" That would be a good song. Maybe you and I should start a band and the heck with this podcast stuff. Yeah, we should definitely not.

You should not leave your day job. That's all I'm going to say. Well, you haven't even heard me sing.

I mean, I actually sing. Are you a musician at all? I was just going to say, yeah, if you want me to bang a log.

My wife says I have no rhythm anyway, so you don't want me in your band. Okay, but while we're on a complete non-sequitur, I remember the word. It's benevolent dictator.

Yes. That's it. So this is a sign of growing older, right?

I know. We forget our words and then they come to us in a bit. But you're right, it's a benevolent dictatorship, and- Yeah ...

I'm not signing up for that. Yeah. And I don't think most of us want to.

So I- But we're getting that without asking for it, right? And we're told to trust the organizations that were the initial innovators of this because they're allowing us to take advantage of this incredible innovation. So- Yeah ...

it has that same benevolent ring to it. It does. It does have that Big Brother kind of we'll tell you what's good for you thing.

Right. Timeout. This thing came out.

But here, Margaret, I do believe in the power of the market. I believe in the power of open source. Agreed.

And I think if the market and the tokenomics of it, and the open source versions are so good, these, again, not blaming ChatGPT or Anthropic because Grok is part of that, Google Gemini is part of it. They're going to be forced to- I agree ... to deal with it.

So- And you know what? It's funny because one of the things I talked about in my keynote in Shanghai was open standards, not just open source, but open standards, and looking at other things that need to be open And if you look at standards from history that were closed and why they opened, the thing I say is these didn't open because the company woke up one day and said, "You know what? Let's just license this technology to everyone.

Let's make this accessible so it's a standard, so people only have to buy one widget and not 500 widgets" or whatever it is, right? And you can look at, I mean, the most recent one would probably be the Apple lighting connector, right? Which led us to the USB-C, because remember, we used to have every phone had its own connection.

Own little doohickey connector, yeah. Right, as well as computers. I mean, the USB-C is now going into a lot of the laptops.

So why that happens is because the market... And I just froze. Hold on.

Oops. I hear you, but you're frozen. I know, but I'm frozen, so let me get unfrozen.

Yeah, but Margaret, don't worry, this records locally, so you don't have to worry about the internet. Okay. So let me just finish that thought.

Okay. It was the market and their customers forcing them to open and create a standard. It wasn't the companies that created that initially, right?

And so there's always this momentum around a consortium of companies, so sometimes it's other companies in the industry that are saying, "This is ridiculous. " But at the end of the day, it becomes a market-moving thing. So I hope you're right.

I hope the market and all of us out here using these models start saying, "We don't want this. We want this more open. We want to see what you're doing.

" So it's funny, you're talking about a hardware standard. Right. But let me give you a software standard and one that's probably near and dear to you and SUSE.

Linux versus UNIX. Yeah. We're old enough to remember UNIX and all the different flavors of UNIX, and everybody had their own UNIX, and it was a Tower of Babel.

It was crazy, right? Yeah. Linux comes along and just wipes it out, right?

Yeah. The power of open, just like a tsunami. And so quickly.

It was interesting because Linux just celebrated its 35th birthday. 35th. Linus put out that first Linux code, and SUSE just celebrated its 34th birthday.

So not only did Linux very quickly become every developer's favorite flavor of the month, but how quickly it started becoming a way that enterprises wanted to consume software, but they wanted it with a little more stability and support. And so for a company to be created one year after an open source project, which we're seeing that a lot now, but 35 years ago, that was pretty rare. Now we're seeing that a lot.

In fact, a lot of companies are built immediately out of an open source project. Absolutely. But I think in that case, it was incredibly fast for something that was brand new.

The open source development model for enterprise software didn't exist before that. No. But now it's so well understood and so ingrained into the market, both here in the US, in China, in Europe, all through the world.

Right. But go back to what you said before. It is a known thing, and yet how many companies truly follow that open source development model, and you can say, I guess, business model, where you're staying true to the code and what you're offering enterprises and governments is a way to have it managed, lifecycled, supported, patched, all those things, right?

" That is still a lot more common. Open core, open components. Yes, there's open core.

Right. In my mind, SaaS, the whole SaaS model made open source business models viable. Prior to that, doing just support and training, right?

I mean, though one could make some dollars, it wasn't going to set the world on fire, right? It's just not enough profit margin there. Right.

But offering SaaS models around it and open core, too. I know a lot of purists in the open source space kind of sneer at open core, but again, if you're an entrepreneur and you believe in open source, but you want to maximize your profit potential, open core is a viable- Look, I'm all about people leveraging innovation in the wild to build products. I think the thing where I draw the line is that if you're going to do that, then contribute back.

And what happens- And that's the key. Don't be a parasite. Right.

Because a lot of times what I've seen with open core is they don't actually give back to the community. So you took advantage of that innovation, it's helping you make money, then give back so other people can benefit from it. Well, and a lot of times, though, the proprietor of the open core project and the one making the freemium modules is the same company.

And the key there is, are they keeping the open core piece of it, continuing to develop it, or do they freeze that in time and just make everything the premium? Correct. That's right.

Speaking of open source, though, I wanted to move over while you were away. We also saw the release of the next Kubernetes version. Yeah.

37, right? You want to talk about open source success stories, I don't know if it gets any better than Kubernetes. I agree.

I mean, it's funny because our head of engineering, Rick Spencer, just recently interviewed Craig, the founder of Kubernetes, on his podcast, and so one of the jokes of that was everybody mispronounces Kubernetes, and so I had a bet going with someone, and I hate to say I'm right, but it's Kubernetes. You say it correctly, so you're fine. But no.

Well, I didn't... What? I was going to say, but Kubernetes has been driving innovation, and the thing I always say is AI workloads are containerized workloads.

Kubernetes is driving the AI workload capabilities. So I think Kubernetes has really become so core to the platform and what we do in the infrastructure space. So now you've got me curious.

Maybe I just live in rarefied air. How do they mispronounce Kubernetes? What do they say?

Kubernetes. Do they? I've heard people say it.

I thought only my Alexa says Kubernetes, but not real people. But we know that AI's always right, so if you hear it from Alexa, then maybe you think it's correct. Do you think Alexa's really AI?

Alexa was more a... What did we call it back then? She has not kept up with the times, and I don't like her new voice either.

No. No, we don't either. We didn't do it in our house.

I told you I renamed my Alexa, didn't I? You had an option to rename, and so my Alexa is now called Ziggy, and it's a real sexy man's voice. Am I allowed to say that?

I'm probably not allowed to say that. Yes, you mentioned that. No, look- I know.

I love Ziggy. My husband does not like Ziggy, for the record. I can understand that.

Right. But going back- And I'm not casting... I'm just going to say I can understand that.

Yeah. But let's go back to the thing, to what you were saying. But we've been back to Kubernetes, though.

So let's talk Rancher. Yeah. Right?

How does this affect Rancher? So Rancher, so it's interesting. So we support two main Kubernetes distributions.

So K3s, which you could almost think of it as a lightweight Kubernetes distribution, ideal for Edge and kind of air-gapped environments, et cetera. And then RKE2, which is more of a standard kind of on-prem or public cloud or whatever it is, Kubernetes distribution. And then Rancher is really what I would call more the management layer, right?

So it allows you to manage Kubernetes multi-cluster, Kubernetes hybrid cloud, Kubernetes so it's like Kubernetes anywhere across multiple distributions. So Rancher and Kubernetes are joined at the hip. It's part of our overall platform, but it's a little bit different unique capability that allows you to really leverage the power of Kubernetes no matter where it's deployed or what the workload is.

Margaret, we've got to wrap up here, but I want to try to bring it together. Number one, I think what both of us are saying is we've seen this movie before. Yeah.

We've seen how open source Can infiltrate a movement, can infiltrate a technology, and change the nature. I'm probably a little more optimistic that it's going to happen sooner than later. Yeah, I am an optimist like you.

I am just a little bit more pragmatic that I'm not seeing the progress as fast as I would like, or I'm seeing too many things that are kind of trying to stop it at the same time. So that leads me to my kind of final take question to you, which is, based upon your views here, based upon what we've talked about, enterprise teams out here. Mm-hmm.

What should they be looking at this quarter? For many of us, we're all starting a new quarter October 1, right? For a lot of us, it's the final quarter of the year.

For some of us, it's a new fiscal year. What should we be looking at in terms of this race to open versus close, go slow, test? What should we be doing?

I'd love to hear your thoughts on this, so maybe I pass it back to you, but I have two thoughts. One is, right now, we can't buy our way out of this, meaning that we can't just spend and openly and as much as we can. We just don't have the budgets because I will tell you every enterprise, and I'm sure government IT department, is under huge cost pressures right now, right?

However, that doesn't give you any excuse or reason not to innovate in AI. Like, you can't do one or the other. You've got to do both.

So, I will tell you, most enterprises are looking to make a bet on a model, on a company, and hope it's the right one, whether if you're already a Google Suite company, then Gemini and its entire-- It's adding new capability all over the place all the time, becomes something very easy to adapt to. Claude and Anthropic are starting to look at different ways to charge from an enterprise perspective. So I think in the very near term, we're going to see a change in the way all these companies are charging because some of them, like Google, are able to do it at a much more cost-effective, and it can't just be token, token, token.

It just can't, right? Yeah. And you're starting to see this token-based charging in applications where they've added an agent or embedded this capability, and then they're charging you by token.

That's not going to fly. Like, nobody's going to pay for it. Yeah.

And so I think the first thing is that- But with that- Yeah ... that is going to force people into the open camp. Correct.

Right? So again, it's like- So they're going to start building their own ... it's open source jujitsu.

Exactly. You got it. Right?

The more you push, the more it goes the other way. And then the more people go to open, what does that do? That puts cost pressure back on the companies that are closed.

And so I think we're going to see that part accelerating, and we already are, as you've noted, and then it becomes, well, what is their business model? Because the valuations of these companies are off the charts, and if the token cost is what drove their business model, and they have to change that because open source is allowing people to just grab it and say, "You know what? I'll just build my own model, my own agents," which has become so easy, right?

So that's my thought is that I think there was a short-term economic spike, just like with the early cloud. Companies are figuring it out. They're refusing to pay ridiculous prices.

They're figuring out ways to get around it and still innovate, and that's going to drive more open. I'm going to go back to cloud too for my final take. One of the things that I just missed, I didn't have it on my bingo card, is that companies would use more than one cloud provider.

Really? Like, I thought people would standardize on either I'm an AWS shop, I'm a Google shop, I'm a Microsoft shop, I'm an Oracle shop, whatever. IBM, whatever.

And then the idea that people would use Azure for this, Google for that, and AWS for over here, I didn't see that coming. I think we're seeing the same thing with open source- Yep ... models and agents.

You got it. And the key for these enterprise teams is to develop harnesses that allow you to plug in, whether it's an open-weight Chinese model, and when we talk open-weight, it is, I think, seven or eight of the top 10 open-weight models- Yep ... are from China.

That's correct. Whether it's that or some of the American Frontier models or Mistral, which is the kind of French EU model- Yep ... that a lot of people point to.

Yep. You need to be nimble enough to use the right model for the job. Well, now we're going to go back to open standards because ideally, whether it's an MCP gateway or interface, right, some way, we also need standards then for these all to talk to each other.

So which we don't have across the clouds easily, by the way. And now we're back to Kubernetes. We still haven't got it with clouds, exactly.

But it is something we need to get sooner than later in AI models. Agreed. Agreed.

Margaret, we agreed at the end. That's a good thing. We did.

Oh my gosh. All right. " In the meantime, rest up, my friend.

You do the same. Good to see you. I will see you, well, it's probably still a month and out, but- Keep it coming ...

June, Salt Lake City. Look forward to it. November 9th, I think.

Yeah, and then AWS re:Invent. The roadshow begins. Right after that.

And we will be there for sure. Looking forward to that too. Sounds great.

All righty. tv. YouTube, Techstrong TV YouTube channel, Techstrong TV OTT channel, which is Android, iOS, Amazon, Roku, Apple TV.

Whatever screen you like to watch this stuff on, you can watch it. Until next time, have a great day, everyone.