Techstrong Gang – March 13, 2025

Hey, everybody. You're watching Techstrong Gang. We got open AI versus Microsoft IDPs, and well, what's gonna make those trains run outta time?

We'll be back in the middle. Hey folks, welcome back to the latest edition of the Textron Gang. We got a new member of the gang we're gonna introduce today, Robert Reeves, who's been in the DevOps space for a long time in the cloud native space.

Robert, welcome to the show. Well, thank you for having me. All right.

And for those who don't know you, give us a little bit of your background there so people know who you, where you are and where you're coming from and what your biases are. Oh, absolutely. Uh, I, I bias have no way to them bias.

Yeah, absolutely. Uh, you know, we hold these truths to be self-evident, that software wants to be free, that developers want to solve hard problems quickly and not deal with human toil. Uh, my entire career has been in service to developers and making certain that, um, they have what they need to live their best lives.

Uh, also I feel very strongly, um, that, uh, about businesses making choices that don't box them into a corner, uh, whether it's with closed proprietary solutions or open solutions that aren't quite as open as we think, uh, think that that drives developer productivity and also outstanding ROI for the businesses that, uh, hire and support those developers. All right. So Mel Gibson's gonna play Robert in the movie.

That's how that's gonna work. Right? Wow.

Hopefully that, that, uh, that, that, that shoulder dislocation scene, uh, in Uhh lethal Weapon Weapon. That's weapon. Yeah.

Yeah, yeah. That's, that's the one. I like that, that was pretty hard There.

There you go. Also, joining us today is Guy Courier, chief Analyst for, um, the visible impact arm of the Tuum Group. I always find that as a mouthful to say, but Guy, how you doing?

Good to see you again. Uh, I'm doing great. Um, good to be here again on this 88 degree or whatever it is, day in, uh, the, in March in Austin.

Uh, Robert, I know you know what I'm talking about, and it is a mouthful. I apologize to everyone. All right.

And speaking of mouthfuls, we have, you know, his Royal Highness, the Czar, apparently, of Silicon Valley, John Schwartz. How you doing, buddy? Wow.

Hey, how's it going everybody? Um, it is, uh, raining and, uh, it's raining, uh, water from the sky, but it's also raining open AI news. My gosh.

They are busy. They're doing a lot of stuff we're gonna talk about. Yeah, Well, let's jump into that.

9 billion deal for AI infrastructure. Uh, left me scratching my head immediately. 'cause I said, well, doesn't Microsoft have a bunch of infrastructure they could use?

You're closer to this, John, what's going on? Well, I think, you know, what's interesting about that deal, by the way, is that as part of it, um, core Weave is giving OpenAI $350 million stake in the company, and the company Core Weave is gonna go public with a valuation up to $35 billion. So it's a pretty nice deal for OpenAI across the board.

But you're right, it, it kind of points to this fissure between open AI and Microsoft, which I guess was inevitable. I mean, Microsoft works with companies and then they work against them, they compete against them. And there've been a series of moves in the last week, but in particular that kind of show this divergence between the two companies.

You know, Microsoft, they have invested $13 billion in open AI and knowing a significant chunk of the company, which Microsoft will not disclose, but I've heard as high as 35, 40%. But in any event, this, this a this, this deal with Coral Reef was just the latest in a series of moves and counter moves. Um, as you recall last week, the information reported that Microsoft is testing AI models to rival those of open ai.

And, um, that could lead to selling to developers and increased competition between the two companies. Tuesday Open AI released new tools to help developers and enterprises build AI agents. So they're, they're fighting back.

I think it's really interesting, and I'll, I'll throw this out to the group, but what, what's going on between Microsoft and Open AI kind of reminded me of their relationship between Microsoft and Apple back in the day, and how Microsoft, remember when, when Steve Jobs came back, Microsoft made this investment in Apple, which turned out to be an incredibly shrewd move based on what happened with Apple Stock after jobs took over and saved the company. And I think in the same way they're doing this with, with OpenAI, they're, they're investing in the company, they're competing against it, but they're also gonna benefit as OpenAI gets larger and maybe has open ai, uh, benefits from the core weave deal. So in a sense, again, it's history repeating itself.

Microsoft is your friend until it becomes your foe. I know. Robert, what's your take on this?

'cause, you know, to John's point about Apple, you know, I'm thinking maybe IBM and OS two, so who knows what that Relates to? Yeah, exactly. Exactly.

Mike, it applies. It goes back, it goes back over and over again. And I'm glad you mentioned that because I was thinking the same thing.

Yeah, it, it's, look, um, AI is, uh, a tool, uh, and we've seen this before, whether it is cloud DevOps, you know, even going back to nascent web, um, there is going to be a lot of, uh, coopetition. Um, but the thing that really jumps out at me with this announcement of this monster deal is that I feel for the CFOs of, you know, uh, uh, fortune 500 global 2000 mid-market companies, because their product teams, their executive leadership, um, are now even more reasons to, uh, to, to adopt AI and to start consuming chips. Uh, it's a lot easier to use that in a cloud like core we provides.

And so I am just really concerned about these runaway AI cloud bills. Um, I'm concerned because I don't see it being tied to the bottom line, uh, for the company. I see a lot of resume driven development.

I see a lot of science projects and, you know, things like this are gonna continue to build this, uh, echo chamber of people, you know, driving people to adopt this technology. And I'm not sure if companies are ready to tie that investment to a return. I'm not sure if they're really, you know, it, it's, it, they're confusing light with heat, and I'm just seeing a lot of heat right now, and I feel for that CFO I'm feeling for 'em.

Wow. That's, that's lovely. Uh, science projects more like science fair projects, if you ask me.

Or maybe, maybe sandcastles, uh, 'cause you know, they're built and they're destroyed. And I, I've been saying lately that, uh, the easiest budget allocation you could get is, uh, the one that requests some kind of AI something or other. Um, not that that means that it's, uh, uh, not gonna be watched and tracked, but, you know, if it doesn't work, I, I don't know, it's like a SpaceX explosion.

It's just like, okay, well, I'll just try another one. Um, I think you're, I think you're, you're spot on, Robert. Um, I, I think that that, well, two, two things.

One is by God, there's so many, there's so much going on all over the place. It's, it's, it's hard to like, keep track of, um, they talk about storming, norming, forming, whatever the, the, those things are, we're, we're, we're in extended storming phase, because like I said, the the money's flowing freely. There's a lot of places people can do all kinds of things.

Um, yeah. John, what was, what, what was the, uh, what was the investment, uh, uh, that, or what was the spending Microsoft had in, uh, core Weave? Um, uh, it was in, in the hundreds of millions of dollars or something like that.

Wasn't that Yeah, Mike, that's pocket change for, for, for, for Microsoft. Uh, there's no reason, uh, they wouldn't just go and spend a whole lot on other infrastructure out of Azure where they have sort of ready made. Who knows?

That's, that might be a sandbox environment for all who know, you know, I think that for our, our audience, you know, keeping, keeping steady and understanding, to Robert's point, what value are you kind trying to drive? Don't just approve and go, um, don't just think that it's, it, you know, AI is this easy button. It, it's not, it requires, uh, whatever you do, it requires curation.

Maybe tuning, maybe some development you're learning, but also in its implementation, whether it's a foundational model or a tuned one or your own little service, um, you, you want human supervision on the whole thing. And all of this stuff that's moving around just points to the fact that there's no one who's found the natural monopoly yet. There's no one who's figured out how to lock in their pet AI model, lock in customers to their pet AI model.

And I think that is the underlying effort here is, is that, uh, whether it's Microsoft OpenAI or anybody else, they wanna become the monopoly and they can't figure out how to do it yet. Yeah. So there's a little bit of paradox happening here, though, because it turns out that the price of AI tokens and GPUs is coming down because of competition.

They're just shrinking their margins and NVIDIA's still getting their money for the GPUs. But the, I've talked to a bunch of people this week who are all saying, you know, prices are starting to drop regardless of the fact that GPUs are hard to find and the pricing from NVIDIA's high, but the rest of the services are dropping. And so they were kind of doing a happy dance, but I looked at it a little bit differently, John, um, are we not buying market share here, essentially?

And maybe people will be underwater, and then the next thing you wake up one morning and discover that, you know, they're broke. Yeah. Yeah.

I know. It's interesting too, the other dynamic, and I I, I don't want to throw this out, but this whole idea, we talk about this AI spending rampant spending, and I wonder if that's gonna continue a pace, given some doubts about the economy. I was gonna throw that out because I think that could throw a wrench into a lot of things.

You know, we've been talking about money as if it's, if it's error, you know, trillions of dollars in these, these budgets or these projects that are being thrown out, bandied about. And the, and the one thing I also wanted to mention though, about open AI and Microsoft, which is kind of interesting, is this whole crazy notion of the $20,000 a month service from an open ai, the PhD level intelligence. And when I was thinking about that, I was also thinking how that relates to Microsoft.

And in a weird way, what open AI is trying to do, I think, and I, this is the only explanation I can get from folks they're trying to commoditize and kind of show some sort of independence and show that they can actually bring in revenue because they are burning through cash. I think they burned through $5 billion just last year, according to New York Times, and they're still trying to raise money. So it's, it's, it, it, it, there's too much, it's overwhelming.

And I think Robert alluded to it for a CFO or a CIO or, or it decision maker, what do you do in terms of all these options? Did you mix and match? I read some where they're up to a dozen, uh, AI agents from different companies would be used within, within enterprises.

The whole thing is kind of head spinning, you know, and I, and I think we there, we kind of have to step back and figure out where all this is gonna land. And I Just have no idea. Well, we, you know, the aphorism about the gold rush in, um, California, you know, the people that made the money were not the gold miners.

It were the people selling shovels. Uh, so if I was a finops company right now, I would be very excited, uh, if I could help Target, uh, not only, um, assigning, oh, you know, not just optimizing, uh, cloud performance to lower the spend, but to say, okay, this department spent X, this department spent y and being able to do effective chargebacks, I would want to make sure if I was a CFO, I'D partner with a finops company right now that could help me point fingers. Because I would like to know if some group has a 30 million cloud and AI bill and then see that they're generating 5 million in revenue.

That's what I would like to know. So, guy, I have another question for you. Um, in effect core, we becomes the infrastructure house organ for OpenAI.

So won't everybody else who provides infrastructure services go rush to every other LLM provider and create some sort of sweetheart deal and say, you know, it's us against them. Uh, you mean like competitors to Core Weave going and trying to do the same thing? The problem is the core weave, uh, started out as a GPU, um, based instance cloud company.

And that was in, what, 2017 or 2016? Something like that? It 17, yeah, 17.

It was originally, this was originally to mine, Bitcoin, or, you know, it was for crypto. Um, but lucky them, they could just, uh, adapt the same architectures, wasn't luck, of course, but, uh, they, they could adapt the same architectures, uh, to, um, to, uh, AI training and, uh, to a lesser degree AI inference. Um, so they have pretty deep roots here.

And, you know, I'm sort of hard pressed to think who else does that might be, uh, an open partner other than Oracle, which is already partnering with, um, with, uh, is that open AI or, or Microsoft Open Ai. Uh, yeah, you know, that's Interesting. I'm already doing that.

They're opening that data, right? So Oracle's been investing. Yeah.

Yeah. I thinking of us considerably for like two years in these same kind of architectures. We know, you know, Microsoft has for Azure, but that's for their own services.

Google the same thing. So Mike, honestly, um, uh, especially given that I, I think John's point about the macroeconomic climate is a really Big one. There, there are at least two startups out there, maybe three with GPU services specific to compete with these guys.

And every cloud service provider, whether it's somebody like si o or, um, some of the, what do you call them, um, dedicated Racks, dedicated hosters, Dedicated hosters, are all sticking GPUs in places. So I think, you know, there's plenty of places to go partner with, but, uh, we'll see what the level of competition is from here. John, what is your prediction here as you look at all this stuff?

So, you know, one thing I failed to mention is that NVIDIA's in a big investor in Coral Weave, which adds another interesting dynamic, but I also, I wonder if Coral Weave becomes part of this Stargate project that we just, we mentioned earlier, this ridiculously overinflated, over hyped $500 billion project over four years to build out AI infrastructure. I wonder if they become a key element of this within the open ai, Oracle, SoftBank triumvirate. I think, um, we're gonna continue to see, and, and it happens literally every day.

It's probably happening right now, even more moves, uh, down the line between open AI and Microsoft. It's, it's, it's inev inevitable. And again, we have to look back at history.

This is will always happen with Microsoft, but the thing about Microsoft that's, that always impresses me is they, they place their bets and they're pretty, they're pretty shrewd on, and, and they're benefiting in a sense from their partnerships and, and competing with some, with some of these same companies. Um, they kind of play the whole table. So I think Microsoft comes out as a kind of a big winner as they always do.

And they're, they, I mean, they are the most admired company, I think, in the Valley right now, in terms of the way they're run. Alright, I'm gonna open up that AI sports book in Las Vegas and take some bets and we'll just like, get some odds going and we'll make some serious money. Well, Actually, can I, can I make a prediction and get John's reaction to it?

Sure. Go for it. Yes.

Um, I think OpenAI is the Yahoo of ai. Wow. I think that's, yeah, I think, Damn, dude.

Yeah. That's a good, that's a good one actually. I, I, you know what?

I see that They made the market. They made the market. They're making the market Well, to some degree still.

And, you know, I don't know how long it'll take, but they will. Didn't Microsoft try to buy? Yeah.

Didn't Microsoft try to buy Yahoo? Remember that back in the day? Oh, yeah, That's right.

Yeah. And it could happen again with, yeah. All right.

Well, Robert, since everybody's throwing around predictions, close this out. Well, it, it's, um, I think, um, that whoever is leading right now is not gonna be leading in the future. I know that, uh, you know, you can be first or you can be best.

Um, and it's just the nature of business. Uh, people are gonna see where, um, other startup founders and other organizations are gonna see where there are missteps and they're gonna come in. I do not think we're at the top of the hype cycle, though.

We have a long way to go until the trough of disillusionment. Um, so my prediction, uh, is that there's gonna be a lot of broken hearts, uh, for consumers, uh, people that are in this space that are providing these technologies, whether it's cloud or tooling around ai, they're gonna do well. Um, it's the people that are gonna be, uh, trying to consume it.

I, I, my heart goes out to the, you know, banks and insurance companies and manufacturers that are trying to figure this out, because they're just gonna be dumping money into a dumpster fire. All right. Robert's got ai, dark horse to win.

All right, we'll be back in the morning. Discover Techron Group, the epicenter of tech innovation. We are your go-to for reaching IT, leaders and practitioners worldwide.

Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients.

Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Techron Group. Hey folks, we're back in.

I know we've been talking about platform engineering on this show multiple times, but every time I go in and have a chat with somebody about this, they point to this thing called an internal developer platform, an IDP. And to be honest, we've seen these things before. We kind of like maybe experimented with it, I wanna say, five or six years ago, and then basically forgot about 'em, and then suddenly they made a comeback under the guise of platform engineering.

But Robert, I know you follow this space closer than I do, but what's your takeaway? What's driving this IDP conversation? Is it just like the simplest, easiest thing to do for a platform engineering team and the best place to get started?

Or is there something else afoot here? Well, um, I think it's just a natural reaction to how we've evolved as an industry, um, with software engineering. Um, so whether it is, uh, adopting DevOps Cloud, um, using AI and to support development efforts, um, we are continuing to get more and more efficiencies out of these individual areas.

IDP platform engineering, what we're trying to do is offer a catalog, a single place inside that company to speed developers on their way. We wanna make it easy for 'em. We wanna remove human toil from this process.

We wanna, uh, get rid of the stuff that is slowing them down. But at the same time, we wanna also remove the chance for unintentional mischief where we can, uh, maybe fire up environments, uh, for development. And perhaps they're not the right ones.

Uh, perhaps, uh, developers will, uh, start working on the wrong environment, and we wanna be able to direct them to the right place. It's just one example. Um, it's just a natural extension of, of what we saw with DevOps and, and I'm here for it.

Um, I certainly love to see, you know, open source projects like Backstage, uh, you know, uh, uh, remember when that first came out? Very excited to see that. And I'm happy to see more and more companies adopting it.

It's all about making developer, the developer experience, uh, easier, uh, more enjoyable. But at the end of the day, it just makes good business sense. Every line of code that you write that is not for your application that lowers costs or increases revenue, I believe is wasted code.

I think that we need to make sure our developers are spending time writing and improving these applications, building and improving these applications, and not on internal plumbing. Uh, and, and the things that stop them from doing what they wanna do and what made them wanna be developers in the first place. Well, let me follow up on that.

'cause let me tell, I kind of understand what you're saying, and mathematically I would agree, but then my soul kind of goes like this and says, well, you know, the first reason we embraced DevOps was to get out from underneath the CIO man in the first place. And now we're moving back to some centralized IT thing that kind of feels like the CIO man is coming back. So how do we do this in a way that Proverbs or preserves, sorry, your freedom, you know, that thing you talked about as being very important.

Mm-hmm. Well, I would hope that our DevOps teams, um, our SRE teams, um, would have a key part to play here. I would hope that this would be driven by internal dev development teams, uh, throughout the lifecycle, that they would be driving this.

I would hope that the CIO would say, we're gonna use an IDP. I hope that's not the case. I hope that it is driven bottoms up and that developers are saying, Hey, let's make this easy to onboard our new developers.

Let's take care of some of the hassle and heartache that I have to deal with. Uh, I've certainly seen anecdotally where these teams, uh, in internal development teams, not executive leadership, um, saying, we're going to bring in backstage, we're gonna make this easier. All these little places that we go for CICD, provisioning all points in between.

Let's go ahead and put that in one place. Let's have our catalog of instances that we can run, whether containers or cloud instances, and let's just make it easier. Uh, so I would hope that it would be driven bottoms up, because when you do that in technology, you're going to have success.

You're gonna win mindshare, uh, the, the, uh, you know, the whole cathedral and bizarre thing. Uh, and, and I would very much hope that it would be driven bottoms up or from individuals that once you improve their lives and the lives of their fellow developers. Mike, isn't this the whole problem though, because I, I think, Robert, you're, you're, you're right.

But I think that requires the platform engineering team to start working a lot more like a product team. And I don't think that's in their, their, you know, their blood. I, I think that, I mean, engineering almost says it all.

Um, in, in the phrase platform engineering, what I'm talking about is, um, that, uh, one of the great outcomes of DevOps was a, a slight reorganization of how a software engineering teams worked so that they started to work like product teams. They had inbound, um, you know, what, what are the needs and requirements? They monitored the user experience.

Um, they, you know, tried, you know, uh, uh, types of development, but reacted to usage and all that other sort of stuff. So a lot closer to the user base considered as a customer. But gosh, like that would be a big change, wouldn't it?

In, uh, so that's why I asked Mike, Mike, like, the platform engineering teams don't work this way. If they started to organize themselves in this way, then this bottom up thing would work because their constituency becomes the developers. But I don't know, I've talked to a lot of ops folks over the years, and they tend to think like, no, we need, we need this engine, this machine that's just humming.

And they almost see the developers and users as like, like pesky, like bothersome things that they have to deal with. And that's not the kind of culture you're talking about, Robert. So Mike, what, what's your perspective?

You've been covering this for a while. I think you're spot on in that they need to come to a mindset where they're treating the developers as their customers and not their hostages. So yes, that has to happen.

But, um, you know, to your point though, what happens is, like somebody sits around and says, we need to scale our investment in compute and storage. And then they just turn everything into this kinda one size fits all approach, and everybody rebels. And the next thing you know, developers who have skills will just go out and do it all over again somewhere else in their basement using a server that they are, have on their own, build their own software, and then they'll bring it to work with them and hand it off to somebody, and we'll be as far off as we ever were.

John, I know you've been talking to CIOs for a long time. Are there kinder, gentler CIOs out there? 'cause you know, when I talk to those people, sometimes I hear phrases like, I hear a lot of I, and not a lot of we, Right, right, right.

It's a top 10. Yeah. Yeah.

I think we're talking to the same people, or it's an echo chamber because it's a, basically, it's a top down mentality, right? It's what they're infatuated with, who they are friends with, who, who they believe in based on their recent biases or their historical biases. And it's not about what the, the vast majority probably are, are pursuing or want see, so they kind of foisted their decisions onto others, which then creates this bottom up problem, which we continually see.

And then they, they continually make adjustments based off of, uh, the whims of a CIO. So you're right, Mike. I mean, it's, that's, that's always kind of be inevitably part of the equation and, and a major flaw, flaw in the system, Right?

So Robert, is there some way to kind of bring these ops folks and CIOs together with the developer community? They kind of have a more enlightened conversation, shall we say? Well, it, it's really, uh, the, with 'em, what's in it for me?

If you are able to identify motivation, uh, for all these parties, what do they hope to accomplish? You can build a coalition of the willing and reach those goals. So for example, are there OKRs, KPIs that our friends over in CIO land are driving toward?

Um, certainly development has theirs, and how can we come together and reach them? Um, you know, it really comes down to, uh, increasing efficiency, lowering costs, increasing time to market for a new product. Everybody can agree on that.

And, um, you know, I really believe that they need to sit down and talk, uh, and they need to figure this out. Uh, it, it is top down, just doesn't work anymore. Uh, you will have, um, you know, developers will vote with their feet and they will leave.

And, uh, you know, some folks might say, well, that's great. We can hire, uh, less experienced developers. Um, good luck.

Uh, you know, every time, uh, you know, when the, uh, you know, you shut down work for the day, you know, at my startups, um, I always, you know, I, I I believe that the most valuable asset that I had left the office every single day or left the computer, whatever. Um, and so, um, maintaining those folks is gonna be key. I think platform engineering, IDP is one of the ways of doing that.

But you have to do that with them. If you're building a tool with them, uh, for them, uh, you have to partner with them, right? I love your Karen, but I'm going to describe the stick.

The stick is gonna be the rise of ai. And right now, when you look at all these projects involving ai, I gotta go get developers and data scientists and data engineers and security people, and all kinds of other folks that are hanging around this project, and it takes a frigging village. It's fundamentally inefficient.

So I would say that platform engineering is gonna be required to drive the next wave of AI applications because it's beyond simply a bunch of developers. We need a more, uh, centralized approach. We just need to do it in a way that, you know, is open to innovation, right?

So I don't get locked into a tool, and I don't have a CIO telling me, well, we're an Oracle shop, so we should only use Oracle stuff, or Amazon, or whatever it is, because, you know, then that defeats the purpose. So, guy, can I have my cake and eat it too? No, I don't know.

This is, this is like the oldest story ever, and it's still going on. The ops people hate the devs, the devs people hate the ops people. I mean, you know, hate, it's not really hate, but, um, I, is this yet another cultural shift also?

I, I just, I, I don't mean to be pessimistic. I really don't. I think that there is an opportunity with this platform engineering paradigm to, to sort of change the process and make it more like a product development process.

And that would be a, a cultural shift, but not actually as huge a one. Um, but you know, most of what I've seen around this is, is the usual story of get user-based buy-in and to get management support and work from the top and work from the bottom and come to an agreement and that sort of thing. And those are always worthwhile effort.

That's a perpetual, worthwhile effort. But I think a more fundamental change would have to, would have to take place. And maybe it's not my idea of, um, running platform engineering more on a product basis.

Maybe it's something else. Um, but I, I, I, I, I don't think, um, I don't think that doing the same thing as before is going to do more than keep the cycle going. There will be some kumbaya, there will be some successes.

There will also be failures to Robert's point developers leaving difficulties, um, that that's been going on for decades. And I would love to see that change. Um, is AI the stick to do it?

It could be. It could be. Um, but I think we are aware of the flaws in difficulties with the use of AI to begin with.

And so that'll have its failures. All right, I'm gonna end this conversation here, but point out one thing, what we're currently doing isn't working so well, so maybe we need a different approach. com is the leading resource for news analysis and education on challenges facing the cybersecurity industry.

com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more. com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more.

com. Home of security bloggers network. All right, folks, we're gonna talk about something a little bit different here, but it turns out that Google and the MTA runs the trains in New York City have been strapping phones to the train to try to figure out, well, when the train might be the delayed.

I'm long time New Yorker. I kind of scratched my head a little bit and said, this is an interesting way of going about solving a problem. But Guy, I know you two have been in New York when you looked at this, what was your impression?

Uh, well, since I've been, uh, the sad elf for this entire show, uh, let me, uh, change my mood here and say that I love this. This is awesome, and it's awesome in two ways. Um, the first awesomeness is this is a fantastic use of generative AI and of AI in general.

It's a fantastic use. Why? Because it replicates a, a, a, something that maybe the most elite of track inspectors can do.

It doesn't replicate it perfectly. I think what is 89% success rate or something like that? Um, and it does it, you know, uh, using, um, AI to collect, you know, reams and reams of data at the point of, you know, at the edge, um, and, uh, uh, do these kinds of analyses that, uh, the human brain and human ear can do with years and years of training and experience.

So when you get this 89% success rate, that's, that's a terrific accelerator. That's great. Love that.

The other way is, um, it, it, you know, this is edge ai and it's edge computing, and it's really interesting that they didn't use edge devices. They just use smartphones. Everybody needs to remember a smartphone is an edge device.

Why they didn't use, uh, uh, dedicated edge devices is a really interesting question, John, maybe you have some insight into that, um, because that would be a typical approach. But instead, Google just kind of went in and said, we just use smartphones. It's easy off the shelf.

We'll install an app, bang, bang, boom. You know, it's done really quickly as a proof of concept that maybe could turn into edge devices. I mean, I love the idea of these R 46, uh, subway cars, which are 50 years old now, something like that.

I mean, I rode them to, to school when I went to the Bronx High School of Science. Um, I'm really familiar with them. They're, you know, they're not quite buckets of bolts at this point.

They were always pretty good, but we're not talking about some new fangled. This, there's, this is a really great example of agile and quick deployment of an AI application at the edge for an immediate proof of concept, positive outcome, and a reason to invest further. So, yay.

I think one of the reasons they didn't use sensors is because, well, if you've been down in the subways of New York City, somebody probably said, you want me to install what? Where have you seen the rats lately? So, Well, yeah.

So it's not a question of sensors because the phones have sensors. They used the microphone to, to, to listen to the sounds of the subways going over the tracks. Mm-hmm.

Um, did they use anything else? I mean, that might have been all they were able to use. Maybe the camera, maybe, you know, you strap it to the front and they use the camera.

What is sort of interesting is how they, they, they picked up sounds and other data, and then they fed it into Google Cloud when they were trying to analyze and spot patterns that would indicate track defects before they became an issue. So that was mainly the, the delays had to do with maintenance. And I think it's interesting too, because Google's been working with, um, other transportation agencies over the years.

It's worked with the Chicago Transit Authority. Um, it's worked on, um, with, with a chat box for them. It's worked in, in terms of connecting street parking meters to Google Maps.

And, and, you know, a lot of cities, not just New York, are looking into this. Um, there have been efforts, I think, in, in Beijing for facial recognition systems, in place of transit tickets and cars. You reduce the lines during rush hours.

Um, I believe in Chicago, they were trying to sense or find, um, instances of gun, of gun use of all things. Um, and I think that was something that they, that was tried in New York too, was looked at at least, um, New Jersey Transit system uses AI to analyze customer flow and crowd management. I think it's just, it's just really interesting to me how AI is being used for something that's very practical.

And, and, and when I, I did not realize the scope and the sheer size of MTA, uh, was like 500 stations, you know, hundreds of millions of rides. I mean, it's, you're bound to have delays 24 7. And I think this is an actually a very interesting pilot project.

Now, whether it leads to an actual contract is another thing given budget constraints, but I think it's a positive move in terms of transportation. And God knows we need to look into, um, enhancing our transportation given the state of, of, of the FAA and other, other, other agencies. Right.

Robert, what's your take? I mean, can AI accomplish what even Mussolini couldn't pull off, make the trains run on time? Well, I, I, I think this is an argument, um, on how AI can support humans in what they do.

Uh, so it wasn't AI that said, Hey, let's put a phone. Let's, let's put all these pixels, uh, you know, on, on, on MTA lines that that was not, that was a human that did that. And of course, you gathered data and, and used AI to, to tell us, you know, and get some, uh, predictive ai and tell us are, are these running on time?

How can we improve? What is the, uh, expected delays, uh, are they gonna compound? Um, I love it when I see people find new uses for existing technology.

It is, um, you know, I I it winds up being a celebration of human spirit and of solving problems. And that's exactly what our engineers do, uh, make it work. Um, and, and dealing with the resource constraints that you have.

Let's figure this out. I'm sure there were probably cheaper ways of building, uh, these edge devices, guy that you, you were talking about, you could get that price down far less than a phone, but how long would it take to manufacture that? Well, exactly.

Get A box of The just cheap, it was speedy. It was off the shelf. Yes.

And so there's always gonna be that payoff between cost and, and speed. Um, but you know, I've also seen the same thing, um, with, um, uh, people building cars. Uh, large European car manufacturers are replacing, uh, $20,000 Windows PCs on the factory that drive, uh, torque wrenches on the factory floor with, uh, little tiny edge devices that are running a container.

They're driving that with Kubernetes. Um, we, they are figuring out how do we use older technology, um, and update it, take advantage of, you know, ease of use for deploying our apps, but also how do we tie this in to get the data to, um, our teams that are trying to build models with ai, either hopefully predictive ai, not generative ai, uh, telling us what's going to happen. Um, another example is a, uh, company that makes, uh, plastic little toys.

Uh, maybe you put them together and they have very old injection molding machines, and they're using open telemetry to get data, and they're using that to predict when those machines are going to go down and to get some preventive maintenance in. I love with old trains, old injection molding, uh, machines, torque wrenches. Uh, you know, we're, we're applying new technology to this to make it, uh, better.

Uh, and, and, and it just, it, it makes me happy. Uh, I like the idea of them sitting around, um, a room somewhere and say, Hey, what if we just use pixels? Oh, that's a great idea.

And, and that one team that did it saved them so much time and hassle and heartache and really made this quite an interesting story for the press to pick up win, win, win, Ga. You know, I've, I've always, I've always complained and bitched and moaned about how tech doesn't address like, real issues in, in real life. And here's an example of, of a practical use of something, a new technology or, or a blend of old and new technology to address a age old problem that actually helps people.

So, and, and again, it's kudos to, to Google and others who are, who are looking into these China practical solutions. Guy, I'll tell you this, my father worked on the trains for 45 years, was, uh, worked for Conrail and all those before that, going all the way back to Grand Central. Um, I'm a little worried from the engineers, is AI gonna start driving the train and do we not need brakemen and brake conductors?

And it's all just gonna go to tech. Yes, AI will start driving the trains. Um, one of the interesting, you just little sort of, uh, uh, example, um, I hadn't really thought about trains so much, but I've been thinking about self-driving cars, uh, for many years.

Um, self-driving vehicles. Um, I don't remember who pointed it out, I didn't come up with it myself, but one of the biggest opportunities is, uh, in, uh, medium haul trucking, um, intermediate trucking. So the most difficult driving, um, issues are like in cities, and that's where self-driving is being tested.

But in fact, the best opportunity probably for removing drivers altogether and using, uh, uh, AI driven vehicles is on medium halls, um, to intermediate, uh, uh, depots or, you know, what, what, uh, depots that the right depots, depots, intermediate depots on the outskirts of cities, at which point human drivers or human assisted driving at least, uh, there's a human in the cab, uh, in the same way. Um, uh, there's a, it is pretty, it's a pretty controlled environment on tracks, um, you know, by its nature. Um, so I do think that that is one of the areas of low hanging fruit for, uh, for all, you know, a hundred percent ai.

Mm-hmm. I don't know, Robert, will I go to a amusement park someday to drive a train just for fun and 'cause AI's gonna drive him in real life? What do you think?

Well, you know, I'm reminded of the scenes in Tim Burton's, uh, Charlie in the chocolate factory where, uh, you know, Charlie's dad loses his job at the toothpaste factory because he was putting caps on the toothpaste. Uh, and then of course, at the end of the movie, hopefully, I'm not ruining it for you. Um, he is, uh, because he, his job was taken by a machine that would put the cap on, and then later at the end of the movie, he gets a job repairing the machine.

Um, I've also seen this with, uh, the move from, uh, typist to word processors. Uh, we, we've certainly seen this, um, over and over again, um, there, yes, uh, a number of people that, you know, uh, uh, farriers, people that put, uh, shoe horses on shoes, you know, how they kinda lost out with the rise of the internal combustion engine. Um, but, uh, there is a whole host of other jobs that were created from this.

And yes, things will change. Uh, the ch accel change will be accelerated. Um, and there will be some difficult times ahead.

I would hope as a society that we would recognize the innate experience that those train conductors and brake conductors and all those people you mentioned, Mike, that they have, and hopefully use that experience to make it better. Uh, let's use AI as a force multiplier and to scale up. Uh, but I would hope that with ai, uh, as a consumer, I would get something better, not just the same, but cheaper for the corporation that would bum me out when you make It any nostalgic.

'cause my mom used to be a typist and used to type 90 words a minute on those old manual typewriters. So there you go. My grandmother was a statistical typist.

She worked in the Pan Am building, um, in midtown Manhattan. She was a statistical typist, and that one went out real fast. There you go.

That particular specialty. What, what is a statistical typist? What is that?

Uh, they, they type, they type spreadsheets on manual or electric typewriter. Okay. So all the characters on The periphery of the keyboard.

Well, yeah. And, and, and putting things in columns. I mean, it was a, it was a specialized form of typist that could format numbers in grids and put them in the right place and all that sort of thing on a page using, you know, typing.

All right. Well, I don't know guys, but it's 2025. I guess somebody can put together an amusement park from the turn of the century.

'cause everything we knew about two decades ago is already becoming obsolete. Hey guys, thanks for being on the show, Robert. Thanks for joining us.

It was a pleasure. Yes, always. Thank you.

So, All right. I want to thank you all for watching this latest episode. Stay tuned for what's going on in the rest of Techstrong tv.

Me, I think I have a bunch of Lionel trains up in the attic. I'm gonna go play with, let's see it.