Techstrong Gang – July 1, 2024

Hello everybody, and welcome to the latest edition of the Textron Gang. Today we're gonna be talking about the economics of knowledge and information and the age of ai. And then we're gonna jump over and have a little, um, chat with, uh, Jon Swartz about some research that the Vitor Group has done related to, well, just what is going on with AI in terms of people's buying behaviors, and what are we seeing these days.

And then finally, we're gonna take a look at Datadog, which is kind of making a move from its observability perch into all kinds of new areas. We'll be back in a minute. All right, everybody, welcome back.

I'm coming to you today from Dobb Sperry, New York, and then we have the rest of the gang. For today, it's Mitch Ashley, who's in his usual Colorado location, correct? Absolutely.

In Colorado. It's nice not to travel this week, you know, but who knows what'll happen next week? All right.

One out of 52. There you go. You gotta start somewhere.

All right. Um, also joining us is Jon Swartz, who's coming to us from San Francisco. So we got the other coast covered.

It is beautiful outside. The sun is out, the birds are chirping. All right.

And then once again, joining us from her very desert kind of climate, it's Tracy Ragan. Tracy, how are you? I'm doing great.

Thanks for having me today. All right, let's jump into our first topic, which I'm gonna go and Mitch pretty quickly on. But, um, Mustafa Slyman gave an interview where he was talking about not only are machines gonna have essentially personalities, but the entire economics of the way we think about knowledge and information is gonna change.

Mitch, I know this left outta you, but what, what's your take on this? 'cause, you know, as somebody who creates content, I'm not entirely sure how free I want to give everything away. Well, this actually, you know, it through, uh, through some, uh, deep thought research methodologies found this.

No, actually was watching a morning news show, and Andrew Sorkin popped up with his conversation, um, while they're up in, I believe Aspen at a, at a conference there talking to Mustafa, this AI CEO at Microsoft. And you know, here you're talking to somebody who is, you know, fully entrenched into the AI mindset and, you know, thinking about AI as emulating human kind of behavior. So he, it, Andrew asked him first a little bit about, so how, you know, how do we, how about dissolution thing, hallucination thing?

Why can't two plus two always be four? Essentially, it's what he was asking. And, and Mustafa gave kind of a researcher answer, which is, you know, human reasoning isn't that clear either.

We don't do that. We actually, we, we learn things by behavioral observations, and that's how we believe facts operate. That's why we believe two post two four in my, in my interpretation.

But the, I think the more fascinating part was he made some pretty controversial statements. I think he said that the economics of information, you know, we're about information and knowledge creation as humans, uh, that the economics of information are about to radically change because we are gonna produce, we are going to reduce the cost of production of knowledge to zero mar mar marginal cost, meaning AI is producing it. We don't need people to produce knowledge.

Uh, and in 15 years, you know, that's when he's thinking that 10 to 15 years, that new knowledge will be created with ai, essentially without humans. That's the way I interpret when he says zero marginal costs. It doesn't take us to help the help AI copilot to create, uh, knowledge.

And so, to your point about, I don't know if I wanna turn that over to ai, the work that I do, so it's, well, what does that mean for society? And he didn't really say, but if that was really true, if it was universally true, I mean, you're talking about humongous disruption to what people's jobs are. You know, a lot of this pie in the sky kind of babble at times where you refer to these new engines to turbocharge discovery and invention.

It was all kind of being brotherish to me, at least a little bit. Um, maybe he was elusive. You know, the one thing I find interesting about this guy is, um, he's gonna work with Sam Altman in some capacity, right?

Yeah. And before Sam Altman, he was the Altman of ai. Um, he has spent most of his time competing with OpenAI, which is interesting now that he is on the same team.

I mean, he was deep mind, and then he left Google under kind of bad circumstances that he started, uh, inflection ai. And then the only reason, I mean, the main reason he went to Microsoft is because Satya Nadella, the CEO, tried to lure Altman, and we know about that whole drama. Mm-Hmm.

And then he ended up hiring Soloman in the process. So there's this, this interesting dynamic that's going on, and, um, it'll be, it'll be, it'll be fascinating to see how it plays out. I scratch my ear.

We're, we're, we're in the hype cycle, guys. We really are. This is the hype cycle of ai.

Um, and if you're not, if you're the, you know, chief officer for AI for Microsoft, you gotta be out there talking about it like this. But it's, you know, it, we, every time we go through a new technology shift, and this is a big one, we have a period of time where there's a lot of hype around it, a lot of investment, a lot of kind of just silliness to be quite honest. And while I agree with a lot that he says, um, in terms of content will become knowledge, and then knowledge will become free, um, there will always be, uh, hopefully the, the process of questioning what that knowledge delivers.

Now, I'm hoping that we become more intelligent humans, uh, and learn to question authority, like what we see coming across our chats, because that is where correcting the knowledge as we move forward will happen. Uh, these are not, you know, these large language models are not static, and many of them are not good at this point in time. And to be quite honest, right now in the us we can't even buy a car.

That cyber hack on the car dealerships brought down the dealerships across the, the United States, probably in other places, and they had a cybersecurity, uh, function to it. So maybe we need to start thinking about securing our software before we start thinking about ai. So I feel like we have the cart, uh, before the horse in some way.

And this is a lot of hype, and it's taking, um, our eyes up fall just Opening. AI was kind of founded with this premise of doing AI safety research. So we won't have a one day evil AI that destroys mankind.

So, Libby, it's interesting just to see, to see where this leads. And also, you're, you're right, Tracy. I mean, well, I had in the next segment, we're gonna probably, I'll, I'll bring up this confusion and hype, which is leading to this piecemeal approach to the way enterprises are building their AI solutions or rushing to build them.

There's a lot of trial and error. All right, guys, wait a minute. Last time I counted, there were 50 plus lawsuits talking about who owns what content and who should be paid for what.

And this seems to be one being filed every other day. Now, it's not a foregone conclusion that the data and knowledge is gonna be for free. And there's tons of companies that collect data, create data, including Frito, that aren't going to just give away data for free.

I mean, you may have one time access to it, but you're not gonna get perennial access to it. Um, I feel like, you know, this pie in the sky little conversation that they're trying to have has, uh, not any basis in copyright law or reality. And frankly, if they wanted to take all the content that we constantly publish, and we'd put it in front of a fire, uh, firewall right now, and then they're gonna say that, that's fair use.

I promise you all that content's going behind a firewall. Mm-Hmm. Yeah.

I think also the other thing they're forgetting is capitalism. It, it sounded like a description of, uh, star Trek next generation. There's, there's no money, there's no, you know, there, there's, I don't even know if they go to the bathroom on the Star Trek enterprise, right?

In the future. They don't seem to, but, um, he was talking about a world where knowledge is free and accessible to everybody. That that's a utopia.

He almost, I just, I was just about to say that He sounds like a cheap utopian officer, let alone the ca IO, right? Yeah. So, and, and he is talking like a futurist, not like, um, an evangelist, if I wanna Put it down.

There's a huge, there is a huge goal for disconnect there, right? I mean, I wa my takeaway from that interview was, okay, in theory, this all sounds really interesting. It sounds like something Isaac Asimov, I'm, I'm stating myself would, would write about, but practically in the real world, and as Mike referenced with all these lawsuits over content, how does that work?

And, you know, they, they're probably trying to figure it out themselves. And meanwhile, series mean, meanwhile. Yeah.

And meanwhile, to make the thing actually work, we have to expose, um, we have to customize those things with our own data that kind of show that the model or let it continue to learn. Um, nobody's gonna check the box that says Yes, by all means. I'm gonna let you take the data I just exposed to your LLM and shove it in your LLM to train your future LLMs.

Everybody's gonna go, Nope. Sorry. Thank you for the use of your L lms, but I'll keep my data.

Thank you. So it's all about who created it, right? Who's managing it, who owns it, what is, you know, so it's the, the knowledge is, is definitely gonna be siloed if that is the, the future.

And we're creating knowledge and not content that we can, you know, get information from, but actually opinions, which is more what I get would see as knowledge, not just decipher some content for me and, and produce it, but actually make an opinion or discover something. For heaven's sakes, I don't see that happening. And if it does, it's gonna be siloed because that, that knowledge is gonna have a cost that's, that it's gonna have value.

Yeah. Hey, I heard interesting analogy that, uh, information or data is, uh, for example, tomato is a fruit knowledge is knowing that you don't put tomatoes in a fruit salad. And that, you know, it's least the sort of like, you know, the, okay, let's put it in context and how do to use it.

To your point, Tracy, YI Just think as we look at all this stuff, unless Microsoft and Google and all these people are gonna start paying people to produce content, the map just doesn't exist for anybody. I mean, who's gonna go out and create content and do the data, or be a researcher in some space, and who's gonna pay these people? They gotta eat, they got families they gotta live.

It's, you know, we're not gonna give them all universal basic income and then tell 'em to go be researchers. That's why it's described as, as Star Trek. Like, nobody needs a salary.

We all have everything we need. Right? It's too silly, John, you're out there in the valley and I at the risk of denigrating the valley, but it seems like a lot of this kind of pie in the sky AI stuff is emanating from a particular region of this country.

Um, why is that? Well, that's what has History is repeating itself. Maybe I'll put it that way, right?

It started, I mean, every cycle, there's a lot of hype. The money comes from Europe predominantly. So we have a lot of VCs, and I'm not gonna point my finger at them, but we have a lot of idea leaders who throw these concepts and ideas out to, to gain as much interest and hype as Tracy alluded to, and create interest in something.

And then, we'll, we'll be, is a classic example to me in a certain way. Let's, let's just, we'll, we'll figure out on the way, you know, we, things will break. We'll, we'll eventually get there.

So what I think what we saw from Soliman was basically the beginning of the hype cycle. He's trying to conceptualize and put it in a framework. I'm sure that what he said, both probably intrigued and terrified people who watched it.

I mean, it's the CMBC audience, so they're gonna be more, they're gonna embrace it. But for the general public, this plays into the fear. And the, I irony for all of you is out here in the Valley, tech is incredibly unpopular as an industry.

Uh, it, it's, we can go, we can do a separate episode on that or segment on that. But there are so many things that come out of this idea factory that we, that only a few profit from. And in a sense, we're at the, at the mercy of their decisions.

And a lot of the things you're talking about, the content, content of absorption creation, it's gonna be completely disrupted based on lawsuits and based on the fact that we don't have any clear understanding of where this is headed. So, in a sense, in a roundabout way, what I'm saying is this area is guilty as much as any for the hype intech. And it also has fostered this distrust, especially among the people who live here.

So we have the vast majority of people who don't believe or trust tech and don't understand it, versus the, the very few who are profiting from it, who are telling us, just trust us. That's an irony. You know, there's another, uh, interesting, uh, uh, Andrew Sorkin at the end of that segment, um, he said something else unrelated to Microsoft.

He said that Apple yes, announced that they're delaying their plans to release AI features right? In Europe Market because they're concerned about the digital markets act. And, uh, that you use the attempt to crack down on anti-competitive behavior.

At least, you know, that's how that could be interpreted. So yes, there's another haves and haves, not your, your regulatory environment could, uh, actually prevent you from getting your hands on some really great technology. There was a, there was a web summit, Mitch, I think it was about, uh, two years ago.

And the Apple had a keynote speaker there, and his entire speech was based on taking on the DMA or direct marketing Act. Mm-Hmm. Which, which went into effect this year.

And believe me, apple is terrified of that. And the fact that Soar can drop that, that was very interesting that you mentioned that, because when he dropped that, I was like, oh my God, they really do have an impact on what they're gonna do in Europe, despite what Apple says. Mm-Hmm.

They've been very virulent about saying, we've done nothing wrong. They almost sound like a presidential candidate. We've done nothing wrong, but we'll abide by the rules.

Can I just call out one other thing about this interview that is completely bonky? So he's saying that the machine is going to be as kind of trustworthy as a human right. And you're gonna judge its validity as you would a human.

Well, okay, fine. I, when I deal with humans, I deal with John, I deal with you guys, and you know, I come to trust you because I know you and I can identify you how the Fri, am I gonna identify an LLM that I trust versus one I don't trust when it's all behind some sort of user interface that's just swapping stuff in and out, and I won't know which one of these things is reliable or is it reliable? So that part of the conversation about, unless they're gonna start naming individual LLMs and telling me what their personalities are, that's never gonna happen either.

Well, that's what I was meaning it's gonna be siloed data, right? We're all gonna choose the LLM that fits our, our narrative, the best Mm-Hmm. Right.

And, and That we, that we trust. So those Of you who watched the debate last week, we know how that works out. You know, I, I think it's, I think it's packaging the three Ps for marketing, but, um, you know, we all, We all live in bubbles.

We all live in bubbles and silos as it is anyway. So what if you said just stand a reason, sorry to, sorry. No, no, no problem.

I'm, I'm just thinking, you know, if it's packaged as an avatar that is, looks favorable, behaves favorably. If you, if you believe his behavioral learning, behavioral observations is how we really understand the, the world. You know, it's not just a, a command line to type in a natural language query.

It's could be a physical robot, could be an avatar, it could be other ways, you know, we could package things in that are in a less threatening or maybe more trustful, um, modes. You know, what if your AI companion was a dog? I love dogs, right?

And I build relationships with dogs, and I can trust them somewhat. You know, I know they're gonna get in trouble, but not too much. So I that, I think that's part of it too.

I have to admit, I would love to have a robot, like on the lo the, the last lost of the space, A danger wear. I want That robot, The old, the old lost in space, or the new one, the New one I the new the robot from the new one. Jay was pretty cool.

Wouldn't that be nice to have a robot like that to pick up heavy bags? And you can, you could have, uh, the robotic dog, like a boy and his dog with, uh, Harlan Ellison story with Don Johnson. That's a long time ago.

Mm-Hmm. All right. I think, I think we're gonna have to bring this chat to a close, but I would just remind everybody that if you're like me, I have friends.

I love them all, but I don't trust them all, that's for sure. You know better. All right, folks, we'll be back in a minute.

All right, folks. And we're back. And we're now gonna talk about some exclusive research that Tuum has done, looking into what exactly are people buying as it relates to ai, because, well, it seems like they're all over the place.

And John has an article up on Techstrong AI that you can check out for more of the details. But j let's get started with you. What surprised you in this report?

Uh, You know, actually what surprised me was the level or the degree of how many of the enterprise buyers and influencers of AI products intended to add or change new vendors this year, within a few months of their last purchases is a roughly half. So, uh, we have, we're gonna start doing this. This is our first foray into data-driven stories that, uh, we're gonna go into the dashboard of Futurum Group and talk to their analysts and get, and get an idea of where things are going and who's winning, who's losing, who's not gaining any traction.

What's interesting here is that I think 43%, again, as I said, of the vendors, were not entirely happy with what they're doing because first of all, they are in a rush to get up to speed again, to meet the hype. Uh, they don't have a number of options. They are relying on the old legacies that they already have in place, and they're not doing a great job of training their employees.

And to quote Keith Kirkpatrick, who is the analyst at Futurum who provided this information, he said, anyone who says they know what they're doing with AI is lying. And that was, I mean, he was blunt, but he is also, it's also true. I mean, I talked with the Axios executive who's the chief operating Officer of Axios hq, and he said that every six months, it seems that the board changes.

He says, it seems as if AI is everywhere in nowhere. How do you package such a superpower? So again, there's a lot of movement, a lot of moving parts.

It's a kinda a piecemeal approach they're trying to get eventually to where they want to be, but they don't actually know what that is. And the things that kind of put them at unease especially are governance, the costs and then the skills, skills training, um, but especially the governance. So what we have is a work in progress to the, to cut it very close and blunt.

Um, and we have a very small number of vendors who are making hay at this point. And then AI application for, for instance, is dominated by Microsoft in the us AWS in Germany, France, and Canada, and IBM in the uk. We also have folks like Accenture and Deloitte on, on the consultant side, who are doing extremely well at this point.

Um, they're just though a handful. And again, uh, Microsoft, Google, A-W-S-I-B-M, surprisingly to me a little bit, uh, they are the leaders, but there is the potential for some breakout to happen. It's just a question of when I, I actually ended up talking to three or four other execs at New Relic, um, as well as Informatica, and they basically told me the same story.

So we'll see. I mean, um, it's a, it's a mad dash. I mean, I don't know how else to put it.

It's a mad dash and people are doing trial and error. You know, John, it's also, it's also a consultant's dream. Yes.

Right? Yes, yes. Nobody knows what to do, which, which racehorse to pick, or should I pick one?

Should I kind of stay with the vendors I have and follow where they're going? Do I go pick this shiny new racehorse that's come out? So especially enterprises tend to rely on the large, you know, consulting firms, um, development firms that work with them to help them figure this out, which is not a bad strategy if they're working with multiple companies.

And sometimes they, they're developing their own centers of expertise, right. Whereas you may not be able to do that if you're an enterprise. You might be able to create a center of expertise around ai.

You know, medium s size isn't, you're gonna experiment and learn some things, but how do you know you're learning enough to be able to use this effectively and not sell yourself down the river? So, as I said, I think this is a goldmine for, for consultants right now to basically learn on the job with customers. Yeah.

As the, as every, which is sort of the definition of consulting and, uh, uh, not a snarky definition, but Well, The, you know, the premium that the decision makers are putting was always first and foremost on expertise and experience with ai. Pricing and contract terms were secondary. Mm-Hmm.

Um, and the number one deal maker, or deal breaker, excuse me, for I, it buyers was a lack of implementation speed and, and timeline. So yeah, it was playing in di the Accenture and Deloitte, BCG, atos, and CGI kept coming up as trusted sources according to rum. Yep.

And they're partnering, partnering with, you know, cloud vendors, Deloitte, I mean, they're partnering with everybody, but like, um, you know, Deloitte is partnering with, uh, AWS and the telco space, maybe others. So it, it takes kind of the technology supplier, if you will, as well as that knowledge of how to implement this that people need help with And why to implement it. Yeah.

Why, why are we, we don't ever, uh, we're not asking the question why, why are we implementing ai? What, what are we doing? Are we trying to provide better services to our customers?

Are we trying to be part of the hype cycle? Are we trying to be, as Alan always talks about in the top three and, and customers implementing ai, I don't know if they're always asking why Is there a compelling problem that we are trying to solve? Or do we just wanna be able to put AI in our URL?

Um, I, so, you know, is AI a product and we've talked about this before, or is it a feature? Personally, I think it is a feature, not necessarily a product. And we keep thinking about AI as a product.

Now, I have to admit that when I first heard about chat GBTI was like, what the heck is that? And now I don't go a day without touching it. Uh, so there is a behavioral change that AI is going to require from all of us, and maybe that is what they're chasing.

But in terms of the, uh, the enterprise companies, they need to ask why are they implementing ai? Is it just because it's the new cool, shiny object? Mm-Hmm.

And if they're doing it for the wrong reason, and they're gonna be unhappy, And you nailed it, because one of the things that the large consulting companies is Accentures, Deloitte's, they bring the industry knowledge, the, the domain knowledge of pharmaceuticals or banking or telecommunications. That's what, you know, a Google or an Azure or, uh, AWS don't have strengths in. Right?

And, and that's to answer that question. So where would would we apply this? What's this good for?

Kinda reminds me a little bit of blockchain. And I'm, I've never been a blockchain fan, but it sort of rumbled around for every ever forever looking for a, a second, you know, uh, killer app besides, uh, besides digital currency, I, you know, think, I don't think AI's gonna suffer from that, but that's part of what people are getting together to say, let's look at the domain, the work that happens in these industries and where potentially can AI help and where can we, where do we start first to really kind of get our legs under us? And why Exactly.

Not every app needs NFTs, Right? I think part of the issue though is we expect, or maybe we marketed that AI is deterministic, right? And it's gonna give us answers that we can trust and we put too much faith in it.

'cause I don't think most people realize the degree to which it's probabilistic, which is it's guessing. And if I put that in a workflow that's gotta be right a hundred percent of the time, then I'm probably gonna be unhappy at some point. So they, a question then becomes, um, due to people who are operationalizing AI and the latest flavors of gen ai, understand when to use what type of model and for what use cases and, you know, or are we over promising and gonna massively underdeliver as far as they're concerned?

You know, Tracy, it's really interesting what you said. Uh, one of the guys I talked to is basically blatantly honest, said the two questions I have about our whole process or our whole plan is how and why, why are we doing it? How do we do it?

Well? So Mitch, as you kinda look at all this stuff, you know, what's your best advice for folks about how to kind of go down this path? 'cause I think there's a tendency right now where, uh, the board is saying, what are we doing about ai?

And somebody stands up and says, oh yeah, well, you've got these proof of concepts, but then nothing seems to move forward. 'cause nobody thought through the actual use case. Well, I, I think ai, AI is a contact support.

Meaning you've gotta, you've gotta kinetically learn, um, how to use AI and what it's good for. 'cause it's, it's not like a, a new programming language, right? Okay.

Rust is better because, you know, it's, um, it has type safety and, and things like that. Memory, safety, uh, it, it's, it's a different paradigm. I mean, I had this, even when I started doing AI work in the eighties doing a logging list.

It's a very different way of thinking about how to solve a problem. And you have to start applying it to understand how an LLM works and how the vector database makes that faster. And how, how you can protect your data with rag.

How does it do that? You wanna understand it in order to be able to apply it. So my recommendation is start experimenting to learn more about it.

Um, yes, you may go to other people to have them do this for you, implement it for you, you know, God bless 'em, have 'em help you. But you'll be able to understand better what it takes and kind of sift out the, do these people know what they're talking about and sell, or these people kind of selling me, you know, smoke, what's going on? So that, that's my advice is, is form some experimentation projects in your organizations with the people who would naturally gravitate to wanting to learn something like this.

John, is the trough of disillusionment right behind us there coming up any second now, or are we already there? Oh, no, they don't. Well, they're, they're confused.

They're, um, angst ridden, but they really don't have a choice they have to forge on because as you said, there's a lot of pressure from the top down to formulate a planning, even if there isn't one. Uh, I had some company, I wish I could name the company, but I can't. But I was told them by them that they literally are like, consider, they consider themselves an AI leader, and they've been getting coverage to that effect.

They said, this is a bunch of b******t, we don't even know internally where we're going. You know, it's just a, it is just a, it's just like a, it's a, a land grab. I mean, just, just see where, let's see what happens.

Let's, let's, let's just break some stuff and try to fix it and move on to the next crisis. That's, oh, I can tell you, doesn't it? Yeah.

I can tell you, as a startup, if you don't have AI in your plan, you're not even gonna be listened to, which is sort of ridiculous, to be honest. That's like saying, you know, you have to have a certain kind of a database in your, in your architecture in order for us, or you have to use blockchain. And there was a period of time that we did that.

So hopefully this hype cycle will start to slow down and we'll start to get smart about AI itself and how to apply it and when to apply it and where it makes sense. At what point do you think that the tech sector's credibility will be tarnished to the point where maybe business executives will stop listening to it, where maybe they already have for all I know, but it seems like, you know, there's a lot of coverage of the tech sector in Wall Street and all these other financial sector publications, mainly about the price of the stock and valuations and such. But do you think the average business executive is starting now?

Look at all this and saying, you know, here these guys come again. If she's awo, if they're a woman, Well, well, We, if there, you know, more wo more, if there are, were more women in tech, we might have a more balanced approach to this Might have. We're not, we're not big gadget people, right?

Women are not necessarily the gadget people, guys like their gadgets and AI's the new one. So let's get more women in tech. Yeah, Yeah.

I'll go with lack of diversity of ideas. That is absolutely true. Has always been the case.

You know, the problem too, with most of these companies, they're convinced that unless they describe themselves as a tech company, they're gonna be left in the dust. And what builds builds into this fear is when we have the annual, uh, statement. John Chambers will always do this annual statement about how half the Fortune 500 companies will no longer exist in four or five years unless they go to the digital transformation, which is like this very spooky kind of general, general arguments.

And people buy into this and they continue to because they think they'll, they'll, they won't lose their job. But eventually somebody's gonna probably say, you know what? We tried to adopt these practices with ai, especially we tried to do this, it didn't work.

We'll do plan B instead, somebody may, there's gonna be a story eventually about some company that's gonna take the contrarian view, because there always will be, there'll always be a maverick or an outlier. It's the pendulum swinging back, right? Mm-Hmm.

It is definitely customers who drive business and customers don't necessarily buy technology. Investors buy technology that are delivering services to customers, paying customers, hopefully. So, you know, capitalism isn't driven by the new technology.

And sometimes I get frustrated with the folks in the San Francisco area because they're so technology focused. And right now it's all about AI for, and oftentimes for no good reason. Right.

But Joe, Yeah, I, well, I think the reason why there's an obsession here especially is because we went through this down patch of all sorts of downsizing after too much hiring. So now they're, they're grasping or looking for something to, to, to kind of change the narrative. And it has changed the narrative.

But I also think this is more about addressing markets and about addressing fear and uncertainty among, among the industry here locally. You are starting to see though, some noise coming outta Wall Street where they are saying, where's the revenue for all this stuff? And when it doesn't quite show up, it's not in the balance sheet, it's not in the books yet.

Um, are we starting to identify people who are quote unquote AI washing and there's gonna be some punishment coming from shareholders? Yeah. And that, that's, uh, oh, sorry.

Just quick aside is like, if you look during, during the earning season, what, and this is gonna start at the end of July, the next cycle, there will be just a handful of companies even that who are gonna break out AI related revenue. AWS may be one of them, but aside from that, good luck finding it. It wouldn't be the first time that, um, several startups get acquired who have little or no revenue, but have, you know, crazy valuations.

So, you know, there's different economic factors that fall into that. It's not always, it's always logical about like, well, how profitable, you know, in fi tough financial times it is, but it isn't when there's a gold rush. Do you mean Now the one thing, one, one thing I do wanna say about this topic is that while I, you know, I, I, I don't believe that we are asking the question why we're implementing ai.

I do believe that tooling that we can deliver toolkits, uh, products that can help companies use AI are going to be what we need to do first. Um, we need more tooling and we need make, we need to make it easier to use AI in development so that we can use it as part of our toolkit. And when I look at ai, when I look at right now, from my perspective, I'm not trying to look at solutions that are already built.

I wanna see tools that I can use to build solutions. And I feel like that, that we need, we need more help and assistance and funding in that area To that extent. Uh, I also think, Tracy, that we need frameworks.

In other words, we don't. We're gonna be exchanging out LLMs and this technology and that technology right over a period of time. So the more you can either construct your own framework or use, you know, frameworks up high Torch or something, um, like that, either from the provider or, you know, in, in languages gives you some, some separation between the technology you're using.

'cause if it's gonna change out, you don't wanna go all write it in a custom language that works with that one provider's technology. And then six months from now we do a, you know, a student body shift. Right?

We're now we're heading this way. Right. All right, well, I think we're coming to the end of this conversation, but I really like that.

I think we need more tooling and we need more frameworks and hopefully less hallucination from both the machines and the people. All right. We'll be back in a minute.

I'm Bonnie Schneider, sustainability contributor to the Techstrong Group. I'm excited to introduce you to a groundbreaking new initiative from Techstrong Research, the sustainability pulse meter. The pulse meter offers valuable insights into how environmental responsibility factors into tech purchasing decisions for key players in the industry.

Position your company as a leader in the industry and differentiate from your competitors with the sustainability pulse meter offered exclusively from Techstrong research. All right, folks, and we're back with our final segment and we're talking about Datadog, which, uh, last week had its big dash conference here in New York. And it was really interesting to attend because Datadog is evolving into something completely beyond just being a monitoring company.

Or then they became an observability company. But now they're talking about including incident management capabilities. They can look for vulnerabilities, they can now interact with LLMs to tell you whether or not they're toxic.

And it just seems to me that they're evolving kind of, you know, another type of platform engineering company. And I wonder, Mitch, your thoughts here, but it seems like everybody who's coming from different vectors to try to become the platform, or at least one of the last standing platforms in this space of what we were gonna call platform engineering. Are we, is everything consolidating, or is it just my imagination?

Uh, I think, I think you're spot on and you use the, the magic word today platform. Um, so that's definitely part of it. I think it's a recognition of, you know, think about a monitoring tool.

You know, we were used to just by a monitoring tool, and then we did whatever outside of that monitoring tool. And now we need that information to follow the workflow, to be part of that process. And whether you're now managing LLMs and monitoring those with, with Datadog, or you're using them as part of their technology, um, and, and understand what's happening there, it, it's trying to put it together, all this information together and make it understandable.

But even more importantly, it's gotta fit into some kind of a workflow. And oftentimes it's multiple disciplines across different teams, security, operations, development, whatever it might be. Uh, the one I kind of scratched my head on was that they added a Kubernetes outta scaling capability.

Um, which is like, okay, um, I need to look at that a little bit more. I didn't realize that they were doing some things there. So I think it's trying to become a more, a more well-rounded platform to their customers and not have to be sort of the old, like, you fit into this silo of a tool and you fit into this silo of the tool.

And that's what, that's what everybody's working on doing. Mm-Hmm. I asked him about the Kubernetes thing just for your edification, and I said, you know, last time I checked, Kubernetes is supposed to scale up and down on its own, sort of, I mean, you just have that capability.

And they kind of smiled and said, yeah, but nobody knows how to make it work. So here we are. Uh, okay, this is a working version.

Did you ask them anything about what they say about DevSecOps, workflows, streamlining DevSecOps workflows? I, you know, I didn't, I really couldn't find a whole lot about that, but that's an area I find very interesting. So here's their position on it.

They believe that DevOps, ITSM SecOps is all gonna converge into one kinda Uber workflow. And you can get there from any path you want using Datadog as your kind of foundational console. But they're betting that eventually that all these processes finally become unified and all these silos start to become, shall we say, uh, less isolated.

So Mitch, I don't know what your thoughts are, but, um, can we get to this glorified unified IBM vision of 20, 30 years ago for making all everything manageable from one console? Well, if I check the, uh, the star log, star date, 27 42 is about when we'll achieve that. What Star Trek references have there been in this show?

It's, it's a Star Trek day. I'm sorry. Okay.

Um, you know, you know what, what I, what the way I think it's evolving, the way I see it involving, 'cause you can see it, uh, people are using these tools and the vendors is, there is no uniform way people do this. There's, for example, DevOps, right? Is DevOps part of the platform engineering team now?

Is it a separate thing? Um, is all if security part of the, the platforms part of the platform engineering, i, it SM part of that as well. People configure these, uh, these functions, these roles, um, to what, to what fits best in their organization?

Maybe they have a strengthen area. Let's add this to that. Um, maybe we have silos and we can't break 'em down.

So is is it all going to one unified theory? You know, kind of the beginning of the big bang? I doubt that.

I think more you flip it the other way is Datadog could work in flexibly in any way with how the customer has configured those operations in their organization. And the unified part of it, Mike, is recognizing that that's, that platform approach is, I don't wanna have to talk to my ITSM tool. Maybe that's Jira, right?

And, and instead now they're doing ticketing and this and, and uh, Datadog. And that kind of alleviates the need for that. Well, what happens if you have Jira already, you're gonna want to integrate this into, so it's never that clean.

Even a greenfield, you may start out that way, but it, it tends not to last. Alright. Tracy, are we all in arms and sing Kumbaya here?

What do you Think? Well, we're trying to do that as a play hub. We wanna gather as much of the DevSecOps workflow as we can because that's where the, that's where the story's being told.

And, you know, in a cloud native environment, um, and I just did this presentation for, um, cloud, uh, I think it's cloud native now on July 25th. Uh, everything's so fragmented in, in, uh, cloud architecture. I think we're gonna have to look at tools like Datadog or Deploy Hub or whoever is pulling together this information so we can see the full story.

And when then we talk about really applying ai, then we have data that we actually could, can, that's being created, not being made up, um, that we can actually do something with, like maybe auto responding to cybersecurity threats. So yeah, I believe that we have to have a kumbaya moment. It, the, the data is, there's data in too many places and in particular around security and, and DevOps.

We've got, those have to be, um, unified. You can't have those separate. Um, I have a question around your story.

Um, it was interesting concept. I think it's, I think it was referenced by maybe one of the executives. So with the rise of AI tools to generate code, do you think observability is increasingly gonna become a requirement?

I know that was kind of touched on by someone there. I I think so, but I'll defer to mention, uh, okay. Uh, yeah with the, you know, the more APIs we write, the more observability we need.

Okay? I mean, seriously, we, uh, everything is so fragmented. The only way we can really understand how these APIs are talking to each other is almost through observability, which is not good 'cause it means we're missing something in our development process.

Um, but, but yeah, the more that we do, we break down and we decouple these monoliths into, into smaller components and APIs, the more we need observability to understand what's happening. Is one particular API being, you know, slammed? Is there a threat because somebody's trying to get into it?

We gotta be able to see that stuff. And, uh, it can be done. Observability is the only way you could do it.

You have to have one of those tools in today's world. You know, to add to that, Tracy is there's using observability tools, right? In your software.

Um, there's also designing your software so it emits better information, more usable information by observability tools. And I think that's where we'll see more of an emphasis, right? So you're, you're not trying to play detective combing through that data or have it do that for you and, and find out I don't have that information.

I don't know what was actually causing that, you know, that spike in performance or whatever it might be. So I think, I think observability and development means observability in software that we're creating designed for operating in an observability environment. I dunno, Tracy, do you agree with that?

You think that makes sense? Well, A hundred, a hundred percent. A hundred percent.

We recently did a discussion on API first and really understanding what APIs you're building, uh, to minimize the sprawl of APIs will be helpful in that, in solving that problem. Here's the paradox. When I talk to people and, you know, I'm a simple guy, so I kind of go down to the brass tacks a lot.

Um, I was explaining observability to them and they were like, 'cause they understood monitoring and a set of predefined metrics and you kind of follow those along and you kind of hope that things stay up. But when I was explaining observability to 'em, they were like, well, I was like, well you can collect all these logs and metrics and traces and then you can launch queries to pinpoint issues before they become a problem, or you can, you know, at least troubleshoot your thing better. And they were like, well, that all sounds great.

And then they thought about it for about five seconds and they said, I have no idea what questions to ask. And so the question I have then is, you know, if I don't have the expertise to go use the tool, what's the point of the tool? And some folks are now telling me that, well, actually the machine learning algorithms are gonna ask the questions going forward.

And so now I'm using AI to manage the ai. I mean, how is this gonna evolve? Yeah, that's the whole point of threat models, right?

We have to have better, we have to have better threat models and we have to automate threat models and we have to improve threat models. And the threat models have to ask the question. And right now we're building the threat models and maybe we don't have the knowledge to do that.

So yes, this would be a a, an amazing application of AI Def It definitely would. And I think if, if for folks that have used an observability tool kind of visually, if you think about it as here's where there's an issue, right? What part of the system, whatever that is at a macro level or a microservice or a Kubernetes cluster, whatever it is, and it, it's connecting the dots of other things that are happening around it.

So it isn't always writing queries. You could write an actual language query and say, here I show me this, I'm looking for this or find where this is. But a lot of it is visual to say, okay, well this is where, this is where we're pinpointing the start of the issue, but it's the tentacles, it's the branches of, well then these other things are happening.

This API is really slow from this third party service that's causing this to back up. And you're getting that all from the telemetry information from those different systems and pulling it together in one way. You can, you can kind of work through and understand how it looks, which sets you up very well for machine learning algorithms.

Um, but I think there's a lot of work in that area because you have to do a lot of preparation of data to be able to used by be used by machine learning algorithms. They talk about features within machine and it's not like features of software, it's, it's prepared data and algorithms, machine learning algorithms. So it, I think it's a great application for it, but it's not a, you know, turn turnkey option.

Well, here's the thing I hear from all the observability vendors and they all say the same thing. They're getting beat up by customers who are like, it's nice that you gave me a dashboard to tell or shows me that I have a problem, but I really want you to fix the problem. And just standing around telling me that there's a problem isn't quite as helpful as fixing it.

And I think that's part of this move to platform engineering is the customers are driving it and saying, you know, you just can't be a, a tool for surfacing up pretty pictures of my mess. I even saw that in Stack State when they were announced the, their acquisition by Susa at Suan. Um, they were very clear to point out, and here's the things that you do to correct this.

Now it's not automated, but it's kind of the runbook of how to solve a particular kind of problem, which is today's technology. Uh, could that be automated with ai? Sure.

But they were, they were, to your point, Mike, they're uh, clearly they're getting asked that by their customers. 'cause even in their demo in a short period, short time, they had to demo it. They made sure and spend time on that part of it, Which is why to our earlier question, it, the DevOps data is needed, right?

Because the DevOps data tells us how to redeploy it and how to rebuild it to auto remediate it. So we have to have all of that data together And that error may occurred, occurred in, you know, not just in code, but in the process, right. Of buildings.

Yeah. So, All right folks, I think we've come to the end of the session here, but it seems like all roads are leading the platform engineering. The question is, which road is gonna get there first in my mind, is it the observability people come and execute, stop?

Or will the people who already execute stuff add observability and we'll see what happens and goes from there? Or maybe only 1, 2, 3 platforms. Five.

At what point does that kind of defeat the purpose of platform engineering in the first place? Anyway, we enjoyed the chat as always. Tracy, thanks for being with us.

Absolutely, John, Mitch, as always, good to see you guys to see You. Absolutely. Good to be here.

Stay tuned everybody. We've got some great content behind us here and the rest of the text on gang TV show, or actually it's the Textron TV show, which the Textron gang is a part of. And we once again, thank you all for spending time with us.

com is the number one online destination for DevOps education and community building. com covers all aspects of DevOps, including DevOps, best practices and tools, DevOps culture, DevSecOps, business impact, continuous testing, continuous delivery and more. com has the largest collection of original DevOps content, featuring breaking news, blog posts, podcasts, and more.

com to learn more. com where the world meets DevOps.