Techstrong Gang – February 19, 2024

Hey, everyone. Welcome to Textron Gang. You know, this is gonna be our first episode where we truly have a gang, because not only are we gonna have our normal Textron team members here, and, and for those who don't know them, my far left.

It's Mike Ard. The lady in the middle is Bo Bonnie Schneider. Joining us from Colorado is Mitchell Ashley.

And, and of course, I'm Alan Shimmel, uh, of Techstrong. But, uh, later on in our show today, we're gonna have the one and only John Willis acclaimed author DevOps cloud expert, and we're gonna be talking a little bit about ai, which is kind of his new expertise as well. Um, so we're excited to have John as our first Textron gang member to join us here on the, on the broadcast, and, uh, grateful to have him available to us.

Before we get to John, though, we have a couple of other topics we were gonna bring up today. Um, we're gonna be talking about some recent news outta Cisco and around layoffs. And layoffs, of course, are not solely just Cisco in the tech industry today, unfortunately.

Um, we're also gonna be talking about some AI and, and some of the recent going, recent going, gone here at Techstrong. com. Mike, It was on, well, he has two, actually, one on digital C xo, and the other one is on Techstrong ai.

And one covers a new mathematical approach, and the other one is about well vector databases. So, yep. And then as we've been doing at the conclusion of Textron Gang, we are gonna be running on Textron tv, our full lineup of shows and interviews and other content that we hope you'll find interesting.

com, and we'd love to hear from you. Alright, so folks, welcome back. Thanks for being here.

I wanted to bring up some news. I actually caught this morning as I was getting dressed listening to, uh, news on Alexa. And that is that the, the story is Cisco might be laying off up to 5% more people.

And this comes on the heels of them laying off about 4,000 people last quarter. Um, 5%. By the way, I think Cisco has 80 something thousand employees, so 5% is probably another 4,000, which would bring the total number here between eight and 9,000.

Look, layoffs are nothing, are not news in tech today. Mm-Hmm. You know, I was on a recent, yeah, not at all.

Yeah. Mitchell and I were on a recent trip to Boston, and every company we met with had layoffs. They called it, they said it was layoff season.

I hope that there's an off season for layoffs that meets, but, um, what I, what I found interesting though is, uh, in reporting numbers and, you know, results, their business didn't necessarily shrink. Right. Uh, security grew, I think it was 4%.

Networking grew at seven to 10%, something like that. A he a healthy number. And, and the big, the big number came out of their observability group, which grew something like 21%.

And granted, that's probably by far the smallest group of the three I mentioned, but still, 21% growth is a healthy number. And I will add that that's not counting Splunk, right? Mm-Hmm.

Uh, so observability is primarily around the AppD business unit that they had acquired, of course, when they acquired AppD. Now with Splunk coming in, we've heard that Splunk's gonna be on the security side of the house ledger with Cisco. I don't know if that's just lip service because Splunk's, you know, a big player in the observability space.

This is true. Mitch, Michael, what do you guys think? I think, oh, go ahead.

I'm sorry. Go ahead, Mike. Jump in.

Yeah. I think a lot of the times, you know, this is a Wall Street driven conversation and the vendors will say things to Wall Street that they think will drive share. So they throw around the word security.

I think the Splunk code is gonna wind up in both the security side and in the observability side. 'cause you know, the techies that run Cisco are not stupid. They're not gonna go, oh, well, we can only use that in security.

Mm-Hmm. So I think that that will flow along. And in general, I don't know, Mitch, what do you think?

I, I think it's all about the deal, all about the acquisition and positioning it with the market. And so there isn't any, uh, kind of product overlap with other, like you, you mentioned down there, o other observability technology. It, you know, every observability is being used in everything today.

Right. It's operations, it's security, it's data, it's, you know, people are all looking at how to get access to this information. So, you know, no, no big predictions.

But I gotta imagine they'll take Splunk and have a, you know, an offering in different parts of the market, not just security. So they're not gonna turn away all their non-security business, that's for sure. Right.

I, yeah, I, it's much more than the code. Splunk has a lot of observability customers. Exactly.

And it's about absolutely converting them. And, and maybe you'll see a combined AppD Splunk kind of observability solution, and it'll be even, you know, make it a better solution. Do You think that might help point to future growth and getting off of layoff season?

I don't know. Um, I, I think to your point though, the layoffs are driven much more by Wall Street than anything else. They're getting beat up for becoming more profitable.

It's not just Cisco, but all these companies. And I think the whole process is somewhat counterproductive because they're clearly taking a blunt instrument to the process, and they're just basically saying, we're gonna cut these people without thinking through, well, how disruptive is all those cuts gonna be to the customer. The customer's gonna buy less because they're gonna be like, well, who's my person?

Who do I deal with? And then that will have a negative impact on revenues and margins. So I, I don't, I don't wanna say this is stupid, but it's pretty close to it.

Well, You know, I think also the fact is that these cuts have an inordinate heavier toll on sales and marketing than they do on engineering. Mm-Hmm. Um, look, I think the perception out there is though, is that tech was on a fat drunken buying spree for the last four or five years, hiring people way beyond their needs.

And that this is just kind of, you know, trimming fat and returning to reality. And, and then there's this other perception, and I think Zuckerberg met, reinforced it, which is, Hey, we could get 25% higher profits while having 25% less people. And you say that to one of the, you know, wall Street guys, they're like, you know, all of a sudden they go all Gordon Gecko on you.

Right? That's the New, that's The new benchmark all of A sudden. Yeah.

Greed is good. And, um, greed is good. Greed's what made America great.

And, and so that's what the, that is the new benchmark, Tiffany. No, Alan, it kind of goes back to what we were saying. You, you mentioned our trip to Boston and layoff season.

I I think there's some cover in doing it right now, too. You know, you do your layoffs while everybody else from Microsoft to meta to whoever that's, you know, I think there's 144 tech companies that have done, I read an article on CNBC. Uh, so, so now's the season to do it.

If you're gonna do it, you're, you're not gonna get as much attention. It'll be Right. The story will be gone by noon today.

Right. It'll be something else. I think too, though, a lot of these companies, to your earlier point, they roll up acquisitions, right?

And I keep adding new companies into the play, and I don't really, um, operationalize that very efficiently. So you do get a lot of overhead over time. And I think, so some of these companies like Cisco, there's probably some fluff in there because there's so many additional people who are selling one or two products when the buyers are looking for somebody who's gonna sell 'em, you know, the entire stack.

com days. There's a lot of that, right? The consolidation of multiple sales teams, not even talking about the consolidation of multiple platforms and technology tools.

But I look until, quite frankly, until interest rates start moving, because certainly the fact that the stock market is at all time highs is not slowing down the, the momentum of layoffs. So it's gonna take something else. Mm-Hmm.

And, but this is the, again, this is the cycles, you know? Mm-Hmm. It'll, it'll go.

We've bottomed out. I think we're on our way up. But an interesting thing I heard from a friend of mine yesterday used to write for us at Techstrong, and he's been a, uh, a marketing person in tech 30 years.

He's been outta work seven months, not able to find, uh, a position. And he's a, he's a good writer. He has great Microsoft connections and everything, and, um, he's ready to go work in Home Depot or something, right?

Because he just can't find a job. And, and that's, you know, we can talk out here all that we want and think we know what Wall Street wants and everything else, but that's where rubber meets the road. When your friends, when our fellow tech workers can't find work, look, a lot of 'em are saying, I'll be a consultant, I'll be a contractor, I'll, you know, 10 99.

And a lot of those people may wind up becoming entrepreneurs who start a business, but it's almost like the forest fire that you need to burn out before the new growth comes up. And I don't think the fire's quite finished yet. Still fuel out there to burn, right?

We'll see how this all plays out, but I would remind everybody, the general economy is doing pretty well, and there's a lot of folks out there still hiring DevOps and IT professionals, so, Absolutely. And security and cyber. Yeah.

Absolutely. Anyway, all right, let's move on from that. But before we do, I wanted to just, let's take a quick 32nd break here so we can gather our notes for the next segment.

And check out this great little, uh, announcement here from US at Techstrong. We'll be back in a minute or 30 seconds. 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. Hi everyone. Thanks for joining us on the Techron Gang.

We are now talking about ai. How is it impacting the workplace and what do IT workers think about this new impact? You know, it's interesting.

Pew Research, uh, just did a survey amongst IT workers and found that, um, good amount of them, 32% felt that this is going to help their partic productivity and innovation, and they'll do better with it. But there is still that apprehension about will this replace us? Are we gonna have any a situation where our, our, our own jobs are threatened by it?

But it really depends on the person. Another thing that's been coming up with AI is the amount of predictive analytics that they're able to get and get more insight into things. So that seems to be something positive that, uh, Watson from IBM, for example, is talking about that they're able to analyze data at such a large scale.

Um, and that's how they're, uh, saying that this is beyond human capability and they are very positive towards it. But I'm wondering what everybody and the gang thinks about this topic. I think everybody's job is gonna change, but I don't think the jobs go away.

I think a lot of the stuff that you hate about your job, AI will help to either eliminate or reduce the amount of time required to do it. And I think that part is good, but there's some natural apprehension because as roles change, organizations are gonna be structured differently. And so who is in charge of what will all be in flux for at least another year, perhaps a year and a half?

So that's where I think the anxiety comes from. com. We'll, we'll ask John Willis about this when he comes on the next segment.

You know, everyone said the continuous testing and automated testing is gonna make testing obsolete. And, you know, DevOps is gonna cost people jobs because as we go to cross-functional teams and become more efficient and automate more, we'll need less people. Before it was DevOps.

It was, it was agile. Before it was agile, it was something else. This is a common theme I've seen in 30 years in it.

And you know what, I've never seen the IT job number go down. So me, this is all just noise by 2 cents. I Think it's no, it's in, you can go back to cloud and that's gonna replace cloud too, rack and stack server job and robots and manufacturing and all that kind of thing.

You know, I think the other way to look at it is, I mean, all of us have seen these waves. It seems like there's an initial group of people who see it as an opportunity, right? Okay, this is gonna change things.

Well, I wanna figure it out. Let me, let me get on board. And you know, John, John Wilson is a great example of someone who early on, you know, saw this and started investing in his, his own knowledge and skills about ai.

Um, but I think in the end, kind of to Mike's point, it it's more of a rising tide. It's gonna, it's gonna lift all boats if you, if you decide to participate, if you hold back and fight it and just See, right. If you're gonna dig your Job going away, you know, that's probably what will happen, at least to you.

Um, yeah. But I think it's gonna change our jobs to Mike's point, but I think it's gonna make us all more productive and, and give us new opportunities as well. Yeah.

I would say the new opportunities is interesting because now you're seeing, um, it workers saying, well, uh, need someone to write prompts for chat GBT, that's a new skill set that we can bring to the table prompts engineering. Yeah. But, you know, I'm gonna tell you something.

This reminds me of an argument I had. My son was accepted to the Kelly School at Indiana University five, six years ago. I went with him out there to the open house, and a professor got up there and said that the minimum wage laws are artificial and hinder, uh, jobs because people won't hire someone at the higher wage.

They'll just automate instead, or use, you know, automation in place of people. And I think, you know, and he was trying to hold himself out as a capitalist. 'cause there's America, god darn it, in Bloomington.

And, and the fact of the matter is, a real American says, if there's an automation that lets me cut ahead, I'll cut ahead. That's what capitalism, and that's what the market's about. So if AI is going to, if your job is so darn easy to do that, AI can come and do it.

Find another job. We're not here babysitting. Let that, 'cause then we'll talk about layoffs next year, right?

Because we just have people who are doing jobs that really shouldn't be there, right? I want, I want good paying jobs. I want jobs that count.

I don't want, you know, this isn't the Tennessee Valley Authority and the Depression with Franklin Roosevelt, where I'm giving people shovels and picks, right? If, if that's, that's the fact of it. I, you know, I'm a capitalist God, don't it?

I, I I believe it was FDR R who said that if you cannot afford to hire people at a decent wage, you should just go outta business in the first place. Yeah. So that's where we are.

That's my take on the whole AI thing. As part of that conversation though, I think that what will change is those entry level jobs that people have been counting on to get into it for all these years. They might not be entry level anymore.

They may be at a higher level. So the bar for kids coming outta school to get into it is gonna be a lot higher. And I wonder how long it's gonna take the colleges to figure that out, because it always seems To me, I don't think they have figured it out from the last thing.

I mean, Mitchell knows this in security. What's the biggest thing we hear from these kids who are graduating with cybersecurity areas of concentration in majors? You know what?

We hear it need experience. How do I get, how do I break in? Right?

Every, every job wants three to five years of experience. Yeah. It, it, again, not a new problem.

Right? My favorite job Fact that we don't necessarily talk about is, um, we, we now are of a age of students who are growing up with AI as part of their right. But part of the tool set they use in their own learning whether the colleges have caught up or to it or not.

Right? They're using it. Right.

And they're figuring it out. So kinda like kids growing up with cell phones, we now have college students and, you know, earlier who, who are growing up with AI as the tool that they use, and they're gonna figure out how to leverage it and help themselves get a job, be more productive, maybe do new things, start a company, whatever. Right.

And bring those skills to companies. Well, we already know how to use do it. We've been using it, and maybe they're gonna show their employers something that they didn't know.

Yeah. It's the only people less out of touch than college or the hr. So I'm looking forward to the next ad that's gonna come up that says, want it somebody with 10 years of generator AI Experience.

They're my experience. Well, don't get started. hr.

We'll ask John Willis about that. He married, he's, he's doing some research on that now. But, um, you know, it's funny, my, my, my son who's in law school now, who did not go to Kelly School because of that argument I had with the professor, um, he actually is in the tech lab at Suffolk, uh, university Law School.

And they're working on AI interfaces, both for lawyers and for people who need legal advice. So there are schools that are already incorporating this into skilling their, their kids. They're skilling their students to be, you know, ready when they come out with it.

Our Last two interns in, in the engineering group at techron are doing AI projects or on teams and labs in their curriculum. So, yeah, I think schools, it's making its way getting ahead of this a little bit. Anybody doing any math on what that will do for billable hours for lawyers?

Um, first thing I learned in law school is never do anything that's not billable and don't work for free. Um, still using, they'll figure that Out more processors and Yeah. And check and check the cases that AI sites make sure they're real.

Yeah, I, well they always figure out a way to make money out of it. Don't worry, take no collection up for the lawyers. Don't wanna go all Shakespeare on us.

But, uh, alright. More on that party. Um, well on, uh, as far as the AI goes, um, you know, we would, I think we're, um, good on that topic.

Um, we're gonna move on to some other topics. All right. Well, we'll be back in a minute.

Okay. Alright. 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 guys, we're back and we're gonna be talking now with John Willis about some articles that he's written that are related to AI and all the things that are going on.

John, welcome to the show. Hey, everybody, great to be here. Hey, John.

John, thank you for being our first text on gang member on, on Textron Gang. Um, you know, we're working some kinks out. Appreciate you coming on and working with us on this.

You know, for those who don't know John Willis, why don't we start there a little bit. John, do you want to introduce yourself or you want me to embarrass you? Uh, you know, why don't you embarrass me?

All right. So John Willis has written, is it seven books now, John? Six or seven?

Well, Depending on what books you count, it's 12, but 12 Books or coauthored or co-authored up 12 books. Yeah. But more than that, John brings literally a lifetime of experience in the IT world, starting back in the, in the eighties, if not the late seventies even.

And, uh, you know, was one of the early, early kind kind of cloud ambassadors, let's call them, uh, who were, you know, uh, evangelizing moving to the cloud when it wasn't as readily, uh, accepted as it is now, let's say. But, you know, I started doing that and then saw the DevOps thing happening. He was like the first US citizen who was at the first DevOps days over in Get Belgium.

And then him and our friend Damon Edwards actually did the first DevOps days here in the US and, and kind of, you know, DevOps took off from there. John, uh, you know, was very involved with Gene Kim and, and, uh, Phoenix Project, and the co-authored DevOps handbook. Him and Gene also did Beyond the Phoenix Project.

Uh, so, and he, John spoken it probably most of the DevOps days at one time or another, you know, around the world in terms of cities. Um, his most recent book is on Deming, uh, the Life of Edward Deming, who many call the patron saint of DevOps. Uh, also lean, obviously.

And, uh, John's working on his next book, which I'm not sure he wants to talk about quite yet. But I will tell you, his area of concentration has really been AI over the last six to nine months as it's really ignited his juices passion as it's as all new tech things often do with John. So, John, did I, did I mess any of that up?

I think you got just about right there, buddy. Alrighty. So welcome, Mike.

Why don't you take it from here? So John wrote an article for us on digital CXO, and it's titled, unstructured Data is The New Bacon. And the basic idea of the article, if I can do it justice in a quick summarization, is that, um, unstructured data is what's driving all the training of the AI models.

And there's a thing called the Vector database that is at the crux of that, and it's become the linchpin of ai. And I'll let John kind of dive into what we need to know about Vector databases, but they've been around for a while, John, it just seems like we're figuring out something interesting to do with them. What do you say?

Yeah, I mean, you know, so just to drop a little bit more on my book, it's really, um, my Demming book is, I like to say is like a Michael Lewis version of what you'd learn about Dr. Deming, although I'm not as good a writer as Michael Lewis. Um, so this is, um, the current book is really gonna be that same style of, if you think of Michael Lewis was gonna write a Moneyball version of the history of generative ai.

Mm-Hmm. So the, the, the thing about your question is the idea of a vector goes back to the 1940s when the first neural network was created. And I won't get it now, I won't walk, walk you through in the next six minutes we have left.

But, but the point is, the math behind that was, you know, basically 60 years ago, right? Um, and, um, and so, and there's been all sorts of in incarnations of it throughout the years. Um, I think the big thing is like a lot of the large models that we've seen, like GPT 3, 3 5, and then four, right?

And then we, the, the general public source is chat GPT mostly, right? Um, that is data that is in an neural network basically inside these massive, massive models. Um, and so you're right, I hadn't thought about it that way.

Like those models were created from like, just ridiculous amount of unstructured data, the internet basically, uh, Redis, all these things that were out there. But what, what, what the article I was trying to drive is that it was an IDC report last year that it said 90% of the data in an organization is unstructured. So you think about chat GPT-3 five, right?

It is all this CSET data of like everything we didn't know except about your company, right? 'cause it's, that data's not out there. So the reason why Vector databases have gotten so popular is it allows you the ability really sort of as a, like, I think they call it like a citizen data scientist, is a referred term to people who don't know how to do training models and all this very complex stuff.

You can literally just sort of load your data into these vectorized, they call 'em embeddings as well. And now you have this advantage of your data looking like it's in chat GPT in a very simplistic description, right? And, and that's why it's become so popular.

I mean, you know, and you, you can list the sort of reasons like one, it it, there's a deferred charge of like hitting, uh, uh, open AI or whatever your preferred, uh, foundational model vendor is. Um, but the bigger thing is it's an isolation of your corporate data. So now I can, like my standards and procedures, I can put my governance data or, um, still questionable, like what's the leakage from a security perspective on how much you want to put in there, depending it's host based or not.

But, yeah. Um, and, and the, the real point of that, that article was that we only manage about 10% of the data that we have in a corporation. That means that 90%, in fact, that same report said it's only used 50.

That data, that 90% unstructured data is only used 58% of the time more than once, right? And, and then, so, um, yeah, sort of Shannon Lee's another sort of friend of our community, you know, DevSecOps, DevOps, and now ai, you know, and she, she said, made this beautiful, um, observation. She says that data is the voice of the organization, right?

It is that, you know, it is Bob's directory, it's Sue's spreadsheet, it's Jane's emails. It's, and, and the thing is, think about we spend 90% of our work on the 10% data, and we don't really manage that other 90%, that there's so much value. And that, and that's why vector databases, the, the whole concept of rags, right?

The retrieval augmentation, there are other forms of effective basis. It's being the most popular. So that's what I was trying to point out.

And, and just for any old Cloudera people listening who are just kicking the table and screaming how I agree, I totally admit that I stole it from a T-shirt you used to make, which was data is the new bacon. So there You go. Fair enough.

Hey, hey, Mitch, I have a question for you. Yeah. I'm, I'm just curious about, you know, it, it's easy to think of, uh, generative AI as kind of a search engine, right?

Dump your stuff in there and then you query it, right? It's actually much more sophisticated. Uh, talk a little bit more about how vector databases make it more accessible to non AI computer scientists to the, the general practitioner, maybe d data engineer in an organization.

Yeah. So to give you that kind of data layer between that and the full large language Now, well, I'm, I'm glad you asked that, Mitch. No.

Uh, so one of the proudest moments I had in my Deming book, which is I was able to explain Schroeder's cat and my mother-in-law understood it. So, so my, my sort of bar here now is there's a famous, um, you know, one of the first real neural networks, it was called mni Database. It was, um, it was, it 70,000 handwritten numbers zero through nine.

And it was used by this, um, John Koon, who basically, they were literally doing a contract at Bell Labs for like NCR to read check numbers on the check, and then ultimately zip codes. And so he basically, and it's the 1 0 1 of learning ai right now. You go out and you find Amus database to 70,000 and you learn how to build your own train neural network.

So like, I'm not there completely yet, but like the, the, I wanna be able to explain that so that, again, my mother-in-Law will be like, wow, I get how this neural network thing works. And, and, and it simply, um, it's basically vectorized data. So like, you have these mo highly dimensional vectors.

You think about two dimensional or three dimensional, right? What, what you wind up doing, like in, in the sort of the, the, the open ai, most of the open AI models, it's 1,536 dimensions. And so what happens then is you take your data, especially in a Vector database.

So if you're gonna sort of use your Vector database in concert with say, GBT four, um, you would then you use this technique to sort of chunk it up, put the words in, in, there's sort of, there's some real cool math about context of words and sentences, but ultimately it's words which get what they call tokenized. And not to get too deep, those turn into floating point numbers represent where it lives in a 1,536. Um, you know, so think of a three dimensional, if we just use floating point for three dimensional, it would be three numbers, right?

It's 1,536 for some context related stuff. And so what happens then is when you do your searches or you do your commands, and depending on how you're doing, if you're using something like an orchestrator, like a lane chain, it'll do a lot of stuff behind the scenes. But if you're just running these plane, you might say, cat.

And what it does is takes that word and it literally does a similarity search. And it's gonna, again, I, I like, someday I'll get this, if it sounds so confusing. There's like a number of different similarities.

One of them is what called co-sign similarity. And it's just to cosign the angle in this massively dimensional of, to tell you how far your word is from the word that you asked. That's in this sort of the vectorized data in Kat.

And the similarity would be some percentage of how close that is. So when you start typing in who is Abraham Lincoln or you, you know, uh, if you're working with your own vectorized data, like what, you know, what is SOP 1 0 1, you know, uh, what are the rules of our SAP for installing routers? Um, you know, it will basically, you know, do that sort of similarity search against these math loading points and just come back with the ones that are closest.

Um, John, I have a question for you. Uh, do you think through these searches that you're finding that, that people are tapping into data that they didn't even know they had, like, let's say a, a client or customer's social media tweet or something like that and they didn't, they didn't know that it was there. Now it's part of their mass database that they're looking into?

That's a, you know, I, now I know why Allen has you on the show that that's a, a fantastic question. Um, you know, I've always thinking about the other way, right? 'cause I'm doing a lot of work on how you take people's data and how do you get a higher similarity, I call it efficacy.

The data science would probably get mad if I use that word. But, but, um, you know, so like I'll, I'll put data in and I'll sort of run some similarities and say, am I getting the right data? But there are really good examples of, you know, there are, there are some data.

I think the more it's the, that is the bacon, right? Mm-Hmm. Is the more you can get that unstructured data in there, I mean, at, at a surface level, you, um, at a surface level, you don't know what all the data is, right?

So like, if I can just start tapping into this 90% unstructured data and putting in a format so I can ask questions about it right off the bat, I'm learning things. I don't know. Like I'm turning, you know, not only data in or data and information, I'm turning into knowledge and understanding, right?

But, but I, I'm, I'm trying to think. There are some examples I've seen, and they're not coming to my head where like, I've been surprised. Um, oh, I, yeah.

I mean, yeah. I mean, it's not quite the undiscovered knowledge, but, um, but like the, um, I went to build a study guide on my Deming book, and it was amazing the, the insight it had, you know, the, you know, the, like, the, like I, you know, I, you know, I had read that book through eight revisions. I probably read it cover to cover 30 times, you know?

Um, and as I was going through it, I was like, this is more advanced or equal to my knowledge of this book. And it did it only one time. Wow.

It was able to tap into, it wasn't a hundred percent accurate. I had to clean some stuff up. But it was just amazing that I did it by chapter, on chapter.

And it, it in the, um, the, you know, I sat back for a second. I had sort of an outof body experience where I was like, like if this was a person, I would be shocked. I would think they would have read this book at least 20 times to be able to, uh, but, uh, but yeah, you're right.

It is finding the understanding and, you know, I think in the sort of tree of cognition wisdom from just getting your data in there because the power of these engines is just the math behind. It's just so incredible. John, I've got two, two questions points I'd like to discuss with you and the rest of the gang.

Number one, as you mentioned, vector databases are not new. They've been around 40 years or more. Um, but what, what was kind of surprising, and I learned this at the Hackathon, which we're gonna talk about that we did down here with John and a bunch of people I guess in August, um, is that a lot of the Vector database providers, vendors were a little slow to realize that they were sitting on here.

Right. You know, they didn't really have that keyed in, right? So like for instance, you know, there, there are some, and, and you know, the database players better than me, John, there were some guys who recognized it right away.

And they quickly gained a big following in, in the AI crowd, right? And then, and the pineapples of the world, right? Yeah.

Sort of purpose built for ai, pine code. Yeah. And then, you know, then you had something like, like Mongo for instance, right?

Mongo has a good vector database, but it, it, they were probably, you know, John, you, you were, you kind of opened their eyes a little bit for them into what they were sitting on there. Well, Did I think, you know, so I mean, you're right. There was, there was some, like, there, you know, it'd be fun to go back to the history.

Like, it, it's like all things, you know, you know, Chad GPT blew the doors open on everything, right? There was sort of an ai, you know, one of the things I'm cover in my book, there were like two AI winters, right? And there was an AI winter before leading up to, you know, sort of the things that were happening starting about 2014, right?

And, um, and, but the, um, so the, yeah, everybody had these sort of embedding tools and these sentence transformers and all these ways to do sort of similarity against sort of homegrown data. But it was when, when, um, you know, so open ai, you know, sort of introduced three five and then chat GPT, the world, just like, like, and then that opened up a bunch of questions. It was kind of like, I, I like to use this like, just a little segue, is that like, there were lots of people using containers.

Google was using it at core. All the passes were using containers, but nobody was using containers other than them. And then Docker came out and commoditized it.

And now like, containers are a way of life now, right? Mm-Hmm. Um, so I think that's the same thing with, you know, like if you look at what open AI did, they literally bust the door open.

So, yeah. So the thing about, um, MongoDB, which is really interesting, is I started playing around with like the pine cones and a couple of the early adapters. Chroma, chroma was the Chromas DB is like your first 1 0 1.

You, it, it's an embedded, you don't have to install anything. I mean, you gotta install the sort of libraries, but you don't have to, it's not SaaS based, or it wasn't SaaS based. And then you sort of upgrade to Pine Cone.

And then the thing I found was, and this is the thing I, you know, I, in all transparency, I'm, I'm, I'm doing consulting for MongoDB right now, but I actually believe, and the reason I went to them, they didn't come to me. 'cause I knew Peter Lander, who was the CMO there. And I was like, you guys have an opportunity that I don't know that people are seeing.

And, and it is that, and it's the thing I, I sort of telling everybody, it's less about AI and more about the data. This is all about a data play. You know, there's a poss it's like, remember the, the magazine called Computer World?

Mm-Hmm. Yeah. Like, where is that magazine?

Like what's a computer like, it's a fabric thing. I, I think AI is just gonna be a fabric thing, and it's still gonna go back to the data. So it's the per the people who manage and deal with the data, and this is their opportunity.

They do data. You know, I mean, who's the only company that does well? I mean, again, there's some, the, there's some of the big data candidates, but like Oracle and then MongoDB has been in the enterprise for unstructured data for, you know, 10 years at scale.

Mm-Hmm. And so I think, you know, that this, and what you find, the reason I like MongoDB not to turn this into MongoDB commercial, but is they treat all the data as sort of one database. So you create, like, it's a document object, which everybody's familiar with, every technical person, in effect, almost every data center how you face is JS om based.

Um, and, um, probably all, but I would say almost all. And it is that, um, it's a document object. And then one of the fields in that object is an embedding, which is a vector.

So now you have this incredible flexibility. So anyway, that's a short Bit. You let me move into another area 'cause we're gonna, we're running on time here.

No Worries. So right now, it's still somewhat novel to hear companies creating their own LLMs, right? And then they feed that on top.

You know, you feed it into a chat bot or, or an AI engine that is also plugged into a, a larger ai other than just your own unstructured data and, and people sort of ooh, and ah, but we're, we're quickly moving past the ooh and a stage of that as that becomes sort of the, eh, that's table stakes. It's anti, right? You have your own LLM running on top of a larger, uh, body of knowledge or data.

Um, we, we, I mentioned this hackathon we did back here in August, and we had some amazing people. Shannon Lee, you mentioned Tracy Ben, and, uh, of course Patrick Dubois was here. And, and Damien Edwards and Alex Honor, and, uh, a bunch of others.

Stephen McGill, who I'm gonna interview next week, um, from Sonatype. But John, you know, that's when I first learned about custom LLMs outta Vector database on top of regular. And this whole, you know, in my simple-minded way of looking at it, you know, you have long-term memory, short-term memory, what do you access first and all of that.

But yet, when I speak to people, especially non-technical folks, this is sort of revolutionary to them. And, um, I don't know what it, I mean, those hackathons seem pretty productive to me. Yeah, no, I think it was a great starting point.

'cause they, I mean, if you think about the, the concept of dev and ops, you can look at generative AI as a DevOps problem. Mm-hmm. Right?

And the one problem is, which I haven't focused most on, too much on, but like the dev side is, you know, like, how do we start, how do we get so citizen data scientists to train models? I think that's the next leap, right? Vector databases allow you to do a lot of stuff without having to know how to run things like PyTorch are very complicated.

You know, I think very complicated from my background and my peers background. So like, I think, and then making it really easy to get the data in, in a way that's sort of already formed and has high similarity, right? That's the future of development here.

Um, on the ops side, what we focused on at that hackathon, which was this, what I call this technical debt tsunami that is coming down the super freeway, uh, because all this stuff's gonna wind up in the enterprise. You know, every time you list like a new vendor that says, oh, we got a, um, copilot at ours. You know, the question that, um, it's gonna come up is like, or, and right now the ops and infrastructure people don't know how to ask the right questions.

Hey, what's the stack? Mm-Hmm. Look, if you're gonna put like this vendor's new, uh, copilot in my shop, can you at least tell me what the stack is?

Well, what do you mean? Well, I know you didn't write the embeddings. I know you didn't write the orchestrator.

I don't know, you didn't write the model. I didn't know you didn't write that. Like millions and millions of lines of Python code that you're using.

So again, can I ask you what the stack is? And so just having that conversation of people who aren't understanding. So the, getting all those people together in that room, step one was try to educate.

We had Joseph Enox, right? He gave this amazing presentation. We, uh, we all learned like our, our jaws drop.

And so I think the biggest thing we have to do as an ops community and a DevOps is like, educate our community to talk intelligently. So now they can understand when the CIO says, are you okay with us bringing in this product? Absolutely.

Like, well, I got a couple questions and the questions they couldn't ask a month ago, two months ago, six Months ago. Well, they didn't know the questions to ask. That's right.

That's right. We, we need to, to wrap stuff up. But quickly, I wanna mention the, the hackathon.

We did do a condensed video version of it. I, I forget how long it is. 40, 45 minutes.

It's available on Text Trunk tv. And we are announcing, if it hasn't already been announced, we're doing a, actually, mark Hinkle was at the hackathon too. Mark is teaming up with us and we're doing a, a tremendous kind of, kind of all day AI virtual session with a group crazy lineup of speakers.

Um, that'll be, Yeah, I'm running the DevOps track for him. Yeah. So, yep.

John's running the DevOps. We'll that's in late May. Yeah.

com. You should, if it's not up now, it'll be up in the next couple days. Start registering for that.

And of course, at RSA this year we're gonna be doing DevSecOps and ai, so stay tuned for that. It's Monday May 6th. Mike, Just, just for reference, the video is on Textron a Ai.

ai. You're right. Yeah.

Right. Because we showed it on Textron tv, but for that one time thing. ai I think as of today.

Mm-Hmm. And so in case you're looking for go check It out there. Thank you for correction.

It will be now if it isn't. Hey John, man, thanks for coming on and, and, and, and enlightening us and entertaining us and, and just for your patience and, and support as always. No worries.

Alright, thank you. Thank you, John. Alright, I think that's gonna wrap up Textron Gang for today.

Stay tuned. I'm gonna be back in a moment and I'll, I'll give you a preview of all the great other content we have on Textron TV today. But for Textron Gang for today, this is Alan Shimmel on behalf of Mitchell Ashley, Mike Ard, and of course Bonnie Schneider.

And special guest star John Willis, Textron gang member. Have a great day everyone.