Atop a Falcon nine booster is set to lift off from Launch Complex thirty nine a the Kennedy Space Center at ten fifty three pm Eastern. I'm Chris Karragio, NBC News Radio, NBC News on CACAA Lomolinda, sponsored by Teamsters Local nineteen thirty two, Protecting the Future of Working Families Teamsters nineteen thirty two dot org. The information economy has a rid. The world is teeming with innovation
as new business models reinvent every industry industry. Inside Analysis is your source of information and insight about how to make the most of this exciting new era. Learn more at inside analysis dot comsideanalysis dot com, and now here's your host, Eric Kavanaugh. Indeed, ladies and gentlemen, welcome to the future. It's careening toward us all the time, but it seems like it's coming faster
these days than ever before. I'm not sure exactly why that's true. They told me when I got older, time I go buy faster, so maybe that's what's happening. But I could swear this Jenai stuff has just put everything at light speed, certainly for the enterprise, but also for small business, for anyone who's played around with this stuff. It is a leap into the future. And as the futurist William Gibson once said, the future is here
already, it's just not evenly distributed. And I love that expression because it really speaks to the pockets of innovation that you see around the world and helps you understand that new technologies and new methods do take time to percolate through society, to permeate to various circles in the business world or in the consumer world, for example. And this generative AI stuff affects darn near everything, and
it's really open the eyes of businesses worldwide. I mean the numbers we're hearing are just off the charts in terms of the number of people using this stuff already. I mean, hundreds of millions of people are using this technology already today and it's more or less brand spanking new. So it was twenty twenty two that it was released in general chat GBT three I believe are three point five. We're up to like four point five now. Of course Google has
barred now it's called Gemini, anthropic has clawed there or other. There's Mistral that has the interesting approach. They're taking with smaller models. But the bottom line is that Jennai is everywhere, and today we're going to dive in with two awesome experts. We've got Baoja Bui from FPT. It's a Vietnamese company, amazing company. I went to Vietnam last year to check out their headquarters and everything they do and was blown away. Really impressive company, very hard
working, very diligent, very focused. They get stuff done. They're not worried about the flash and anything silly. They're just focused on getting the work done and building cool technologies. So we have Baoha to talk about that, and Jim Wilt as well is a fractional CTO and distinguished engineer. And these folks are really sparked. So in studio of audience, live studio audience,
don't be shy. Asked them the hard questions on a first to say my take on what Jenai is, and I'll throw it over to Booja and over to Jim as well. Really, these Jenai engines are predictive engines. That's what they really boil down to. They are predicting either the kind of image they think you want based on your prompt or the kind of text you want
based upon your prompt. Now, they were trained on massive amounts of data on the interwebs, not on everything we learned a couple months ago and one of our first shows of the year that even the folks at open Ai realized early in the game they can't just train this on any data that's out there. They have to focus on certain data sets to get the end result they want, which is an engine that will reflect back the kind of text that you want. And it's very powerful. Folks. If you haven't used this
stuff, you need to start playing around with it. And you first start, then you understand what it does. Then you can kind of get digged deeper into the weeds to figure out how to use it. But they're basically predictivens, is what they really are when you get right down to it. So at that Boha, tell us a bit about yourself and what you see. What is Jenny and I to you, and what should people use it for? Yeah, well, thank you Eric for that comments that we are
not flashy. I want to be flashy, but we just we techy people just can't be flashy, So it's a skill I have to up myself for. But that said, though, so bo. I've been in IT and technologies for almost thirty years now, or even more than that. I stop counting the moment we hit above twenty my so been with IT technology. I've been CIOs, I've been vps, I've been on the application development side, been on products. I've been but since twenty fifteen, No, wait,
twenty ten. I you know, I got introduced into data management, and you know the journey starts there. I got passionate about just making informations out of data. You know, how do I get that information into the hand of the users, the executives, the people who's driving trucks, not needing to know when I have to deliver a certain product to a certain customers of patients. I was in healthcare for ten years at that point. Sale my
whole the introduction to data, I was with the HIE. You know, if you still remember those terms, you know, the data for healthcare was a big thing back in two thousand and five ten. So that's my introduction to data. Now I've been with FPT, I've been their customers, I've been working with them, and I really appreciate your confirmations that you know, they are hardworking, very smart as a customer. I was constantly amazed by the level of professionalisms that I got from Amputy. So not a plot,
I'm just kind of recounting my journey here. What is jen Ai? You know, Jenei? Like you said, it starts out to be or aio, all right, It starts out to be predictive. You have data, you have the ability to use code to really mind that data and then generate information based on that data or predict based on the patterns that came from that data. And then it gets into the ability to really now using patterns to start producing new material. So that's what Jenei is for me. And where
are we going with jen Ai? It's everywhere, it's you know, Jim. With Jim and I talked a lot about jen Ai and the growth of jen Ai and Ai overall. But Jim, it took fifteen years for cloud to be adopted. I still, you know a few years ago there was a company that's still not looking at cloud. But jen Ai in two years
we got widespread adoptions. I talked to my son. My son is doing a co op at a farmer company, and I asked him, so, what are you doing over you know, your bread time oh, he's using jen Ai to do some optimizations in their lapwork RNA recognitions RNA pattern So you know, it goes from a corporate level to students. You know how many the professors trying to prevent their student from generating ai jen ai for papers.
But is that not a bad thing because at the end of the day, is why reciting information that you can get from Google just put it in your own words? Is that you know, so again permeate every aspect of our life nowadays. So we all have to get onto that bandwagon. That's a great way to put it on. Like the way you said that, where are we going with Jenny? I everywhere? I'll throw it over to Jim to comment on that. I have to say, I mean I watched the
whole big data craze take off fifteen seventeen years ago. Let's say that was the biggest thing I'd ever seen. This is like five to ten times bigger than that in terms of the number of people getting in, in terms of the impact, in terms of the number of people and impacts, it's just off the charts. But what do you think, Jim, Yeah, it's interesting. You both have really covered a lot of the basics that we wanted to, you know, get out there in terms of the history and where
it's come. Baha's correct, and looking at adoption of the PC was you know, about ten years, fifteen years to get to fifty million users. That took you know, about ten years to fifteen years for cloud adoption to really land. And when we look at machine learning and generator of AI, machine learning has been around for ten years plus, it's been around. There are people that have been working in the space for over twenty thirty years with
neural networks and so forth. But the big difference with generative AI is is it's really applicable to the common person. Anybody can get into a prompt situation to some learn some prompt engineering, if you will, and start getting results that are just massively interesting. And I think that, you know, the statistic that hit me that Baja mentioned is actually it was two months generative AI has reached one hundred million users. That's twice what it took these other technologies
what ten you know, fifteen years to achieve. You know, the Internet took five years to achieve fifty million users. It's already surpassed the airnet's adoption in just two months. What's also exciting about Generator of AI is that we're traditional AI is using what's traditionally called supervised learning. Baja mentioned it's and you did too, Eric. It's mentioned that it's based on a lot more data,
and that's called unsupervised learning. What that allows it to do is to create, when I'm going to say, patterned recognitions and results that they're out of our view. So it looks like it's thinking, if you will, on its own, but it's working with such massive data. It's it's really really exciting to see that it can find patterns that we normally wouldn't find ourselves. And I think that's one of the things that makes it that aha moment
that oh this is different. It's accessible, and it produces results that are pretty transformational because it's putting together patterns that we expect to see that are rallied. Yeah, that's a great way to put it. And maybe Booja,
I'll throw this one over to you. What I've seen with AI and other capacities too, and now of course with gen AI is that almost invariably it surfaces something that I did not expect, and that's very interesting because typically any AI algorithm is just a recipe that is designed to address a particular type of situation. Let's say a particular function, a particular array, or something that happens in the world. You're trying to predict what happens or optimize what happens.
And you would think that we know ourselves well enough to not be surprised by our own behavior, but that's just not the case. Like and so every time I use one of these engines, it throws up something I did not expect, and that's good news because that's it triggers the creative process, and it makes us think more. What do you think? I think I absolutely agree with you. I think that we, as great as our human brains are, we have limited capacity or you know, how fast we can
generate, oh, how much we can retrieve. So it's not I don't think that it's a it's a I think it's more of a retrieval, you know ability, whereas the machines it's gonna take back. It's gonna take everything that it knows and give it back to you, and it reminds you of things that hey, I knew this, but I didn't remember. And that's where the benefits or the efficiency gains in using generative AI is so great.
It's because it really takes away your the need for us to remember everything that you have to you know, that's why we're saying the old day, but it's really is still there. You know, you have document template because people cannot remember everything every single time, so you need template to repeat to save time. You don't have to recreate things all the time. Generative AI is able to really do that retrieval of your template and give it all back to
you very timely. But you know, retrieval and knowing what to retrieve and when to retrieve is also important because yeah, as much as we like to get everything back, you don't always want everything that JENEI give back to you, at least quite yet. Well, and that's a comment I had made
before our show. I was talking with Mark Palmer, who was a real smart executive in the tech space, and he mentioned that he read some research recently that showed something like eighty percent of people who are using jen ai right now are just taking whole hog, whatever it comes out of that engine and using it. That has a very bad idea right about how you want to
always vet that stuff Oh. Absolutely, it will give you good ideas, but you have to be careful because number one, it does hallucinate, so it does make stuff up. It is generative by nature, so it generates new things based upon old things. Right about what it's seen. All it's doing is reflecting back patterns of what it's been trained on. But you have
to be very careful about vetting that stuff, right huh. Oh, absolutely, you have to because it's trained on informations out there, you know, especially some of the open source gen ai open ai for example, it's being trained on the Internet, and not everything on the internet is usable. Just
be careful. Not everything is accurate, right, Not everything is accurate and even when so that's kind of going into some of the rest that when we well, you know, what we have to do in order for us to be ready for gen AI is some of those vetting the under the training of your resources to make sure they know how to use it safely and effectively.
Yeah, and that's the curation process, right. So we keep hearing about RAG models, retrieval, augmented generation, which, if I understand it, really what you're doing is you are trying to use your own trusted, governed data as an anchor, and then you want to be able to spin up text using the LM as a text generative capacity if you will. But you wanted to use your facts that you give it, and those facts are posted as embeddings either in the model itself, which some people do, or in
a vector database which is then adjacent to the model. And people understand the whole concept of persisting something in the database and then recalling something from the database. But here we've added a third element, which is the language model to
generate stuff. But still the same concept applies. Right, We're persisting something in a data store, and then we're leveraging that along with the generative capability of the large language model to create fresh new text that is ideally accurate with what our corporate data says, right bouh In a way, yes, so you know we you know, if you're using your data alone and retrieval and then to supplement it with the large language to to make the interactions easier or
to generate it's in text and it's limited. But the ability of gen AI to you know, so large language model gives you the ability to read the data that in a database just not possible you know pdf my You know some of the older companies that I have worked with, the Banks Insurance Company, they have one hundred years of the documents that's probably still in some kind of fish or images. Large language model nowadays give us that capability to really retrieve
that information. And you said either ambed or vector base it, so that now you can retrieve it. So that's the advantage of some of the newer capability. And then based on that, then it gives the ability for us to interact with it using natural language. So that kind of expand the resource pool of who can rechieve that kind of information versus before all the tech people
have to write some kind of query. Right now you have to you know, a much larger resource who can comp for those right And see, this is I mean, you just hit on one of the biggest transformations that we're seeing right now. And I heard it from my buddy Loo Simon with a company called up Tom here in Pittsburgh. He's the one who first said it to me. And this is so big and we'll talk about this after the
break, but this is a seriously big deal. We used to always have to learn how to speak in computer language in order to talk to computers. So we had to learn to speak SEQL to struct query language to talk to databases. We had to learn Python to talk to various Python apps or that's used for risk management, lots of different stuff. Honestly, you had to learn cobol in order to talk to a mainframe for example. Now you can talk in English words to the computer systems. That is a huge deal.
We'll be right back here listening to Inside Analysis. Respect, welcome back to Inside Analysis. Here's your host, Eric Tabanac. Take us all right, folks, take us to the future and see if we're talking about jen Ai. It is futuristic. It's amazing, folks. Let's be honest, play around with this stuff. It's going to blow your mind. A friend of mine turned me on to mid Journey, which is through uh discord. I
guess it's a discord channel, and oh my goodness, it's crazy. The amount of artwork and the sophistication of the art that you can create with this stuff. It's just wild. So things are changing rapidly. I remember hearing who is saying, whoever wins this war wins the world. Now again, there are use cases for this stuff. There are dangerous sides of it, deep fakes, things of this nature. We're going to talk about some of that, but this segment I wanted to get deeper into are you ready and
how do you know? So? How do you know if you're ready to leverage this kind of technology? Organizational readiness? As we said we picked up last segment. We're talking about how you can now talk to computers in English words, in prompts right as supposed to having to learn SQL or Cobo all these other languages. And you can also use it to tell you what code does. You can feed it code and say what does this code do? And it'll tell you. It won't be always completely accurate, but it'll be
pretty darn close. And I've heard from enough people now to know that it gets pretty close. It gets to about eighty percent in terms of coding of writing what you want, and then you just fine tuned to get your last
twenty percent. But Jim Wilde throw it over to you. In terms of organizational readiness, well, I mean that the most clever thing I heard in our pre call last week was to start your first use case internally, so nothing that faces your customers or your prospects, your partners, but do some project that faces just your internal staff to learn about these things. Because it's like a new pair of skis or something with jet engines on them. You
have to figure out how to ride them. But what do you think about all that, Jim, Yeah, that's a good way of putting an analogy. It's like seize jet engines. You wouldn't go out on the slopes with other people around if you have jet engines and your skis. I think one of the interesting aspects is there's fear, uncertainty, deception sometimes and organizations.
When you're looking at the adoption of jenator AI, where we're excited about it and public in general is excited about it, a lot of organizations have, when I'm going to say, a healthy fear, which is in most situations pretty healthy. Right when it came to PCs or the internet or cloud, it was good to come in it kind of slow. It's not going to work this time because you really are not prepared for the adoption that it has
gotten in the general public. And so if you have a policy in your organization not to allow people to use generative AI, be aware they're using generative AI no matter what it's on their phone, and I know that we want
to protect corporate assets, corporate secrets. Find that's great. But I think if you go about the adoption we talked about, I think you're going to find you can ease your way into it and get what I'm going to say, a lot of those fears dismantled and a lot of the excitement generated because you're doing it with internal uses, internal purposes. You mentioned about code analysis,
I think it's a code companion. Every developer should have, you know, a time to their hip, because it will help them find errors that they're putting in their code. So fast. You don't let it write the code for you outright. What you do is it's your companion. It's going to find your blind spots. It's going to find where you made a mistake. I wrote an example of this using multi threadic code, and I put
an air on it in purpose and it was amazing. Not only could the generative AI find the air, which none of my colleagues could, by the way, it also fixed the air, which all of my colleagues could do, but it fixed it in a very concise and correct way. Now I'm in charge of the code therefore I'm the one responsible accountable. I should never just let it fix the code and put it into production. That's just silliness.
But if you do think about it, there's a lot of value it can be used internal to organizations that makes the organization execute much more operationally efficient. And that's something Bohan I have been doing with other organizations is coaching them on ways to enhance their internal efficiencies using generative AI. Yes, everybody wants to be the first one out there, and there are some amazing products that are coming out as products from organizations, but if you're new to it,
you really need just to kind of frold it inside. One of the retail organizations I highly regard as highly innovative has taken and create a degenerative AI tool and given it to fifty thousand of its employees and they don't know what they're going to do with it. They just want to learn what they're going to learn from it and what they're going to do from it, and that's that's
a great attitude. You give them a safe tool that's not going to put, if you will, your corporate data in danger, and you see how they're using it, you monitor, you see how they're using it, and you find out what they insights, you know what blind spots they're finding, what got you moments, they're getting those those things you said, you you know, Oh my gosh, how did it come up with this? Yes,
exactly. Everybody should grabbing that. And I'll kind of pass a lot about how because I think she can probably talk a little bit to the concept of encouraging operational efficiency in organizations first in terms of learning what you can do with the tool before you go crazy putting it out there for everybody to see. Well, there's just so many things that we can gain efficiency, right you know. So Jem and I been working together on a few of them.
One is to lavish AI for sound analysis, you know, to look at what can we do to use sound to detect issues in machines and predictive maintenance or just assist in the diagnosis of it. There's others, uh use where we have products already out there using visions too to detect issues on say, you know, winterbine, if that's a crack, how do you know?
You know, so monitoring using vision AI. We're also working on code assists, like Jim was talking about, every developers nowadays really need to leverash ai jen ai to help them just make you know, as a double person, Jim was, I don't know, Eric, have you ever worked with any kind of coding. But at the same time, you know how tedious
it was to sit there and type all these commands. If there's something that really accelerates that coding for you so that you can focus on the big concept, that's where optimizing the efficiency of every aspect of the organizations will be wonderful. So that's that's where we are seeing. You know, organization really should adopt look at what are the tedious things that nobody really want to do or they want to do but they wish it could be fast. That's where we
apply AI to help. But how do you get the organizations ready as well? Right? What are the things that let's just say, if you create the tool to set, I don't know, predict something within your organizations or to generate some content within the organization. First of all, you need to make sure that your model is chain on information that you want your employees to use and don't just train it onto the internet because who knows what's out there.
So you know, organization readiness, making sure that the organizations, the employees are ready and understand what are some safe AI practice should be. And then also from a technology standpoint, setting up the tool that can deliver that safe products. You know, I came from data. Data governance is my preach of the day these days, you know, every single time, get your data governance in place, get your data clans, get your data in a place where your model can get to so that it can be trained.
Because if it takes your data scientists pages to get to the data, it really doesn't get your organization ahead of the game. That's and that's a really good point that you just made. And so I'm going to throw out a thought here and maybe a throw it over to Jim first, and then baoha. You can ask it what am I doing wrong? You can ask it
all sorts of questions and feed it a document. So one of the I think relatively underappreciated and incredibly powerful use cases or functions of these tools is summarization. So you can take a very long document, let's say a thousand page documents, feed it in and it's almost like a dynamically generated cliffs notes. When you're in high school and college of these cliffs notes. Right, It's like Boom's like for you. It check out all the filler words and it
come dons for you and get you going so fast. You know, Jim and I were just experimenting what last week we point a model to some users document user guide that we found on the internet and then feed the engine that we create a bunch of code and TELP to interpret it for us. It did it just like that. Jim got the engine all set up in what fifteen minutes fifty minutes to get the first setback, and then you know it takes a little bit more training to get to some more eligible result, but
less than a day. Yeah, that's that's crazy. And I was talking to some folks from Matillion, a software company MDM style company. They had a use case where they said, look, let's just try to throw it at chat GPT and just see what it says. Just and some people were skeptical, like, just give it a try, and so they did, and their minds were blown how accurately it came back. And so the point being, think about the whole side of discovery. When you're doing data analysis.
The discovery side is so important because you're trying to find the unknown, right. I mean, it's one thing if you're trying to find a known like with a query to the database, give me what have already persisted, But if you're trying to look for patterns that have not yet been identified, this kind of technology is incredibly powerful because it will on a moment's notice, absorb tremendous amounts of data and then throw back seemingly random reflections that are very
accurate and very compelling. So on the discovery side, I mean I kind of view these engines as being a new front end for almost everything in the enterprise. Eventually, what do you think about HAA, I agree, I agree the discovery. That's exactly what Jem and I were trying to figure out last week is how we have an obscure platform that not that many people know about anymore and very hard to find resource for even if you had resource.
Some of these platforms that have been in an organization for fifteen twenty years, the amount of organic growth and customizations and that governance on this platform are just mind blown. But you know, how do you really accelerate that discovery process so that you really get the engineers to where they produce results versus doing these tedious, unvalue added activities. Honestly, so Ai, that ability is just it amazed me still and get me very excited. Well, and you do.
So the guys from Matillian we're mentioning they did then need to go, so they figured something out about how to do it, and then they actually had to go do the hard work with their existing information systems to get the design work and the engineering done. At least they knew which way to go. At least they knew what's possible, Jim, I'll throw it to you. Being able to suss out what is genuinely possible in any given situation is
half the battle. Then you have options that you can choose to go in this direction or that direction. But if you haven't effectively discerned the nature of the challenge or the range of possibilities, you're still in the dark, and you're shooting in the dark. So this discovery component of these large language models, I think it's very compelling for being able to understand what's happening and then
have a really good idea of where to go next. What do you think, Jim, that's interesting because the discovery process can be enhanced by asking the generative a engine give me ten questions. Ask me ten questions as I'm going to discover something, and then we'll prompt you with best practices of discovery to
feed itself information so it can start offering you better analysis. The example how I was talking about was so much fun because it's an obscure platform that really hasn't been leveraged, let's say in tens of years, and we are able
to find an old syntax manual with two thousand pages. So doing the pag approach for grounding or language chaining, you feed that two thousand page manual in and now it's making what I'm going to say, very precise outputs that are specific to the rules of the language and the syntax of the language that it
was built in. And so you can take things that are ancient in the computer world and make it very modern to the answers that can give with the very accuracies that you're seeing, and you know it could be one hundred percent accurate eighty percent of the time, and the times that it's not accurate, you're going to have to have enough knowledge to call that out. The hallucinations of the biases the blind spots. However, what we found is that as
you're going through a discovery process, it's those those gotcha's. We're basically things I wasn't thinking about. It will think you know, have you thought about and it will give you some things. Have you thought about this? Have you thought about that? And it's like, no, actually I haven't.
So it's a companion, right, It's kind of that person that that can can have a conversation with you and throw things out at you, but you're not going to reject them because you're asking them to do that for you. If that's a really good way to put it, and we'll talk about this in the next segment coming up here in just a moment. But using it as a companion, like an incredibly knowledgeable intern who will tell you whatever you
ask it may not be right all the time. It kind of reminded me of the old Yogi bearra quote, Right, ninety percent of the game is half mental or something like that that can be completely correct. Eighty percent of design the clock the broken clock is correct two times a day. So well, don't put that to help folks, will be right back. You're listening to a very cool episode of Inside and Out. Welcome back to Inside Analysis.
Here's your host, Eric Kavanaugh. All right, folks, back here on Inside Analysis, Rocket and Rolling today with baalja w from FPT and distinguished engineer and fractional CTO Jim Wilts. This guy knows a lot and a lot of people too. You know, there's a lot of cool people. He does a lot of stuff which is always fun. And we want to talk about pitfalls. There are always pitfalls. I think I'll just throw it out there. There are already these anecdotes that get thrown out. It's amazing how
fastest stuff happens. Everyone's probably heard about the lawyer who went out and threw citations in and it turned out to be nonsense citations. Well, shame on that fool for not just taking the extra step to make sure. And this is where you get into the rag model concept right. In other words, you want to use a trusted data source like whatever sources they have for cases, for historical cases. And I'm telling you, folks, lawyering as an
industry is in the crosshairs. Accounting as an industry is in the crosshairs of AI. All these things that we've done traditionally are absolutely positively in the crosshairs of AI, and as Jim was kind of suggesting eighty percent at the time, it'll be one hundred percent accurate. You do have to know when it's the twenty percent to take a hard look at stuff. So that's the biggest pitfall I think is don't just rush in and start throwing this stuff it generates
out into the world. Be careful about what you make that's customer facing. There's several issues already, but those are probably the biggest ones. Jim Will, I'll throw it over to you. What are any other major pitfalls you've seen that people should watch out for with jen Ai. I think one of the things I run into and you almost have to take a pitfall as comical
and enjoy the entertainment aspect of it. When I was writing an article, I was tweaking hyper parameters just to see what happens when you break it right. And I had a test explained quantum entanglement, and I did it with different temperatures and I chose one that was out of the parameter range and it created gibberish. I mean, my you know, one year old granddaughter could have done better. But it was just random characters group together. So it
wasn't even English language. It wasn't any language at all. It was kind of thing. So it will break and you and part of the learning process, you need to know when it breaks, how to break it, buy it breaks, you know, and you need to be on guard for these kinds of things. But as I was finding things, question the answers. That's probably one of the things I would say in every situation, what it's
going to be critical. You talked about adoption in another earlier segment. You want to start with none mission critical and work your way towards mission critical. You don't start with mission critical. That's the lawyer example you gave and others that are out there in the last week. You don't want that. Start with non mission critical and work your way with confidence and learning and maturity.
I've asked times when it gives me a quote, can you give me the source of the quote, and it will say, oh, I'm sorry that I made that up. I mean literally, that's the words that we'll use. I'm sorry, I made that quote up. That didn't come from that Forrester Gardner article, or if it does refer to an article, I'll say Can you give me the number enumeration of the articles so I can look it
up and it will give me a false number. So you have to drill down into, you know, the material before you publicly go somewhere with it. If it's just helping you make a decision, it can be somewhat wrong and you're not going to hurt anybody with it. But if it's anything you're going to be then promoting forward. You have to be accurate about the data and the results, and you could ask good questions. Can you give me the source for that answer? Can you give me references to where I can
refer others to learn more about the answer. If it can't do that, you're probably in the hallucination zone and you want to get out of there right Well, you know so, Jim, you talked about making sure that you check the words right. And part of the pitfall of adoptions also is moving
too fast for the organizations. I think that we really need to watch for is the organization ready and if not doesn't mean that you shouldn't adopt, but you really have to build almost like a safety net and an on ramp for the organizations to adopt. So meeting your organization, your user where they are, Making sure we understand where people are with the ability to use what we're going to put out that and whether or not we have provided the appropriate level
of handholding for this kind of work is very important too. You're reminding me of something. I'm so glad you brought this up because you put this concept of an on ramp, So you want to be careful at this pace at which you go. You want an on ramp, and you also want an off ramp, right, I think you want an off ramp, And what I'm referring to is that, And this is going to be an ongoing question.
Where do you actually insert the generative capability in your workflow. There's going to be some point at which it comes in, and you're going to want to know, of course where that is, and you're going to want to have a toggle to turn it off or on. So if you use it, for example, for customer service for chat bots, this is going to be probably one of the most common use cases is fueling a chat bot and
feeding it enough information from past conversations to know where to go. As the beauty of having an existing system is that you can just take the database of everything your chatbot has already said. Do some analysis on when the chatbot did a good job and when the customer was satisfied. Okay, that's the behavior we want to model and mimic. But then at some point, if it starts going walky, you want to have the off ramp. Oh, turn
it off like this, shut it down. It's just like an employee you maybe drank too much before going into work that night, and like, okay, you better go home, buddy, it looks like you're both well. That come up and comes out in any chatbot is so important. You really need to be able to collect the information when your chatbot start making no sense that you know, be able to really train that because it's a continuous training
as well. Right, Well, that's the thing too. So my buddy Usama Fayad, who runs the Institute for Experiential AI, I throw us over at the gym. He said, the whole thing about experiential means human in the loop, and he makes a really good point. And there is this narrative in the media that oh, AI it's going to just take over the business and run everything. No, no, no, no, you sure
don't ever want to do that. You want to have someone watching it all the time to make sure that it does what you want it to do, because it will move it blinding speeds, and so if you don't course correct in time, it's going to be with blinding speeds when careen off the tracks, right, Jim, Yeah, I think the chatbot example is a good segue into what I'm going to say augmenting. I'm going to say services and everything that's augmented means as a human it's in that loop. They have the
accountability and responsibility to have the final say. So maybe what we'll see I would love to see in the future is opposed to just chatbots doing it randomly and anonymously, that it connects you directly to a person with information that's augmented by the conversation, so the person on the other end can understand your situation much faster, and then through the generative AI get you to your resolution much
aster. But that human is going to if you will facilitate that process or you know in the sense that it's not completely blund But you brought up a good point too. There's an article I wrote where I started it with the tractor didn't replace the farmer, you know, and generative a. I won't replace your job, right, However, I am the article with the tractor
didn't replace the farmer, but farmers with tractors did replace those without. And I think that's a big distinction to what you said, Eric, because it's very specific. The human in the loop is a necessity. The power of the human and the loop, if you will, in terms of accuracy and impact, is going to increase astronomically, and that's going to now create what I'm going to say, what Baja was saying earlier. Am I going to ban papers from being written by generator of AI? Or am I going to
allow it to augment people who are writing papers? Different question, But that's kind of the thing you have to get through. Well, Jim was a T shirt. I have never been a T shirt. So I'm advocating for allow Jenny I to write paper. I think we're augmenting, and I think one of the cool okays augmenting. Yeah, you've got I wrote something. Read this and tell me what I wrote. Okay, Okay, that is not what I was trying to say. All right, correct it for me
a little bit. Let it correct it and I can say Okay, that's good. Now I understand where I made a mistake. Now I'm going to continue writing from that, not take it. I'm going to take what it corrected and make it my voice and my intention, and then have it read it back to me again. So it's a it's a give and take when you think about it, well, it's go ahead, Eric, Yeah you
made you brought up such a it's such a great use case. And I'll give an example of why, and then we've got a podcast ponus segment coming up here next. But think about in customer service, and this is what really fascinates me about these engines is how it decides to choose what it chooses, right, And that's why you have to try it, to learn and to kind of figure it out, because there is a method to the madness. Obviously, it's a computer system, right, I mean, it's got
rules and instructions and things that's supposed to do. But you think about customer service, and historically you have a customer service center, it's going to bring on bringing up your system, bringing up your record, et cetera. It's a very slow process and it's typically from one system where there are lots of other systems where customers have touched the organization touch points as they call them, on the call center, on the website, in a store, maybe I
wrote a letter or something. You never know. But these GENAI engines can, in a moment, like as you're getting to the customer service rep, whip up a whole bunch of stuff that's accurate and relevant and just give suggestions to the call center person. So now it's not this slow, painful thing. It's going real fast and the human can decide what to use not to use. That's why it's so powerful. Folks. Podcast bone a second, I'm coming up next. You're listening to Inside Analysis. All right, folks,
time for the podcast bonus segment here and a fantastic inside Analysis. We've been talking to Boo Ho We of FPT America's and Jim Wilt fractional CTO and distinguished engineer, and we're talking all about jen Ai use cases, what to do, how to get started, and as we were getting excited about that last use case, maybe about how I'll throw this over to you first. You know, one thing I figured out early is well, wait a minute, if this thing can write in German and French, and Spanish. I
bet it can write in HTML and JavaScript and C plus plus. And short answer is yes, it's your can and it can also absorb that kind of information too. So think about systems and information systems talking to each other in bits and bytes, and you know you get log files basically, which to a human if you know what the log file means, okay, now it's
meaningful. But the aale of log files you get, like thousands and thousands of log files you're trying to scan through, well ninety eight percent of them are the same, and then the ones that are different, Ah, that's when something probably went wrong. You can even throw that stuff at these engines and say hey, I got all these log files, what the heck does that mean? And get back some interesting answers. What do you think boja?
Oh absolutely, actually that was cybersecurity. I think leverage AI at least a few years ago. You know, there's been systems that knew exactly what you said. If you in a network management at all, one of the dreaded activity is read the log file. You know, you can monitor your network all day long, but if you don't analyze that lock file and read
the log file and figuring out what it's telling you. It's quite less and useless, right, So that's one of the applications that had been produced, and it basically general use gen AI or a large language model to read the log file and then just detect the differences the abnormality and call out and abnormality and then do adjuscency analysis and to understand what kind of of abnormality it is
and where the threat is coming from. So absolutely, well that was specific for cybersecurity, but you know applications application debugging, right, Oh my gosh, right, can you those things? Because it's brutal, I mean very brutal. I'll throw it over to Jim. Debugging has historically been one of the hardest things to do, and even with sophisticated technology. And I remember fifteen sixteen years ago interviewing a guy I always mentioned this. He had such
a cool name, Zohar Gilt, that was his name. He was with the company called Precise and they got bought by Idea. But what they would do is help you, this is a long time ago. They would help you do monitoring of system and application dynamics and why did this break? Why did this go down? Even back then, you would have interesting heuristics. You'd see CPU usage went down on a timeframe, and the application speed went
up, and all this stuff happened. But the human beings still had to parse all that together in his or her head to make sense of it, and then you would learn patterns of things. Oh okay, if this goes up and that goes down, it's usually because there's a disk failure or something. Well. Now, in the world of development, in part because of Kubernetes, but in part because of just the complexity of the cloud, we have containers and containerization, and we have all kinds of observability. I mean,
there are so many tools out there. I joke about this. There's a company Mesmo, which just sits on top of your observability data and helps you marshal it and go in the right direction. So there's so much activity that there is a meta tool now that sits on top of it. That's how much observability we have. But it's still up to the developer to piece
it all together. So this Gennai stuff can be tremendously powerful at absorbing massive amounts of these log files, finding some correlation between them, and at least getting you on the right track. What do you think, Jim, Yeah, there's definitely the ability to scan and look for anomalies or anything along the non what I'm going to say, traditional paths, because logs can be mundane and you're looking for that needle in the haystack can be very frustrating and to
broker it across multiple logs. What's really you're going to be able to start doing with AI is pitting multiple logs against each other. It could be through the time stamps, there could be through events. Whatever you can do. You need to train it enough to understand that this log is connected to this log this way. Then you're going to be looking at multiple facets of the same problem from multiple viewpoints, so that you can have better ways of understanding
what's going on that. The other thing is, you know with generative AI, you can write your test cases. It can look at your code, look for anomalies in your code. I think one of the things it's really interesting is it can read back to you what your code is doing. If you sometimes are trying to understand this code that you don't understand what it's doing, comments don't make sense. Who comments to their code anyway? Then you can make a correction to the code, or it can it augment you to
making a correction to the code. Then you get another model to look at that change that you made and explain what it's doing. Is it still doing what it was said before or is it doing something different? Is that what you wanted for an outcome? So we're going to get to where I'm going to say, we're going to leverage these tools increasingly more complex, you know, around the way that we can solve problems that we never would have attacked
the problems before before. I would look at one log at a time, or I'd get a room. I've done this and telcom where you put people, you know, a dozen people in a room for six weeks to look at all the logs to find out where did the error occur, you know, because the CEO was contacted by a text and you can't find it. With Generator AI, we can start training the models to look at things multifaceted so that we can actually start finding errors that we never would be able to
find on our own before because it just has more eyes. If you will on the screen, then we can put go ahead baha, well and you know, just on top of that, you need experience engineer to be able to read the lock and understand the lock and really recognize the issues when you come across it. Now, with jen Ai, you really can take all of that and abstract it and make it so that the people you'll need one
experienced engineer versus you need a roomful of engineer well to do it. That's such as Yeah, that's such a great way to put it, and it's exciting that I think we have to close on that line because that's brilliant. The future is here and it's exciting what folks will be talking about. Ha. We of FPT America's look up this company FPT. They're here. Huge. I mean they're a massive company. I have a wonderful origin story about
where they came from. Very interesting stuff. And of course Jim Wilt distinguished engineer and fractional CTO. We do archive all these events for later listening and viewing. Send me An Nemei if you want to be on the show. Info at Inside Analysis dot com. We'll talk to you later. You've been listening to Inside Analysis and now the Voices of ACAA was an exciting announcement. Want to hear NBC News or KCAA anywhere you go. Well, now there's
an app for that. CACAA is celebrating twenty five years and our silver anniversary with a brand new app. The new KCAA app is now available on your smart device, cell phone, in your car, or any place. Just search KCAA on Google Play or in the Apple Store one touch and you can listen on your car radio, Bluetooth device, Android Auto or Apple Car Play, catch the KCAA buzz in your earbuds or on the streets. Celebrating twenty five years of talk news and excellence with our case aac Just do it and
download it. K c AA celebrating twenty five Yeah. Hi, this is Graham Smith and I'm running for County supervisor and District three. I know you're getting inundated with messages about the election, so i'll keep this short. If you want a supervisor who won't waste your tax dollars, who will be a watchdog over the ten billion dollar county budget, who understands the danger of relying solely on warehouses to keep our economy going, and who isn't taking donations from
corporations or sitting politicians. Vote for me Graham Smith for Supervisor in District three. I've worked for twenty years in finance and business. I know how to grow our economy, create jobs, and make sure your tax dollars aren't being wasted. I'm not a politician. I'm just a resident of Samberdardino who's tired of unqualified leaders who make promises they can't keep. And I care about people, affordable housing, and honest and open government. We need to address our
challenges in a realistic manner. A county supervisor needs to pay attention. So if you're ready for a supervisor who prefers skills over sound bites, someone who wants to serve the people instead of themselves, I think we need to try something new. Vote Smith for Supervisor. You can go to Smith for SB dot com s M I T H F O R SB dot com. The election is March fifth. Mailing your ballot today paid for by Smith for Supervisor.
I'm Graham Smith and I approve this message. T hebot Club's original pure powd Drgo super Ta helps build red corpusoals in the blood, which carry oxygen to organs and cells. Our organs and cells need oxygen to regenerate themselves. The immune system needs oxygen to develop and cancer dies in oxygen. So the T is great for healthy people because it helps build the immune system, and it can truly be miraculous for someone fting a potentially life threatening disease due to
an infection, diabetes, or cancer. The T is also organic and naturally caffeine free. A one pound package of tea is forty nine ninety five, which includes shipping. To order, please visit t Hebot club dot com. T Hebo is spelled T like tom, A H E E B like boyo. They continue with the word T and then the word club. They can pleek. Website is to hebotea club dot com or call us at eight one eight six one zero eight zero eight eight Monday through Saturday nine am to five
pm California time. That's eight one eight six one zero eight zero eight eight to hebot club dot com. It's that time of year again, No, not the holidays. Medicare open enrollment and if you have questions about Medicare you should talk to the local experts. Paul Barrett and Associates. All of his agents are certified with plans that are accepted by most of the medical groups in our area. Call nine oh nine seven nine three oh three eight five.
Their service is free and after forty two years of the business, their agents are trained to help you pick the plan that's right for you. KCAA Radio has openings for one hour talk shows. If you want to host a radio show, now is the time. Make KCAA your flagship station. Our rates are affordable and our services are second to none. We broadcast to a population of five million people plus. We stream and podcast on all major online audio
and video systems. If you've been thinking about broadcasting a weekly radio program on real radio plus the internet, contact our CEO at two eight one five nine nine ninety eight hundred two eight one five nine nine ninety eight hundred. You could skype your show from your home to our Redlands, California studio, where our live producers and engineers are ready to work with you personally. A radio program on KCAA is the perfect work from home advocation in these stressful times.
Just type KCAA Radio dot com into your browser to learn more about hosting a show on the best station in the nation, or call our CEO for details too. Eight one five nine nine ninety eight hundred. I always hear from our clients who hired another firm that they wish they'd hire DNA Financial first. Don't have regrets about your IRS tax case. Just hire the best in the first place. One owed one hundred fifty thousand to the IRS and it's spent
thousands on another firm. We stopped the levies, negotiated a payment plan, and had their penalties forgiven. And while every case is different, cauarantee that we'll find your perfect resolution and get it done right. For a free consultation, call us at eight sixty six to zero one zero one five six. That's eight six six to zero one zero one five six. Then you can say DNA did right by me. There's never been a better time for men to be whoever they want to be. Yet it's never been less clear who
men really are. Guys Guy Radio, starring author and Guy's Guy Robert Manny, is coming to CACAA every Wednesday at eight pm. Whether it's relationships, sex, wellness for spirituality, join Robert as he interviews the experts and shares his insights on building a world where men and women can be at their best. Guys Guy Radio, Better Men, Better World NBC News Radio. I'm Chris Karragio. House Intelligence Chair Mike Turner says CIA Director at William Burns told
him that a ceasefire in the Israel Lamas war is close. Speaking on CBS News Face the Nation, Turner said the potential deal comes after weeks of negotiations with international leaders. I was just briefed by the CIA director who Burns Friday. Personally. He is the one who is conducting the ceasefire negotiations, and he believes that we're close, and I think that is very, very important
to accomplish. The Congressman said, it's important to get the deal done because there's hostages still being held by Hamas and there's a major need to get aid into the Gaza strip. This comes one day after Israel reportedly said yes to the framework of a potential temporary cease fire and hostage release agreement. The US official so the deal would involve a six week ceasefire and the release of at
risk hostages. The Supreme Court will issue at least one decision tomorrow. The ruling may be the case of former President Trump's eligibility for Colorado's presidential primary ballot. The opinion or opinions will be posted online at ten am Eastern. This comes after the Colorado Republican Party asked the High Court to act before Super Tuesday primaries. This week, Crews and Texas are battling strong wins as they fight
the largest and most destructive wildfire in state history. As of this morning, the Smokehouse Creek fire has wartched over a million acres across the Texas Panhandle and into Oklahoma since Monday. More than five hundred structures have been destroyed, and the fire is only about fifteen percent contained. Critical fire conditions are expected to continue today, with the National Weather Service predicting wind gusts of fifty miles an
hour, with red flag warnings posted across the Central Plains. NASA and SpaceX are hoping for good weather for tonight's launch of the Crew eight mission to the International Space Station. Unfavorable conditions prompted officials to scrub yesterday's scheduled launch attempt. The SpaceX Crew Dragon Endeavor, atop a Falcon nine booster, is set to lift off from Launch Complex thirty nine A the Kennedy Space Center at ten fifty three pm Eastern. I'm Chris Karagio. NBC News Radio, NBC News on
CACAA Lowland. Sponsored by Teamsters Local nineteen thirty two Protecting the Future of Working Families Teamsters nineteen thirty two dot org. The election is March fifth. Ballots are in the mail. Mail them in today. Now there's something to vote for because there's a new Marshall in town. Vote Derek Marshall for twenty third