AI, Privacy and the Art of Keeping It Simple with Gary Murry

Perry Robinson: Welcome, Gary. How are you doing today? Okay, for folks that don't know, I'm Perry Robinson. I'm RocketDoc's CEO and ⁓ one of the podcast hosts for Ground Control. ⁓ And this Bryan that's over one of my directions here. Yeah, Bryan's co-host. Bryan say hello. Gary Murry: I'm doing well, thank you. Bryan Jenkins: Yeah. One of the ways. Hey, everybody. I'm Bryan again, back again. I to customers on a daily basis. My title is now fancily down here. So if you want to look me up on LinkedIn, I'm there. Looking forward to talking to Gary today. ⁓ Gary Murry: Yeah. Perry Robinson: All right. Gary, ⁓ you and I met I think now a little over three years ago, yeah? Gary Murry: Yep, three three and a half, yeah. Perry Robinson: And to be quite honest, this is the first time I've gotten to see an unblurred background ⁓ of where you take your calls and work from. so remind me again, would you study in undergrad? Gary Murry: under ⁓ so my degree's in mathematics, mathematics. ⁓ Perry Robinson: Yeah, yeah. Except mathematics, right? And ⁓ this looks like a mathematician's math cave. Is that a pretty fair description? Gary Murry: Yeah, yep, that's pretty reasonable sort of statement. Yeah. Perry Robinson: Yeah. But these days you're not a professor at San Jose State teaching mathematics. You're doing something completely different, which I think is probably on the title below your name there, right? CTO, right? So how'd you go from math into the world of technology? Gary Murry: ⁓ so basically, you know, living in Silicon Valley, right? Everyone has some it seems to play with Thai Tech. And, know, right after getting through college I got a ⁓ a contract position at Apple and you doing ⁓ quality assurance with Apple, and you know, ⁓ met some other fellow mathematicians who ended up being there. and ⁓ so it you know, you just in high tech you meet some super smart people and eclectic, as again, my background probably kind of shows, which keeps interesting, right? And so that's kind of right. And back then we didn't really have computer science degrees, right? Your computer science degree was a math degree with some sort of computer science studies. Perry Robinson: Yeah. Gary Murry: So I my first study was Fortran. So I didn't have punch cards, but it was it was a mainframe. Perry Robinson: The eighteen nineties were a different time. It was tough without the phones and the Yeah. so Apple, ⁓ give take what year was this that you're doing that that you know initial work out of out of undergrad? Gary Murry: Eight nine or so, somewhere around then. Perry Robinson: Early days on on the Apple side, what for for the for the Silicon Valley natives and the ones that like to geek out, right? So different campuses, different points in time in Apple's lifespan, where was this lifespan for Apple at that in that year? Gary Murry: This was, you know, the Mac ⁓ right, ⁓ I think Mac had recently made their big splash with the ⁓ the large colored versions that they were sending to all the colleges and stuff. ⁓ for those who know the geography, right, infinite loop was the the the place to be. We were across the street from it 'cause we weren't the people to be there. ⁓ but this was well before the donut got out there in ⁓ Cupertino. So yeah. Perry Robinson: Yeah. Yeah. And ⁓ and these days you haven't gotten that far away from Apple because you're married to someone who works at Apple, right? Gary Murry: She's got my wife's been at Apple for mm about five years now. Yeah. Perry Robinson: Yeah. So the the that circle, that infinite circle is definitely a an awful thing. All right. So yeah, so rocket docs, right? I know how how you kind of came into rocket docs, but I don't think any of the other folks that are listening might have that story. So what brought you on t into rocket docs and what made it interesting? Gary Murry: There you go. So RocketDocs was an interesting creature. ⁓ right. It was ⁓ in a space I didn't know much about, right? But you know, ⁓ RFPs originally was the the main focus. ⁓ and you know, ⁓ obviously rocket docs is expanded beyond just RFPs and you know, answering any sort of questionnaires. ⁓ but it was an interesting sort of position because They were on the cusp of going from this older technology to this newer technology was coming on the horizon that was really making a big splashes. Everyone, I'm sure, listening knows, right, with the the impact of Gen AI and all of this new technology coming. It was a really interesting time to be on a company that's been around for thirty years and seeing how they can, you know, to transform with this new technology and Perry Robinson: Mm-hmm. Gary Murry: kind of capitalize on their experience with all this new sort of ⁓ almost green field. Right. And we we made some choices early on that we were excited about because well, we weren't sure if they were gonna be the right choices, but then, you know, a year or so later we saw other companies kind of following suit with this whole private AI sort of approach and, you know, security first. ⁓ it was it was really neat to just see that Transformation. Perry Robinson: lot of changes. Yeah. And I think a lot of the experience that you'd built up in different companies along the way, really the the common factor for you was going in and and figuring out ways to make sure that the technology that was being delivered was being delivered in a way that worked. Right? Gary Murry: ⁓ I mean right, ⁓ yeah, I've been fortunate enough in this valley to work on everything from slot machines to playing with viruses to, you know just every day sort of, you know, helping to get ⁓ large data sets to work and it's always been fun to try to figure out how can we do that better. Absolutely. Perry Robinson: Very cool. one last ⁓ kind of ⁓ question. So you have a twin brother, ⁓ you already mentioned your wife, you've you've ⁓ you've got diverse interests, ⁓ I think that we we might say, but what what does one find Gary doing? outside of his working hours. What's what's the thing that that makes yeah, what what what do you find yourself enjoying the most, Gary, when you're not working? Gary Murry: I find myself turning to most often just tinkering. Right. ⁓ again, ⁓ we live in a time with such availability of technologies and with things like, you know, well, microcontrollers, like ⁓ Raspberry Pis, etcetera. You know, there's just so many different ways you can put these different sort of Legos together and do some really cool things with, you know, ⁓ I've taken some of my son's Toys that have RFID tags in ⁓ and use those to control lights. Right? So put a little RFID reader in a sensor and you know, be able to turn on different color light effects because they put their Skylanders on a particular portal. So that Perry Robinson: Yeah, and and with with a a graduating senior, a graduated senior now, right, going off to University of Chicago to go experience the cold for the first time, right? You're gonna probably be doing a bob a lot more tinkering, I'm guessing. Gary Murry: Yes. Have more free time to tinker, yep. And then probably to send them to Chicago. Yeah. Perry Robinson: Yeah. There you go. And and one of the things I was alluding to as well is that you spend a little bit of time in inside of a dojo piece building. So yeah. Gary Murry: Yes, I've I've been teaching martial arts for forty some odd years. and so I've ⁓ teaching tempo, yep, and then also been known to teach fencing. I'm actually a maestro in in fencing also. Perry Robinson: actually just a maestro in fencing also and that's not just inside of the martial arts. We're talking about classical fencing like we see with the columns for the face and yeah, the white pajamas. Yeah. Okay. All right. Well I think Gary Murry: No, they're talking fancy. Mm-hmm. Perry Robinson: background wise, like we've gotten to know a little bit about you. Let's find out what you think about work related things a little bit more. Bryan, let's see. This is the part where we can kind of turn the show towards seeing if we can ⁓ if we can stump Gary on topics. Gary Murry: Not hard on you can do it. Bryan Jenkins: Yeah. Very first I I think you gotta give everybody some sort of ⁓ I guess instructions on how to manage their time so well. You're fencing, you're doing like everybody I talk to is busy in the industry. Maybe you can shine some light on the how to manage to be good at all these different things. Perry Robinson: Okay. Gary Murry: give up sleep, right? No I mean actually in in all sincerity, right, ⁓ the people I seem to find have the least amount of time are the ones who are retired. It's if you have a busy schedule you're forced to keep to a schedule. Bryan Jenkins: Give up it's an extra twelve hours. Yeah. Mm-hmm. That's true. Yeah. Perry Robinson: Time management by Gary, just be busier. Gary Murry: There's big Bryan Jenkins: She's busier. De another hobby. That's Tinker More. Perry Robinson: More time will be created. Just don't sleep. Gary Murry: Color code your calendar. Bryan Jenkins: That I bet is something that is very helpful. So if people aren't doing it, they should do that. ⁓ so we have a new release coming out, Gary. a bunch of new features. ⁓ do you ⁓ do you prioritize like the features that you're that you're gonna push? Because I know there's there's so many. Gary Murry: Yeah, nope. ⁓ so one of the things we often try to do is we work closely with our ⁓ customer success managers to ⁓ you know, see what common themes we're hearing from our customers, ⁓ to help ⁓ you know, figure out one what can give us the most bang for the buck. ⁓ and you know, help the most people as as best as possible. ⁓ and the other thing too is we always try to put in some features that will lead us farther down the road. Right, that may be addressing thoughts, concerns that customers just haven't even run into yet. ⁓ but you know, so it's a balancing act between that. I mean there's you know, keeping the customers happy as much as possible during the day, but then also leading them to something new and exciting in the future. Bryan Jenkins: Yeah, absolutely. I mean i you've you've seen stuff all the way from Apple in the eighties. You've you've watched the you know, the internet come around, you've watched has anything been faster than AI? Gary Murry: You know what? When the internet first hit, we were thinking, God, this thing is going quick. This thing is really changing everything and I it was hard to imagine. And AI's hit and it's just changing things even faster. ⁓ and ⁓ having seen this be something like this before, I I hesitate to say it's hard to imagine anything going faster, but who knows, right? It is ⁓ McQuinet ⁓ Bryan Jenkins: Who knows? Yeah. Definitely. add different universes in there. That'll make it faster. Yeah. ⁓ Perry Robinson: Computers. Yeah. Yes, please travel. Which we also thought wasn't gonna happen all that soon, but you know. Bryan Jenkins: We're there. Basically. I saw like ⁓ we've got like a few new features that I was pretty excited about that I know customers have been asking about. Simplify Excel. ⁓ I also saw we have ⁓ I was actually giving a demo this morning and I saw three or four new features that got published last night. So do you wanna give like a quick rundown on some of the stuff that you're most excited about? Gary Murry: I mean, So well, so you you'd mentioned ⁓ simplifying Excel. I mean, one again, going back to listening to our customers and hearing some of the ⁓ pain points they're running into. ⁓ you know, ⁓ of the mm common sort of pain points we were hearing is customers were doing these web portal sort of interfaces and would often download these Excels ⁓ would ⁓ you know give what the questions for the web portal were gonna be. Bryan Jenkins: Mm-hmm. Gary Murry: The thing that we were finding which was odd is that for really no great reason they were ⁓ the the download of these Excel spreadsheets were complicating them for I mean no real reason because the customers were just using them as guides to re-answer the the ⁓ the actual web interface. And so one of the things we took approach to is to just simplify that, right? of the guiding principles we often will use is If I give a human a questionnaire or an Excel spreadsheet and they can't figure out where to answer the questions, you know, that's a problem. And so one of the things we do is help to make it easier for them to understand, okay, there's where the question answers pairs will be. And it's much easier now for you to answer these ⁓ web interfaces ⁓ with this simplified form. And so and you know it It wouldn't have occurred to us necessarily, but we heard our customers, and our customers are telling us that this was, you know, a pain point. And so and it turned out to be several customers. So it was super exciting to see that come out. the other thing that ⁓ ⁓ and exciting to see come out is the ability mark questions from within launchpad. ⁓ one of the ⁓ Common sort of principles we've also tried to follow is let people work how they already naturally want to work. And with Launchpad, which you know, again, I'm sure a lot of our customers already know, ⁓ you know, a plug-in for Word, they're already in Word. ⁓ And it's a much more natural way for them to interact with the document they're having to fill out and say, hey, here's my document. ⁓ you know, just look through these pages for me, look through these sections for me. And answer, you know, find all the questions for me this way. Right? I don't want to have to go through some other thing to do it. I'm already in this document. ⁓ and this gave our customers a lot of flexibility because one, they can, you know, work at their own pace, right? It's it's not quite as daunting. ⁓ but two, they could also fine-tune their their answers by saying, Okay, for this group of questions, I already know this is really related to product. A B C, let me really pull my knowledge base focused on that product. And so solved a lot of problems for our customers pretty darn quickly. and probably the last big thing that came in with this release that we're kind of excited about is ⁓ with SharePoint. you know, ⁓ one our big concerns always at ⁓ RocketDocs has been to be as secure as possible. be as you know being able to let ⁓ your your IT team trust us so they don't get in your way to use the tool. They don't get in your way to do your your job. ⁓ and you know unfortunately right the IT teams they're they're the ones who are responsible for making sure things are secure. And so sometimes they're seen as, you know, blocking you. Well we made it easier for them to to feel comfortable with the security. And we be able to use the ⁓ principle of least privileges, right? And so with the sh new SharePoint update, we give the IT team the ability to lock down even more what's shared on SharePoint with RocketDocs ⁓ again, it makes the review process with the IT team go much because as soon as they see what the what it is and hear, you know, what it does, you know, they they half hour meeting goes down to about a five minute meeting. So, you know, that's always great. Bryan Jenkins: Yeah. No, absolutely. And I know reminders was another thing that that popped up. Like I know that we skimmed over, there's a bunch of gems in this ⁓ in this release, but I know mm-hmm. Gary Murry: ⁓ Thank you. Perry Robinson: So much into it. Gary Murry: So so with reminders, absolutely. Thank you for reminding me. ⁓ Giv the ⁓ but ⁓ you know, the the ability to be a lot more customizable on reminders. ⁓ right, one of the things we were hearing from customers is, you know, I've got a month before the project's due. I wanna remind my people every week. But, you know, as we get closer, we need to all right, now I need to that that's timeline's getting impacted. I need to up that. I'm starting to get nervous. ⁓ Bryan Jenkins: Ha ha good. Gary Murry: The ability to automate those reminders really helps. And then as you know, right, we curate the your knowledge database. As things get out of date, you know, people kinda you know, you've got all sorts of other things you're doing. Sometimes it's easy to forget that, ⁓ I you know, I I meant to do that knowledge database entry and I I didn't do it yesterday. We also now give the ability to periodically remind people automatically ⁓ when things are out of date. ⁓ again, everyone's busy. This is just a a convenience for the RFP writers and the SMEs so that they can just you know, ⁓ yeah, they just ⁓ that's in the background. Give me five five seconds here. Yeah, that's right. That we're good and move on. So yeah. Bryan Jenkins: Gary, can you talk about like, I mean, your days at Apple, right? Like you lot of ⁓ lot of companies, right? You can either add a lot of things or you can start to take things away. And I think sometimes there's not as much weight in things that you decide not to add or not put in. So like here at RocketDocs Docs, like when you're getting requests from ⁓ on one off things, and you're also getting a lot, obviously a lot of requests for different types of things. Like, how do prioritize not implementing so much that the system gets confusing because it could be easy to do. Gary Murry: Yeah, absolut ⁓ absolutely. Yeah, yeah. No, I'm right. There is this natural inclination, especially if you have an engineering mindset, of let me throw everything in there. And you know, 'cause 'cause we believe people sincerely want and believe they need it. but to your point, absolutely it you know, it becomes more and more there's how many different ways can I do the same thing or or you know Perry Robinson: Mm-hmm. Gary Murry: Do I really need this one particular flavor of Dunes? So what we try to do is try not to actually solve the one particular flavor. Right. So the example I like to use is we've got one customer who wants, you know, their text to be blue, and we have another customer who wants their text to be purple. Well, really what they want is the ability to customize their text color. Right. And so whenever possible, you know, as we get these requests, if we can generalize them and actually make something that's even more flexible, more powerful for the ⁓ the customer, that's really the right way to go. you know, and that's usually what we try to do. Now again, obviously the complexity of giving a color picker versus just a blue green text is a little more ⁓ a little more involved, but you know, it's you know it's more sophisticated. Right. And again with sophistication comes flexibility and that gives you a little bit more power. So we try to generalize as much as possible. Bryan Jenkins: Yeah, I think there's something to that, right? Making things simple and easy for people to use. Sometimes reminds me of the Henry Ford quote, right? Like people if you ask people what they'd want, they'd say a faster horse. Right. But they don't even know that a car is something that is an option. And you being a CTO, being at the forefront of tech, seeing kind of where things are gonna go in the next couple of years, like how do you put weight on what people think they want versus what they might might want in a year versus two years versus three years? Gary Murry: ⁓ absolutely. Well, and and nowadays, I mean, I think I might actually even change your time frame from two months to six months to a year because just how rapid things are happening. Absolutely. ⁓ you know, part of it it comes down to, you know, there is the pain you're feeling today that someone may be feeling today. ⁓ you know, can I solve that pain in a week? Bryan Jenkins: Who knows if we can predict that far? Yeah. Gary Murry: With the idea being that I can give you a much more powerful and long term solution, then I think you know both parties would probably agree that waiting a week isn't a problem. Right. And so we're constantly trying to balance how long will it take to get to an even better solution? How long will it take to get to a solution that won't become obsolete also in the space of, you know, two weeks. ⁓ and so that's really the big balancing act is, you know. ⁓ How long does it take to get to something meaningful? ⁓ one of the concepts of Scrum is that every release you try to release meaningful software, right? With the idea being that you have these gradual steps to a more larger story. and so that's usually the approach we try to take: is, you know, we want to get to this end goal. We need to figure out what are some meaningful steps we can take. And, know. ⁓ Hopefully those meaningful steps will also alleviate maybe some of the these other requests that are coming in by again generalizing them and leading towards a long term solution that will be even more valuable even if you don't necessarily realize where that's going yet. Again, our analogy is absolutely right. A faster horse versus an actual car. Yep. Bryan Jenkins: Yeah. People couldn't even envision a car, right? At at some point. And I think that's how some people feel with AI and like for somebody listening right now that is like, you know, working a proposal job, like, how do you how would you recommend they like not be uneducated or or how do they find what's going on, right? Like in the course of their day to day lives, how do they make sure they don't fall behind? Gary Murry: Absolutely that's that's a good good question. ⁓ so I'm a firm again, as I said earlier, I you know, I in my spare time like to tinker. Tinkering is playing. I'm a firm believer in technology. You you want to play with technology, right? ⁓ there's so many different arenas now that AI is available to play with. And you know, this helps with several things. One, it gets you so you're comfortable with AI, right? We I I frankly I think playing is a super natural important method for people to learn. ⁓ because it makes us relaxed and more engaged with something and it makes us less concerned by the AI, ⁓ you know, by dealing with AI if we're just playing with it. ⁓ but it also gives us the ability to start exploring and one the the things that I often have to do when I talk with customers is to discuss the current state of AI, especially generative AI, and then what's science fiction. Right. And and ⁓ playing AI gives us that ability to kind of go, all right, I I understand a little bit more. I understand what it can't do yet. And you know, and that in some ways that's even more important because while imagination's super important, Perry Robinson: Mm-hmm. Bryan Jenkins: Mm-hmm. Gary Murry: And I do think imagining where AI is going is still in a little a prognostication. I think understanding where it is is also really important to, you know, so we are not so concerned by it, right? I mean ⁓ it it it is something to pay attention to, but it's not something to ⁓ right? Just like you know, the I I think probably going back to cars, there was probably some fear ⁓ of these giant metal machines running around that have no sort of survival instinct like a horse does. I mean it's so I think absolutely playing with AI wherever you can, however you can, is important. And that's what I've been telling my my son is, you know, just play with it. Just see what it can do. See under see what you can imagine it doing. Perry Robinson: What can it not do today? Well, at least in the art space. Gary Murry: What can What can't it do right now? So again, one of the guiding principles we've been doing is, you know, if if I can't give it to a human and they can't figure out what a question is asking, AI's you know, i i it it it probably won't figure it out either. of the the biggest limitations I'll say ⁓ as as can't do limitations I find with AI is ⁓ It can be incredibly convincing and incredibly wrong at the same time. It'll walk you down this rose path and then just makes a whole different sort of conclusion, right? Again, a very common one you'll hear nowadays is, you know, I've got my car, it's dirty, ⁓ I need to get it washed. ⁓ the car wash is only down at the end of the block. Should I walk there or should I drive there? Right? ⁓ Bryan Jenkins: I saw that one. Gary Murry: Very common sort of ⁓ logic question. And again, again, I can give it to a seven year old, and a seven year old will, you know, be able to answer it with no problem. ⁓ but that leap of right, not saying, Well, it's close enough to just walk, right, eludes the AI still. ⁓ you know, there's certain things that are we still think are common sense, and I think that's really where com where common sense is where AI starts to break down. You know, it is getting more and more clever. ⁓ but it's still yeah. The other place that ⁓ AI breaks down a lot is math. Just because AIs currently generative AIs are mostly predictive models and they're just predicting what word comes next. And in math it's just the the the nature of infinity is hard to guess what comes next. Perry Robinson: Where should people pay most attention when they're looking over the work that an AI system has created for them? Where where you said that they're that it's good at making things look really good. It'll walk you down that primrose path. Yeah, how we talk about human in the middle, but what should they actually be doing as the human in the middle? Gary Murry: Path, yeah. That I mean so this is this is where it can become fairly taxing on an individual to truly to truly follow this path and make sure that no logic leaps occur that are perpendicular to the direction you're going. that, you know. So ⁓ I would look at usually is my experience has been it's usually that last leap. that seems to be the where it goes astray. Right. Good, good, good. And now go do that. And you know, I'm not quite sure. I mean, there's this idea that the farther you get down a prediction, the more uncertainty you've introduced through each of those sub predictions that you've added. So it makes sense that that last logical leap is where it would break down. So what I often will do is I will start from the end of the Perry Robinson: Mm-hmm. Mm-hmm. Is it? Gary Murry: Right, what its final assertion is, and then work backwards to see if it's working correctly or seems reasonable to me. ⁓ so yeah. Perry Robinson: that one so so the last leap right so thinking about places where there are a lot of systems that are designed today ⁓ the first place they're supposed to look is inside of relevant information ⁓ But a lot of AI systems are connected to lots of other information, right? And you know, is sometimes that last leap when it can't find a good answer inside of the relevant context and then goes to other sources? Gary Murry: So so this is so this is what's referred to as hallucinations ⁓ in in many cases that you're looking at. ⁓ AI systems really, really, really want to answer your question, really want to solve your problem. ⁓ and so oftentimes we'll just make up stuff so that they can get to the end and say, I've solved your problem for you. if if ⁓ For example, as we do in RocketDocs Docs, right, we use a rag-based approach so we can say, hey, here's the information I want you to work off of. But the really important part from our side is we also tell the AI, here's the information to use. But if you don't have enough information, tell me. Don't don't make up an answer. We give the AI permission not to answer. And I I mean, when we first implemented our Gen AI, we didn't have that. Perry Robinson: Mm-hmm. Gary Murry: And we we absolutely we're seeing AI trying its best to answer and making up stuff. But as soon as we said, hey, you know what? If you don't have the information, I understand. That's fine. Don't answer. And that's the same thing you would do with a junior RFP writer, right? You would tell a junior RFP writer, hey, here's our knowledge database, or here's our Excel spreadsheet, which is often what we hear. ⁓ try to answer this. ⁓ but understand not everything's in there, so if you don't find it. Perry Robinson: Mm-hmm. Gary Murry: Just let me know and we'll add that. And that's the same permission we give to our AI that, you know, the other thing we often found that's even better is, you know what? You've already answered the question before. You already have knowledge in your knowledge database that's correct. Let's not even generate anything. Right? Let's just use that. And that's really the nice thing is, you know, Joe from accounting already said that was right. Right? And so we don't even go to the gener generative response yet because it makes no sense, right? It it it it is not it's an unnecessary step at that point. Bryan Jenkins: That's back to that simplicity we were talking about, Gary. Sometimes like it's when you really think about the problem, you're just trying to solve the problem, not introduce a bunch of features. So Perry Robinson: Mm-hmm. Gary Murry: Exactly. Yeah. And if the actual problem is I want to answer correctly, then we've done that in the past. We're good. Absolutely. Yeah. Absolutely. Yeah. Bryan Jenkins: It's as fast as possible. Yeah. Without a review cycle that'd be better. So yeah. So if I had to put a big theme on this release, it feels like keeping content healthy was the main driver behind it. Like where do you see it going from here? Gary Murry: So absolutely, yeah, keeping content healthy is, you know, ⁓ is an often reoccurring thing for us, ⁓ and curating healthy. ⁓ I think really where the future with that in mind is going to be again AI, right? More and more having some of this curation, some of this keeping it healthy ⁓ done through the AI, but the idea of being Not that the AI necessarily does it for you, but really the AI percolates up to you the things that need attention. ⁓ you know, there's very few p in talking with customers, very few things I've heard from customers where they're want to be completely hands off and let the AI take care of it completely. What they really want is AI to help them help them know what to look for, what needs attention now, what needs attention soon, right? And so that's really where I see us going. with this in the future. Perry Robinson: So Gary, ⁓ first one for ⁓ favorite of all time. What is your favorite technology of all Gary Murry: Favorite technology of all time? Raspberry pies. Perry Robinson: Raspberry pies. Why? Gary Murry: ⁓ y there's they're a real world interface to program to ⁓ a pro a computer. So I can put them around the house and have them monitor hair air quality, the garage doors open, you know, other sorts of things. I like that real world world interface. Perry Robinson: Love it. Okay. you're ⁓ things that you would ⁓ use Claude or Chat GPT for. What are what are the things that you think like those big solutions on the AI side are perfect for this? Gary Murry: You know what? I th I I absolutely think they're good for brainstorming, right? ⁓ for ⁓ things where I want to bounce some idea ideas off ⁓ you know, really quickly and have a a a conversational sort of approach. I absolutely think they're good for. ⁓ also I mean they are super useful for coding, right? It's having a junior coder that you can just say, Hey, you know what? I don't want to write another program to read a a C C S V file. Perry Robinson: Mm-hmm. Okay, flipping the question around, would you never, ⁓ ever ⁓ Claude or Chat GPT for? Gary Murry: ⁓ anything I want to keep private. Perry Robinson: Why? Gary Murry: ⁓ so there's two concerns. One, I don't I j I just don't know how they will use that data. And two, and to me this is actually a more insidious and a deeper concern for me, is day I talk about this thing that ⁓ I'm a little concerned about privacy, and then some other day, some other time, some other computer I talk about this other thing that I really don't want known. And AIs are really good about connecting and joining information. And now maybe I've created some information I really don't want shared. Right? And and it's not like I lead an exciting life. But, you know, I do value my privacy. Bryan Jenkins: That's debatable. Perry Robinson: Yeah. Bryan Jenkins: You're like a fencing champion and a right? In a tinkering with raspberry pies and seems pretty exciting to me. You're modest. Perry Robinson: Teaching martial arts, multiple black belt. Gary Murry: This Perry Robinson: Fakers doesn't sleep. Gary Murry: Let the AI connect all those dots and see what happens. Yes. Perry Robinson: Shibumi. Gary Murry: Yeah. Bryan Jenkins: Yeah. So rapid fire question from me. Gary, why'd you pick a rag model versus any other model? Gary Murry: It it was the cutting edge for reducing hallucinations. I think I still think it's cutting edge for release reducing hallucinations. Bryan Jenkins: Do you think teams are are thinking about the fact that maybe they're replacing their efficiency with AI, but they're actually just replacing the fact that they're finding questions now with reviewing more information and then like what are some of the problems that you think people run into using AI like Claud or open AI? Gary Murry: again, I think that double checking things is probably the biggest problem they run into when dealing with these ⁓ other ⁓ with ⁓ generals general AIs and right, and then also ⁓ with things that are, you know, qualitatively the same, even if they just changes the words a little bit. And now I've gotta spend the mental energy to say are these two sentences equivalent or not? Right. So there is a lot of mental energy that's going into those sorts of things. Bryan Jenkins: Professional skills wise, like for an average worker person right now just trying to get better at their job and, you know, be relevant in the next five years, what what skills do you think they should be focused on from the technical side? Gary Murry: I you know, ⁓ you cannot avoid having the skill of knowing how to write prompts. Start dealing with, you know, AI in general. Bryan Jenkins: Where where do you feel like they go to to find out how to write a good prompt and what goes into a a good prompt? 'Cause I think people might not fully understand what what you mean by that. Gary Murry: Yeah. ⁓ nowadays, I mean there's so many YouTube videos that ⁓ I think are I mean that would that's often my first go to is you know what I've heard about this new thing. Let me go see if there's anyone, you know, and and there's I've never had something that I couldn't find a YouTube video on. And those are often, you know, type it along while they're doing it. Absolutely. Yeah, that's probably the easiest way nowadays. Bryan Jenkins: Earlier you referenced like playing, so that seems like the best way to go about it. Go learn about it and then try to actually put it to use. Actually make something or prompt on your own and try to get it to do something that you want it to. Gary Murry: Yeah, the YouTube the YouTube is just to let you learn the instructions on how to play the game. And then yeah, absolutely. Just play the game. Bryan Jenkins: Go play, yeah. Perry Robinson: I think that's Gary's takeaway from this one. Go, go play, go tinker, go figure out new cool ways to make things work. Gary Murry: And and figure out how to break it. That's always fun too, right? I mean, you know what the limits are. That's right. Trying to break things is how you figure out limits. Perry Robinson: But not on customer environments. Bryan Jenkins: Ha ha Perry Robinson: Yeah. Well Gary, thank you for joining us today. ⁓ we appreciate having you on ground control and excited to see when you build a little miniature rocket for the next time that we have you on the podcast so we can see ⁓ rocket ship taking off. ⁓ Gary Murry: Thank you. It was my pleasure. It's it was fun chatting. Bryan Jenkins: Yeah.