Subscribe
Share
Share
Embed
Branding is a powerful tool that creates lasting impressions on individuals. Although people may only see ads and logos, creating a successful brand takes time, effort, and creativity. We Built This Brand is a podcast that delves into the process of building a brand by interviewing founders, marketers, and creators who have successfully created powerful brands. Through this podcast, listeners will gain practical applications and a better appreciation for the brands they encounter every day.
Chris: Welcome to We Built This Brand. I’m your host, as always, Chris Hill and today I’m here to talk to the founders of a local Knoxville community, KnoxDevs. And we’re going to talk about their latest conference, Devmoot. In addition to that, we’re also going to be touching base with Jeff Prosise to learn about his work as the chief artificial intelligence officer at Atmosera. And if none of that sounds like it has anything to do with branding, you’re wrong. We’re going to talk about why.
Because there’s a lot of branding and there’s a lot of marketing that goes into starting a local community building and creating a conference, as well as starting a business and also the personal brand that comes along with writing books, communicating with people, and promoting that content. We’ll also, of course, dive into AI. We’ll talk about some of the pros and the cons and the challenges of artificial intelligence and what applications that has to marketing professionals. So, I think you’ll get a lot out of this interview. It was a really fun interview to have all the way through, and it’s pretty neat because I get to interview three people, not just one on this episode. So, go ahead, check it out, and can’t wait to share this episode of We Built This Brand with you. So, let’s get it rolling.
Chris: Well, hey there everybody, this is Chris Hill, and today I’m coming to you from the Relix Theater. Is it the Relix Theater now? Actually, now that I say that I like—
Cody: Relix Venue?
Chris: I’ve always heard it, like the Relix Venue.
Adrian: Relix Variety Theater.
Cody: That’s it. That’s it. That’s it.
Chris: Relix Variety Theater? There we go. Yeah, we’re here at the Relix Variety Theater in downtown Knoxville where, in just a couple of weeks, the KnoxDevs community is going to have their first-ever conference, Devmoot. So, why are we here today? Why are we talking about KnoxDevs and Devmoot for marketing and branding?
Well, they started a community, they’ve built that community, and now they’re launching their first conference. And I think that’s a really cool thing to explore. So, we’re excited to have with me today, Cody and Adrian, who are the founders of KnoxDevs, and we’re going to be talking with them a little bit more about this conference, Devmoot, and KnoxDevs. So, with that said, let’s dive into it. Cody, why don’t you introduce yourself?
Cody: All right. Well, like Chris said, I’m one of the co-founders, along with Adrian, of KnoxDevs. I’ve been in software development for a long time, 20 years, 25 years? I don’t know. Currently, I’m Chief Architect for a Canadian company, mdf commerce, and then, you know, running KnoxDevs with Adrian and a bunch of other talented people.
Chris: Awesome. Adrian?
Adrian: Yeah, so for me, I just got involved with KnoxDevs, started back in 2015. Somehow, it’s been eight years. And we’ll talk a little bit more about the history here in a little bit, but we’re excited about this upcoming conference.
Chris: Yeah, awesome. Well, it’s great to have you guys with me today. And let’s talk about KnoxDevs. What is KnoxDevs?
Adrian: Sure. So, KnoxDevsOps is a grassroots 501(c)(3), a nonprofit corporation. It was basically started, and the goal is to just, this very simply, at a grassroots level, raise the level of technology in Knoxville, providing opportunities for the people in tech community. It’s mostly developers, but we have entrepreneurs, we have network administrators, database administrators, project managers, all kinds of people. And everyone in tech is welcome and have represented by all kinds of different jobs and roles. And we try to fill that niche. There is also the Knoxville Technology Council that handles things at the corporate level and we work together with them to kind of handle both ends.
Cody: Yeah, I think really, a big part of what we do, and starting in 2015, was organizing the community. You know, there was already, you know, somewhat of a community, but it was a little bit disorganized. It was unclear what existed and what didn’t, and so I think we just kind of stepped in and tried to clarify and create an umbrella for a lot of the existing meetups and technology leaders of the community.
Chris: Yeah, and it’s neat that you all have done this from, like, a grassroots effort. It’s not an organization that came in and, you know, just said, “Hey, we’re going to try to start a group here in Knoxville.” You all decided to do it on your own.
Cody: Yeah, yeah. In fact—well, it’s interesting we’re doing a conference. It may be our last. I don’t know if we’ll do it again.
Chris: [laugh].
Adrian: [laugh]. We’ll see.
Cody: Yeah, it’s our first. It’s exciting and fun and it’s been good so far. And I think it’s going to be wonderful, a wonderful event. But I, you know, Adrian, I came at this kind of different ways but at the same time back in 2015. For me, I was at CodeStock, actually, which is a local conference that’s, you know, been really successful for many years.
But I was at CodeStock and recognized, you know, the need for an organization because we had a pretty good, you know, size of developers and all that, but you know, again, lacking the centralization. And it kind of—to me, it was kind of a lightbulb. And I was talking with Greg [Ostermeier 00:05:15] at CodeStock, and he said, “You should talk to Adrian.” Whom I’d never met at the time. And he was actually doing a talk at that CodeStock, so Greg introduced us. And Adrian was kind of thinking the same thing and so we just combined efforts. And you know, here we are, eight years later.
Adrian: Yeah, so Cody at this conference. So Greg, talk to him about this thing called Slack, this new tool, it was kind of cool at the time, and a lot of other tech places were using it. And so, Cody went ahead and just registered KnoxDevs.com, KnoxDevs.org, and signed up for a Slack account.
And so, by the time CodeStock was over, I think we had—there were 100 people on or something. And so, he had started from that direction, and at the same time I had been meeting—I was leading Agile Knoxville, a local meetup group, and I was working with James [Horry 00:06:02] at the time, who was hosting and leading different meetup groups. And we would sometimes have conflicts where one group would have a meeting on a Tuesday night and someone else would also have a meeting on a Tuesday night. And then, you know, people would say, “Well gosh, I can’t go to both of them.” So, James and I had been talking about what if we were to get all these different meetups together and let’s try to coordinate, we can cross-pollinate, we can, you know, not—you know—we can share, like, well, you know, who—what recruiters are helpful, things like that. What are—where can we—what are meeting spaces?
So, he and I had started that and then Cody came along and started this kind of grassroots Slack group. And we said—James, and I said, “Oh, my gosh. We need to talk to Cody.” And Then it turns out—so we just—and so at first ever, I’d say the first-ever meeting was July 7, 2015, at Panera in Cedar Bluff. And so, we got together and a whole bunch of people were just running different meetups. And we said, “Hey, let’s start a—let’s start—let’s put this together.”
Chris: That’s awesome. So, you grew from there and it’s been eight years now. How did you all fare during the pandemic? I would imagine that was probably a really challenging time for a community to be keeping going.
Cody: Yeah, absolutely. Well, you know, one good thing—like, that’s the—well, I’ll back up a minute. You know, when we were talking about, like, back in 2015 at CodeStock one thing, you know, Greg Ostermeier mentioned was Slack in Chattanooga. So, that there’s a Chadev, you know, they had a Slack group. And the lightbulb for me was Knoxville is so geographically dispersed, you know, we’re really spread out more than a lot of cities. There’s no town hall, so to speak. And so, that’s a challenge for us.
And so, when I heard Slack, I thought, you know, it’s the virtual town hall, you know? And it’s like, it totally made sense. And I think it resonated and that’s—you know, we’ve got over 2000 members now. But I think because of that, that helped us a little bit in the pandemic because we’re virtual. You know, a lot of what we do is virtual.
Now, meetups, definitely. You know, in-person meetups have suffered and, you know, we’re always looking for ways to help the community, whether it be meetups or other things. So yeah, it was challenging. And we also kind of just laid off a little bit. It’s like, we just relaxed. Like, let’s not kill ourselves during this pandemic. Let’s just, you know, see how it plays out. And so, you know, here we are, still kicking.
Chris: Yeah, that’s cool. And I mean, like, coming out of the pandemic, you know, people start to be able to meet again—and I know a little bit of what this is like because I’m actually on the board of the American Marketing Association here in Knoxville, so we’ve dealt with the same kind of similar, like, do we come back? Do we not? When do we start having things? So, I would imagine, like, there’s probably been some shifts. What have you all seen now as a community?
Adrian: One thing we’ve seen is we’re—so used to be—we’re as an umbrella organization that kind of helps local tech meetups. We’re seeing a lot less participation in people coming out to meetups, for example, where it used to be, you know, maybe 20, 25 people would come out, would show up on a Tuesday or Thursday night, you know, and talking about technology, you know, a lot of them are struggling.
And some organizers who used to run these, you know, have now just said, “Yeah, I’m not doing anymore. I work from home, I’m busy.” Or, you know, just different stages of life. So, we’ve had, you know, a couple of we—KnoxDevs events. We call them quarterly events, although they’re not really quarterly. But we have had—like, I think our last one we had 87 people there or something. So, you know, that’s a pretty great turnout, you know, for a tech event in Knoxville. So.
Cody: I think it does help. The quarterly events are good becau—you know, monthly, it’s tough. There’s a lot of events out there, people are constantly on video calls and so the quarterly event is good for us because you know, people really crave that social interaction and it’s a pretty good kind of rate for us to meet every quarter, you’re looking to reconnect with people. And so, I think that that’s helped us.
Chris: Yeah, that’s great. That’s great. And then, of course, we get to Devmoot. How did the idea for Devmoot come about?
Cody: I would say, it’s pretty simple. You know, I mentioned CodeStock, which is a great conference. They’ve ha—you know, 1000, 1500 people every year. It’s a great thing for Knoxville and they—it didn’t happen this year. And you know, that disappointed a lot of people, you know, myself included. It’s a great place for people to get together and socialize, you know, not just from Knoxville, but all over the region.
And, you know, for various reasons, it didn’t happen this year. And so we, you know, a lot of people were eager to do something and we thought, well, let’s—we’ll just let KnoxDev step up. And so, we’ve organized this. And again, we don’t know if we’ll do it next year. You know, hopefully, CodeStock will be back. But we’ll deal with that next year.
Adrian: Yep. Yeah, I think as soon as CodeStock announced that they weren’t going to have the conference this year, people immediately, that same day, started messaging KnoxDevs and posting, “Well, what’s KnoxDevs is going to do?” And you know, which we’re, you know, kind of, in a way flattered, you know? Because, like, we’re in, you know, in—kind of terrified, at the same time thinking, “Well, no, no, that’s not—we cannot replicate CodeStock. We don’t want to.”
So, but when enough people asked, we started talking, and we said, “Well, you know, we could probably do something smaller, easier, simpler, you know, not trying to replicate CodeStock.” But even if they were back next year, where it could be something that’s, you know, that’s adjacent to it, you know, different time of year, different topics, that sort of thing. Additive.
Cody: And we had some early, strict tenets from the beginning, and, you know, one was keep it simple. So, we chose a single track, you know, a single day. And we wanted to avoid the mundane, so, there’s a—you know, one challenge we have is just about any topic you’re interested in, you can go on YouTube. And in rather than, see, you know, somebody local talk about it, you can see the guy who created it, you know, or guy or gal who created it, and hear it, you know, straight out of their mouth. So, we have to find ways to make it interesting, you know, to have this conference here.
And so, we really wanted to avoid just, like, the basic topics and we wanted to focus on more cutting-edge things. Like we’ve got a few AI talks, we’ve got quantum computing, and you know, Web3, which is always fun to discuss. So. And then also social was a big part of it, you know? Keep it social because I think people crave that, you know? They come for this, they come for the talks—or they really come for the social, you know. The talks are a real reason to get out, but I think, you know, a lot of people come for the social aspects of it, the networking.
Adrian: Yeah. And I think that’s part of the reason where you’re planning to have a cash bar, you know? And so, beginning around lunchtime and, [unintelligible 00:13:30] with appetizers and hors d’oeuvres, you know, later in the day, just a happy hour for people that want to stay around. You know, because sometimes at events like this, especially everyone working from home—or not everyone, but a lot are working from home—we only see people, what, you know, maybe once a year. So, I know at CodeStock, it was a time I would just reconnect with people that I only see at CodeStock, you know? And I can hear what they’re doing and what’s new, and what are they working on, and how’s your family doing and things like that.
Cody: Yeah. So, circling back to your point, though, it does—the pandemic did make things challenging. And now that there’s so many people working from home, it’s a different landscape. But people crave that socialness and need those opportunities more. So, you know, in some ways it helps, I think.
Chris: Yeah. I think with more people working remote, you see a lot less people coming out to events during the week because they’re not coming from work—
Adrian: Right.
Chris: They’re coming from their house. They’re leaving their comfort zone. So, to do something like this is more like, “I’m dedicating a day to be out there? Yeah, I’ll do that. That sounds a lot of fun. Like, a lot of fun.”
Cody: And it’s strange, you know, you cut out your commute, you’d feel like you’d have more time, but that time just gets sucked up and—
Chris: Oh, yeah. Yeah.
Cody: [laugh]. Well, one other thing I always feel really strongly about and wanted to mention as far as, like, building a brand. You know, back from when we started KnoxDevs, to me, there’s this kind of exponential aspect of working with others and standing on the shoulders of others, you know? Like, you know, I feel part of the impetus or the reason that we were able to get KnoxDevs off is because of CodeStock. And you know, and CodeStock had founders. And, you know, there was like the .NET Meetup group, you know, many, many, many years ago that turned into CodeStock.
And so, like, I really feel strongly about how individuals can make a difference. And, you know, really see that with CodeStock and other groups. And that’s—I like and—my goal is to inspire other people to, like, you can make a difference as an individual, and then, you know, inspire others and then, you know, hopefully, Knoxville and areas like this continue to blossom.
Adrian: Yeah, and I would add to that. So, one of the things that we’ve been able to do, like, even if you’re inspiring people is great, feel like, just start a meetup or start a similar group in another town, or, you know, we’ve had people mentoring high school students, for example, you know, people taking time to mentor, you know, new people who are doing career switches. And I can’t—I lost track several years ago of the number of people who have gotten jobs just through KnoxDevs, through just word of mouth, you know? Someone will mention something or you’ll meet someone, you know, and then next thing you know, you’re like, “Oh, wait a minute. I remember, I heard so-and-so was looking for someone,” and you make that connection and next thing you know, they have an internship or a job. Which is really exciting to me.
Cody: That’s fantastic. It’s the only way to hire, you know, through those social connections almost. You know, it’s so much more valuable than, you know, a paper resume or whatever to get that reference. And yeah, it feels great to see that people getting hired and, you know, starting ventures and all that sort of thing. It’s amazing.
Chris: Yeah. Yeah, that’s—and Knoxville has a great tech community, too. Like, I think it’s very low key, I mean, during Knox Startup Week and things like that, it becomes more prominent and more prevalent, but I mean, there’s a lot of tech entrepreneurs around here, there’s a lot of developers and people that are doing a lot of high tech stuff to begin with. So, you know, the fact that you all are fostering that community and hopefully inspiring even the next generation of organization leaders, whether it’s KnoxDevs or whatever the next iteration is, I think, is really cool. So.
Cody: And, you know, we’ve got the benefit of ORNL in our backyard and a great research university. And, you know, even like, Pellissippi State is a fantastic community college, they’re doing a lot of good things there. So yeah, it’s a great, great place to be.
Chris: Awesome. Well, why should people come to Devmoot? Besides the bar opening at lunch [laugh]?
Cody: [laugh]. Well, you know, we’re trying to make it a really cool event in all aspects, you know, great social, and great talks. Jeff Prosise, I think you’ll be talking with him, he’s a really good speaker and, you know, wrote a book on AI recently. So, some great talks. The Relix is cool because, like, right—I don’t know if you can see it, but over, there’s an alley next—you know, right, out this door, so there’s a kind of a sealed-off alleyway, so there’s a great place to, for people to go conversate. And we’ll have some tables set up out there. And, you know, some good swag, some good food, plenty of coffee, and you know, a lot of great people from the community. So, why not come?
Chris: Yeah. And where should they go to register?
Adrian: devmoot.com. Just real simple. You can see what’s it all about, the schedule, speakers, link to tickets, just all kinds of information there. And so, I would add to that is when we started this, we said if we can’t do the conference well, then we let’s don’t do it. And so, this is our first conference, then we will make some mistakes, but so far, things are coming together pretty well.
And one great thing about that is we had a bunch of great sponsors. So, you know, one of those, Premier Staffing Partners, who’s actually been a sponsor of KnoxDevs since 2015, we are a very low-budget high-volunteer organization, so we get a whole lot done on a shoestring budget, but we’ve always tried to do it well. And Premier and you know, especially [Chris Ann 00:19:03] has always come through for us there, and James, and really appreciate their support. But in this conference, we’ve had a lot of other sponsors step up so we’re going to have actually a great lunch, we’re going to have refreshments all day long. Cody mentioned coffee, but there’s also tea and lemonade. And then we’re going to have, you know, hors d’oeuvres in the afternoon, very nice hors d’oeuvres. So, it’s going to be a good conference, it’s great talks, you know, good food, just great sponsors, a good quality overall, a great venue. And it’s going to be social and laid back but also very professional.
Cody: Yeah, and I’ll say, we’re a nonprofit, and you know, the way nonprofits work, people are doing this because they want to, you know? Nobody’s getting paid for this. And it takes a lot of people. You know, Katie Cleveland and Andy [Cowell 00:19:51] and just lots of other volunteers really helping make this thing successful. And throughout the years, we’ve pulled off a lot of cool events, we’ve done, you know, hackathons with hundreds of people and we’ve had events with the mayor of the city and county and, you know, events in the Sunsphere and at the Museum of Art. So, I think this is going to be just another fantastic event.
Chris: Excellent. Well, we’re looking forward to it.
Adrian: And speaking of branding, and kind of on the shoulders of others, one thing we should mention, is we should mention Alex [Pawlowski 00:20:27]. He was here for several years and created our logo, I think, and just so much of what we use today in KnoxDevs was created—but it was—he worked circles around both me and Cody. And so, it made us look great when he was doing most of the work. And we tried to tell people that, but, you know—so.
Chris: Yeah, awesome. Yeah, the visual elements meet a lot when you’re creating a community, creating a brand, anything in that respect. I mean, even down to how the K-D is connected probably means something when you went to design that logo.
Cody: It’s a secret, though.
Chris: [laugh]. Well, it’s a secret. We’ll leave it at that. But yeah, so Cody, Adrian, thank you all for being here today. I’m excited to be coming to the conference and I think there’s going to be a lot of people that are going to be excited to be there. So yeah, if you’ve not gotten your tickets yet, devmoot.com. We’ll have links in the [show notes 00:21:21] here for anybody who wants to check it out and get registered. And with that said, we’re going to move on to the next part of this episode with Jeff Prosise, where I’ve got a great conversation ahead. So, stay tuned.
Chris: All right, Jeff, welcome to We Built This Brand.
Jeff: Great to be here.
Chris: Really excited to talk to you today. As you may know, as listeners already know, we’re talking a lot today about the Devmoot conference. And you are going to be the keynote speaker at the conference, so I’m really excited to just be able to sit down for a little bit and have a conversation with you about, I mean, frankly, some of what you might be talking about so people get an idea of what’s going to be going on at the conference. And then learn a little bit more about you and your background.
We Built This Brand is primarily a show about branding and marketing, and also we like to talk to founders and people that have started their own business, their own brand, from the ground up and you have done that. And not only have you done that, but you’ve managed to successfully leverage that into selling that company and growing into an even bigger brand. So today, I’d love to just dive into that a little bit first and talk a little bit about your background there. So, take me through, where did you get started?
Jeff: That’s a great question [laugh]. I’ve been at this for 40 years or so, so there’s probably not a short answer, but I’ll try to make it as concise as possible. So, I started my career as an engineer, went to the University of Tennessee, and studied Mechanical and Aerospace Engineering. Graduated in the early ’80s and, you know, I worked as an engineer for a few years. But also got introduced to computers, personal computers, and discovered a love there.
In fact, I felt I was a lot better at writing code than I was at designing engineering things. So, some of our listeners may remember a magazine that was pretty popular back in the day called PC Magazine.
Chris: I do.
Jeff: Yeah. I got to gig writing for PC Magazine. I was contributing editor there for several years. Also at Microsoft Systems Journal, which was really a great periodical in the hardcore tech industry back in those days. 1990, I decided to make a big career jump. It was a little scary at the time, but I guess in [laugh] retrospect, it paid off.
I retired from engineering and started writing my first book, continued writing for the magazines, wrote my first book, hit the speaking tour and things like that, started doing consulting and training for companies, and basically did that all through the ’90s. Then in 2000, I got together with some very dear friends and we launched a company named Wintellect. It was originally based in Knoxville, but we eventually moved to Atlanta, in part because it was easier to find talent down there, dev talent. So, Wintellect started as a training company. Microsoft was our biggest client.
Me and two of my partners spent many, many years flying around the world to all the different Microsoft offices in Beijing and Shanghai and Hyderabad and Dublin and everywhere else, where our job was just to make these software engineers at Microsoft the very best they could be, better than Google and better than Amazon [laugh] was our mission. So, that evolved just a little bit because even in the early days of Wintellect when we were focused on training, we knew that we wanted to start a consulting division as well, not just to teach customers how to build products, but when a customer wanted to help build that product. So, we launched that in 2005. A gentleman who’s still a business partner of mine today, and a dear friend, Todd Fine, launched that, quickly grew it so that its revenues were a lot larger than the training revenues. So, Wintellect spent many years not just training developers all over the world, but helping companies large and small build software solutions.
We did that until 2021 And then we merged with a company named Atmosera, located in Beaverton, Oregon. I work for Atmosera Today. My title is Chief AI Officer. So, I have a lot of fun because I have a lot of meetings every day with customers and potential customers, understanding their business processes, the challenges they’re facing, and figuring out creative ways for AI to help them improve their business, then working with my team at Atmosera to actually build those solutions for them. So, I often tell people I’m having more fun right now than I’ve had in probably 25 years because AI has changed everything. I sometimes pinch myself in the morning because I feel so lucky to get to do this for a living. It’s fascinating stuff.
Chris: Absolutely. I mean, AI is even changing the industry that we’re working in now. I mean, for those that listen, they know, we’re a podcast production company at our core, so we see this stuff coming down the pike all the time with new changes and new, you know, new technology because it’s directly applicable to what we do. So, that’s really cool. So, let’s go back to when you started Wintellect. You said you wrote a book. What was the name of that book? What was that topic?
Jeff: My first book—because I wrote several, over the course of about ten years—my first book was on DOS 5.0, if you remember that. It was a big deal at the time.
Chris: [laugh]. I do. I do.
Jeff: But a lot of people don’t remember that. My first book was on DOS 5.0. And probably the book that, when people recognize my name—if they recognize it—they probably do so because of a book I wrote for Microsoft Press called Programming Windows with MFC. It helped Windows programmers, a generation of Windows programmers, learn to write Windows apps in C++.
Then I wrote a book after that, called Programming Microsoft .NET. That was in 2001, and at that point, I decided I wasn’t going to write any more books. I’ve been doing it for ten years.
Things had started to change. The internet had really come along and frankly, people learn in a different way these days. They rarely go to Barnes and Noble and, you know, look for a book on the technology they’re looking for. So, I stopped writing then, but I did write a book last year on AI. It’s called Applied Machine Learning and AI for Engineers. The publisher is O’Reilly.
And the reason I wrote it was because, number one, AI is so incredibly important. It is literally changing our lives every day, mostly for the good, and we’ll continue to do so. I often start off talks by saying, “When we one day cure cancer, we will thank AI for making it possible.” And I firmly believe that. So really, with zero expectations for how the book might sell or anything, I did decide to write one more book last year.
It was a labor of love. I’m confident it’s the last book I’ll ever write, and I was determined going into it for it to be fun, not a chore. Because writing a book can be a chore [laugh]. And it was fun. Now, I’m hoping to do a second edition of it next year because things are moving so quickly in this industry. You can know everything about AI a month ago and there’s new stuff to learn today.
Chris: I can imagine keeping up with that is quite the challenge. And any tech book. I mean, you mentioned writing on DOS, I mean I remember being—not to date myself—a being a little kid and my parents’ first computer at home was a DOS prompt computer. So, to play any video games or do anything like that, I had to figure out DOS.
Jeff: You had to figure it out. Yeah.
Chris: It was a necessity to get by and, you know, ‘cd’ this and make big changes and all that stuff. So, I have young memories of that, young memories of, like, Windows and what that was. And I mean that’s so cool that you got to write those, essentially, textbooks for a lot of developers and engineers. And how do you—like, I don’t want to get into the, like, the specifics of it too much, but I’m curious, like, how do you go about, like, not only deciding to write that kind of a book but, like, getting a book like that published? Like, I would think, just in general, like DOS, Microsoft should write a book on how to do it.
But like, obviously, there’s a need for other experts in the field to talk about it because there’s more nuance to programming than just, hey, we developed this code, you can read this book and find out this information. But how do you go about that?
Jeff: You know, writing a technical book is different from writing a work of fiction. If your desire is to write a novel, anyone can write a novel. It’s very, very hard to find a publisher for it. It’s a little bit easier with tech books, especially if you’ve been out of in industry, writing for magazines back in the day or have a popular blog today, there are publishers that will talk to you. And you can probably find a deal if you’re a decent writer and have a little bit of a name.
So, the way my first one came about, PC Magazine was owned by a company called Ziff Davis Publishing. And the ’80s was the heyday for Ziff Davis because they had not only PC Magazine but some other popular magazines as well. And their magazines were so popular, they decided around 1990 to set up a book publishing arm called Ziff Davis Press. So, because I was a writer at PC Magazine, they approached me and said, “Hey, we need someone to write a book on this thing called DOS 5.0 coming up next year. Would you be interested?” And I said, “Sure.”
And actually, it wasn’t quite that simple. You know, I had a full-time job as an engineer, I had a full-time job in the evenings as well writing for magazines and speaking and things like that, but I ultimately decided to, you know, make that jump. I remember it was 1986, I believe. Bill Macron was the editor-in-chief of PC Magazine. He passed away a couple of years ago, unfortunately, but one of the most incredible people I’ve ever known. And he flew me up to New York City, gave me a tour of the office at One Park Avenue, and then sat me down in his office and tried to talk me into giving up engineering and doing computers full time, coming to work for Ziff and PC Magazine.
And I remember saying, “You know, Bill, I’m really flattered, but these PCs are a lot of fun, but I’m not sure that we’re ever going to be anything more than a hobby, really.” So, [laugh] so I declined. But by 1990, I kind of had this inkling that maybe these things were here to stay. Maybe they were even going to become a part of the fabric of our society. So, when I signed that first book contract, I left my job, my wife left her job as a school teacher and gave birth to our first child. That all happened in the space of a few weeks.
So, we went from, you know, a young couple, two incomes, no kids, to a young couple with no income and one kid. And it took me a year of 80 and 100-hour weeks to write that first book. It just about killed me. But, you know, I wouldn’t trade it for anything now. So, you know, I frequently speak to people who are thinking about writing a book and they say, you know, “What’s your advice?”
First word of advice is, “Don’t.” [laugh]. But the second is to just realize, you know, it’s going to be hard. There are going to be times when you wonder, you know, why am I doing this? And the really hard part comes about halfway through a book because you’ve been killing yourself, probably for months at that point, you’re halfway through and you think, “Oh, my gosh, can I, you know, can I continue this slog for another, you know, three months or six months, or whatever it takes to write the book?” And at the end of the day, it’s about willpower. You have to go into it bound and determined that you’re going to finish no matter what. And if you don’t go into it with that attitude, it’s going to be a really hard slog.
Chris: I can only imagine. I mean, I can relate to a degree. Not that I’ve written a book, but I started HumblePod right about the time of the birth of my first child, so [laugh] I can definitely relate. My wife likes to joke that we had two babies, not one [laugh].
Jeff: I totally get that [laugh].
Chris: I totally get that [laugh]. Oh, man. Well, that’s really neat. And I mean, you know, of course, that writing that book led to, you know, growth in your career. I mean, you said you made the speaking circuit, and I’ve definitely seen a lot of, you know, tech entrepreneurs, thought leaders, that kind of hit that circuit after they’ve written the book or after they’ve started a big podcast or what have you.
And, you know, what advice would you have out there for, say, someone who is a developer right now or a technical engineer, listening to this episode? Because this is coming up before Devmoot and I know, we’ll have some new listeners to the show. Like, what advice would you have to them about doing what you’ve done?
Jeff: Yeah, it’s a good question. You know, today, you would do it a little bit differently, I think. In 1990, if you wanted to get your name out there, and, you know, get people inviting you to speak and consult and stuff like that, a book was a great vehicle for doing that. Because, you know, in the early ’90s, that’s how people learned. You know, we learned from Microsoft Systems Journal, and PC Magazine, and from the books of the day.
It was a great way to get your name out there, and you know, for me, it led to—I didn’t really intend to hit the speaking tour, but your first book comes out, if it gets some attention, you know, people start calling you and saying, “Hey, would you come and speak at this conference or at that conference?” And I foolishly said, “Yes.” [laugh]. No, I’m kidding there. I still do, typically, four to six conferences a year, mostly international.
And it’s just, I enjoy it; it’s kind of part of who I am. But today, making a name for yourself is a little bit different. You can write a book, most likely no one cares because books just don’t sell that well anymore. People have other ways of learning. So, you know, having a popular and well-regarded blog is a really good way to get attention.
Being proactive on social media, especially the so-called professional special media, like, LinkedIn, you know, put out a great blog post that solves a problem people are having or answers a question that they’re asking and then, you know, splash it on social media to get attention, that’s a great way to kind of build that name, if you want to, you know, parlay that into speaking or whatever. Very different today than it was then, but it’s still possible.
Chris: You know, for sure. I mean, I know some colleagues of mine that have—other entrepreneurs that have been in that space, and even recently been approached about writing technical books and things. So yeah, definitely still opportunity out there for those. That’s why I ask and I’m curious because I’m sure somebody listening to this will go, “How do you get to that point? Like, how do we get there?”
Jeff: Well, and you know, writing a book today, you know, there were a few books that sell really well; nothing like it was back in the ’90s. I was lucky. I guess I wrote nine books back in the ’90s and a few of those sold really well. If you had a good-selling computer book back then, you could make a living off of it. Very, very difficult to do that today.
But there’s still other reasons to write. I mean, one thing about a book is it focuses you. You may know a subject inside out, but I guarantee that when you start writing a book about it, you’re going to have to dig in and learn about the things that you don’t know. So, it’s a really good way to not so much expand your knowledge, but for me, kind of fill in the gaps that you may not even know were there. But then yeah, when you, you know, speak or go into a company to do consulting or something like that, it’s always nice to have a couple of copies of the book and drop them off. You know, whether you deserve it or not, you get instant credibility when they hear that big thump on the table and there’s that book with your name on it.
Chris: Yeah, yeah, it’s definitely good credibility builder for sure, I can imagine. I definitely know folks, a Knoxville luminary, marketing luminary who will remain nameless, but I’m sure people can figure out pretty quickly, like, he started his career I could tell, like, he wrote his book and that was his foot in the door. Like, “I’ve written a book.” Like, self-published, self-written, all that, and it definitely helped. And I know other folks that have gone more the traditional route like you and that’s really cool. So awesome.
Well, let’s take a shift now. Like, we were talking about the book writing but, like, I’m really curious about what it means now to be Chief AI Officer at Atmosera. So, what is a Chief AI Officer? Because I don’t think I’ve seen that title before.
Jeff: Yeah. In fact, you’re seeing it more and more. Before I took on that title, I did a little research and yeah, it’s a real thing these days, but it’s still relatively rare. So, what I do at Atmosera as Chief AI Officer is pretty simple. I have a lot of meetings with clients and potential clients.
And the high-level objective is to help them figure out creative ways to use AI to improve their business and improve their productivity. So, it typically involves a lot of meetings. Most of those are remote, sometimes I go to the client site. And understanding what their business is, what their business processes look like or what software they’re writing if they’re doing that, and understanding where the challenges are. And then what we typically do is brainstorm and brainstorm about ways that AI could help them.
And what I usually advise a customer is, “Hey, AI is very broad. There are a lot of things we can do—a lot of things we can’t do, by the way—but let’s pick two or three high-value targets, problems that we can attack with AI where we can build a solution for you and we can easily measure the results of that solution, the increase in productivity or the decrease in time required to do something.” And if we’re successful there, if we pick those two or three high-value targets and you prove that it makes your business better, then let’s talk about other things we can do as well. But my approach is, let’s focus and identify those HVTs, and let’s see if we can’t do something magical for your business. So, it’s a lot of fun. A lot of brainstorming, a lot of whiteboarding, I get an incredibly wide variety of questions and asks from customers.
And sometimes the answer, by the way, is AI can’t really do that. One of the things I often do with customers is, educate them about what we can reasonably do with AI and what we can’t. Customers come, you know, in different levels of sophistication. I’m working with some right now, one company, for example, has a 30-page two-year plan, very detailed. It’s a strategy document about how they’re going to increase productivity by 25% over the course of two years, using AI.
Well, for me, that’s a dream client because they already know a fair amount about it and now I can read that document, go have discussions with them, and we can figure out tactical ways to help them meet that goal. At the other extreme, we often have clients who call and say, “Hey, we’re trying to figure out this AI thing and we don’t even know where to start. We’ve got this idea that it could help our business and we’re certainly afraid that if we don’t do it, our competitors will, so Jeff, can you help us out?” So, you know, whether they come in at the high end with a 30-page strategy document or, “Help us figure this out,” it’s a challenge and it’s fun. And that’s what I do.
And we have some very talented engineers and software developers at Atmosera. If we end up working with that customer to actually build a solution or a solution set for them, I usually participate in the design and architecture, maybe a little bit of the coding, but I hand it over to my engineering team and, you know, they make great things happen in code.
Chris: So, as you look at the technology itself, like AI, I think there’s a lot of misconceptions, like you mentioned, like, it can do some things, but it can’t do other things. Like, what are some of the main misconceptions people have about AI in your field?
Jeff: A lot of folks because they haven’t delved deeply into it, look at it as something magical: it can do anything. I’ll give you an example. I had a customer ask me a few months ago, “If we are designing a new aircraft and we have all the parameters for that aircraft, can AI tell us what the proper angle is for the aircraft’s nose gear?” And I said, “No, it really can’t. Unless you have data on millions of aircraft that I can train up a model on, I can’t really help you there.”
So, I think a lot of folks, just because they haven’t dug real deeply into it, they have problems, they have challenges in their business and they’re hoping AI can solve it. Sometimes it can. Sometimes it can’t. So, that’s why, you know, education is a big part of it, helping them understand what we can do and in what we can’t.
Chris: And what do you mean by AI? And I know that sounds like a very basic question, but like, I know it’s artificial intelligence, I know that there’s a lot of different options for what AI can be, but like, do you have a tool bag of applications you use? Is it custom-built in-house? How does that work for you?
Jeff: A little bit of both. We build a lot of stuff in-house. We build machine learning models slash AI models. We also use tools like OpenAI’s ChatGPT and GPT-4. Large language models are a really big deal these days and probably 90% of the calls that I get from customers these days have to do with, hey, can you help put an LLM—a large language model—over our internal documents or over our database or something like that. So, in that case, we build a solution from scratch, but we lean heavily on those super advanced, super expensive LLMs like ChatGPT or Llama 2 or others.
Chris: That makes total sense. I was just curious if it was something that was developed or not. So, that makes a lot of sense.
Jeff: We do a little of both. And, you know, ask the question, what is AI? AI technically is a superset of machine learning. However, machine learning is a superset of deep learning, which is mostly machine learning done with deep neural networks. And today, when people use the term AI, they’re really talking about deep learning.
Because when you look at today’s most sophisticated models like Stable Diffusion or Midjourney—which do image generation—or Llama 2 or ChatGPT—which does text generation—those are all very, very complex neural networks. But I often tell people what AI really is, is a set of clever solutions to specific problems that computer scientists and data scientists have evolved over the years. There are really two areas where AI is really good these days. One is dealing with computer vision problems, like, “Look at a live feed from a security camera and tell me when there are people in that feed,” or, “Look at photos of parts coming off an assembly line in a manufacturing facility and tell me which of those parts might be defective and should be inspected by human eyes as opposed to the non-defective parts.” So, the field of computer vision has advanced a lot since about 2009, 2010, and those kinds of problems we can solve pretty easily.
We can even build object detection models which look at photos or frames and videos and identify the objects inside it. I was working on a job recently, for example, where a customer wanted to use a camera feed with an object detection model over the top of it to tell them when a truck entered their dock area, and even to identify what type of truck it was, to identify the company, who owned the truck from the logos on it. Ten years ago, it would have been very difficult to solve that problem. Today, it’s relatively easy to do.
The other area where AI is really good today is with natural language processing. And that really—well it started a long time ago, but it was in 2017 that we encountered a real milestone when Google introduced what we now know was the world’s first large language model, it was called BERT. BERT was essentially the progenitor of ChatGPT and the other GPT models. And with those models, we’ve gotten very good at doing things like translating text or speech from one language to another and even at generating text. So, computer vision, problems that involve language translation, document summarization, even text generation, we’ve figured all that out. But when you think about it, in the universe of what’s possible, that’s a very small piece.
Chris: The possibilities there are crazy. I mean, even from my work in the digital world, like, being able to tell a computer what I want it to come up with visually and it pop up and be almost exactly what I was expecting is just insane.
Jeff: It is insane. I’ve spent quite a bit of time recently understanding how image generation models work, models like Midjourney and Stable Diffusion and DALI-2 and a handful of others, and I’m with you. You know, as a former aerospace engineer, I know how airplanes work, and yet, every time I fly, when we speed down that runway and we rotate and lift up in the air, it’s thrilling to me. It amazes me that it works as well as it does.
It’s the same way with image generation. I know how it works, but it still amazes me, especially when we’re talking about prompted image generation, also known as text-guided in the generation where, as you said, you just type in a description and boom, something comes out on the other end that matches that description. That is fascinating stuff.
Chris: Yeah. And what’s even crazier is you can not even have a good idea of what you want. Tell ChatGPT and it’ll give you the prompt to give [laugh] Midjourney, and then you’ve got your image. It’s insane.
Jeff: Yeah, exactly right. And you know, one of the remarkable things too is, even though it’s text-guided, you provide a description an image is generated. If you provide that model with the same description one thousand times, you’ll get one thousand different variations of that image. That’s because of some randomness that they built into the image generation process. But yeah, it’s remarkable.
And, you know, as models like Midjourney, and Stable Diffusion continue to evolve, they just get better and better, you know? Everyone knows some of these models aren’t really good at rendering hands, you know? Sometimes those hands may have seven fingers or something like that. That’s not really a fault of the model. It’s that it didn’t see enough human hands when it was being trained.
As the model is further developed and further trained and it sees more human hands, it gets better at that. And some image generation models are better at that than others because the ones that are better were trained with more images that contained human hands. So, you know, it’s usually not a deficiency of the model with the technology; it’s really a result of how the model was trained. And, you know, that’s a point I often make to customers and to audiences. You know, we talk about bias in AI models, and certainly, there is bias. The bias comes not from the model itself, but from the data that it was trained with. At the end of the day, all of the data that model was trained with was generated by humans, right? Which means there is going to be some amount of bias in the output from that model because all it knows is what it learned from the data that we trained it with.
Chris: Yeah. And I think that brings us to an interesting point because I know in the marketing world, there’s a lot of fear of AI. There’s a lot of, “Oh, my gosh, it’s going to take over my world. It’s going to—you know, I’m not going to have a job in a few years because of it.” And even in the tech world, there’s a little bit of that with developers and basic coding.
I mean, I’ve literally gone to ChatGPT and said, “I need HTML code that does these buttons and these links,” and it spit out exactly what I wanted. And it’s crazy that it can do that. And you know, if it can do that, then why do I need a web developer anymore? Why do we need some of these smaller people? Or, I shouldn’t say smaller people—
Jeff: [laugh].
Chris: That was that was misspoken. But, you know—
Jeff: I know what you meant, yes.
Chris: [laugh]. Why do I need some of these people helping me? Everybody’s a very critical part of that production chain. But yeah, like, how do you overcome that? Where do you see that going?
Jeff: So, there’s no doubt that AI is going to change things. And yes, there are going to be jobs that get obsoleted. My view of it is, AI is going to make us more efficient, is going to make us more productive, make us better at what we do. And yes, there may be less need for people, you know, just write the boilerplate HTML code. But I think history shows that, you know, people adapt, people evolve.
We have some very smart people who write that boilerplate HTML code; they’re going to find other things to do, perhaps ways that they can make a greater impact like writing a little JavaScript code to make that HTML interactive or something like that. But you know, to your point about ChatGPT, it is remarkable. I often tell audiences, when you go to the ChatGPT website and ask it to write you a poem or a marketing slogan, it’s cool, but it’s really just a parlor trick. It’s how they got attention for it. The real value of these models is—well, there’s several, but the number one use case for large language models today is putting them over your company’s internal documents to easily surface information.
I’m working on half a dozen projects like that right now for customers. Another is putting it over your databases. But a lot of people don’t realize how good ChatGPT is generating code. I mean, it is really good. It was trained on more than a billion lines of code from GitHub.
In fact, there’s a class action lawsuit winding its way through the courts against Microsoft, OpenAI, and GitHub right now because, you know, there’s some valid concerns there. Even if they used only code in public repos, a lot of—most of the code on GitHub, even in public repos comes with open-source licenses, which may require for example, attribution to the author. ChatGPT doesn’t honor that, although OpenAI is working on that right now.
But it’s very good at generating code. And that really, that makes us more productive in a couple of ways. One, as a software developer—and I still write a lot of code, part of my job is writing POCs, Proof Of Concept apps for customers, to—so rather than just describe to them what we can do for them, I’ll spend a weekend or Saturday or something, building a POC, push it up to Azure where they can play with it and it’s literally a case of where a picture’s worth a thousand words, especially when they see how well it works. But as a programmer today, most of us programmers used spend a lot of time on Stack Overflow, you know, trying to figure out how to write a piece of code to do something, go to Stack Overflow, you know, ultimately find an answer by piecing things together.
Today, most of us go to ChatGPT and have it write the code for us. A few weeks ago, I needed a little piece of Python code, a Python function that would take a PNG file as input and would look at the alpha channel in PNG and generate a numpy array of ones and zeros from that alpha channel. So, I started to write the code and I thought, “Well, you know what? I’ll bet ChatGPT could write it faster.” I went to ChatGPT, typed in a description very much like what I just gave you, and boom, it spit out three beautiful lines of code that were probably better than anything I would have written. Plugged them right into my code and used them.
So, it is increasing programmer productivity by allowing us to do mundane things more quickly. Now, there are a lot of companies that I’m aware of that are not allowing their developers to use it because they’re very concerned about, hey, if Microsoft loses this lawsuit and they have—this company has put ChatGPT-generated code in one of their products, what’s going to happen? So, a lot of developers at large companies now are at least limited, if not completely forbidden from using ChatGPT to generate code. That’ll all work itself out, eventually, but you know, the scary thing is, it’s very possible that the plaintiffs will win in that lawsuit and that large language models that generate code will just go away.
And that’s going to impact not only developer productivity, but the other thing that’s incredible about ChatGPT’s ability to generate code is this: if you want to put ChatGPT over a database to easily surface information from it—and it’s probably the second most common use case today—guess how we do that? We use ChatGPT to generate a question that a user asks, to take that question and generate a SQL query. Then we execute that SQL query against the database—ChatGPT is very good at generating even complex SQL queries—and then we often take the results of the query, pass it back to ChatGPT, and say, “Phrase this response in human terms for me.” And by doing that you can work magic. If you take away the ability of a large language model to generate code, then it’s very, very difficult to implement that scenario.
Chris: I mean, just from the marketing perspective, like, being able to understand—like, one of the hardest things to understand right now in marketing is attribution, right? Like, where did someone come from, you know, how do they buy this product on my website and things like that. Being able to tie all that data together, put it into the database, and tell ChatGPT to suss it out, and then just give you a plain English answer of where everybody’s coming from, that’d be amazing. That’d be insane.
Jeff: Yeah. And that’s really where—how LLMs are changing the world. You know, Google for years has been able to take a natural language question that we type and provide an answer. But the deep-learning models they were using under the hood could only provide verbatim answers from documents identified through a vector search technique. The beauty of the LLM today is that we can pose natural language questions and we can get natural language responses.
So, it’s really upped the game and it’s why Google feverishly started working on BART once ChatGPT came out and they saw the impact it was making. They literally saw a threat to their dominance in search. So, LLM’s have changed the world. You know, in a Devmoot, in the keynote, one of the things I’m going to talk about is what ChatGPT really is and how it works. We’re going to talk about that seminal event that occurred in 2017 that ultimately led to it.
And the mic drop moment is that ChatGPT is nothing more than a next-word generator. When you ask it a question or give it a command, all it’s doing under the hood is repeatedly calling itself, generating the first word in the response, then the second word, then the third word, and so on. It’s very good at doing that, but it’s just a next-word predictor. And once you understand that, once you understand how it works, it gives you a lot of insights. For example, we know that GPT can’t do math.
If you ask it to add two plus two, it’ll probably get it right because in the massive volumes of data it was trained with, it probably saw two plus two somewhere. But ask it to add two arbitrary floating point numbers and in all likelihood, it will get it wrong. Why? Because it can’t do math. All it’s doing is looking at similar questions and responses that it saw during training and generating one word at a time in that output.
That also helps explain why it’s liable to hallucinate. Guess what? It was trained with essentially a snapshot of the internet in September 2021, plus two massive books databases that are part of a separate lawsuit now. So, some of the data that was trained with is not accurate. Believe it or not, there is misinformation out there. All it’s doing is—
Chris: No way.
Jeff: —mimicking—yeah, believe it or not.
Chris: [laugh]. [Exactly 01:01:13].
Jeff: All it’s doing is, you know, mimicking what it learned during training, misinformation and all. The cool thing though, is when we put a large language model like ChatGPT over a company’s internal documents, we do have techniques for not eliminating, but exponentially reducing the chance of hallucinations because rather than allow it to rely on all the data was trained with, we limit its scope. We say, “Here’s a question and I want you to answer it from this context.” And then we feed into it pieces of documents that we’ve identified through vector search, which is just a similarity search mechanism. And I wouldn’t say we’ll never hallucinate, but it’s far less likely to. So, you know, that’s just a terrific use case.
We had a company reach out to us a few months ago, this company was founded by a bunch of engineers about 30 years ago. They had developed some cool IP and patents and products around it and stuff, but these engineers are retiring and this company is scared to death they’re going to lose that institutional knowledge. So, they asked us, “Could you take internal emails, internal documents, transcripts of video interviews with these engineers and things like that and basically give us a knowledge base that would allow someone, once these engineers have left, to go in and ask a question and get an answer to it?” Thanks to large language models, we can do that. It would have been very difficult to do a couple of years ago, but that kind of solution is table stakes these days if you work with LLMs.
Chris: One of the things that makes me a great podcaster is the fact that I’m a verbal processor. I have to talk and think and process everything out loud to myself right now. And that just helps me think through things and get answers to things. I am one of those people that can walk up to somebody and completely monologue and get an answer that I needed just by sitting in front of them and talking out loud. So, that’s my background.
And why that’s important to what I’m about to say is that I have been finding a unique use for ChatGPT, that falls right in line with what you’re saying. And what I will do is I start my Mondays with some weekly planning. You know, as a business owner, I’m sure you know, you got to be prepared for your business, you got to be ready for the week. So, what I’ve started to do is I use an application called Descript. I open it up, and I just ramble.
I say, “Here’s what’s top-of-mind. Here’s what I’m feeling. Here’s what I’m thinking for the week.” It allows me to use my verbal processing skills to think through things and process things and help me on some level there. It’s almost like therapy [laugh] at times because I’m able to talk out loud and do all this stuff, right?
And then I say that and then it gets auto-transcribed—which isn’t perfect, but it’s still pretty good—auto-transcribes all of what I’m saying into text. And then I’ve got a whole script set up in ChatGPT. And I go, “Hey, imagine you’re my assistant, and you’re going to take everything that I’ve said, and you’re going to create action items for my week from everything that I’ve just said.” And it does it. And to your point, it doesn’t hallucinate.
Like, I was wondering, like, why it doesn’t make stuff up because it feels like everything there that I’m reading, I’m like, okay, that was good. Sometimes it’s not perfect, but overall, it does a really good job of, like, hitting the mark of what I need it to say and do. And it’s just been amazing for me because I can just take those action items and I’ve got them formatted. I’ve specifically told it to do it in a certain way so that I can put it in Todoist. So, I’ve got a digital to-do list once I’m done with this ramble. And it’s just amazing.
But yeah, like, it’s a bit of a humblebrag, but it’s just one of those cool things that I’ve been processing as you’ve been talking, like, that’s one really cool application for it that I’ve seen and makes me excited for stuff like ChatGPT.
Jeff: Yeah. One of the reasons I love talking to customers about how we can use AI—and specifically large language models to help them out—is hearing the interesting and creative ways that they’re using it. So, we have a salesperson in our company, for example, he’s a young guy really, really smart, when a salesperson in our company completes an initial call, we call it a scoping call with the client, just to find out, you know, what kind of what are they looking for us to do? What kind of problems are they having, you know, what’s their budget, if they’re willing to talk about that? We require the salesperson to fill out what we call a scoping document that answers some basic questions from that call.
This salesperson takes the transcript of the call from Teams, runs it to ChatGPT, and says, “Answer these questions for me.” And uses ChatGPT to generate his [laugh] scoping documents. I love it.
Chris: [Yeah 01:05:59].
Jeff: Yeah, I mean, it’s about increasing productivity, right, and, you know, finding unique and creative ways like that, that’s why it makes us more productive, you know? You mentioned earlier, a lot of people are afraid of AI. Every time I turn on the news these days, it seems I hear some news anchor or pundit saying, you know, what’s the government going to do to save us from AI? And indeed, there have been some really smart, famous people like Dr. Geoffrey Hinton—who was a Google; he was one of the people credited back in the late ’80s of inventing the backpropagation algorithm that we use to this day to train deep neural networks—and he said, “Whoa, we need a pause on this.”
I’m glad there have been other equally smart and well-known people step up and say, “No, no, no.” Any technology can be used for good or bad, but if we put a stop to this, we are disadvantaging ourselves. We have to keep going. We just have to be very responsible at what we’re doing. And you know, I’ve wondered, when people like Dr. Hinton are saying we need a pause, I wondered why… I mean, he—I’m sure he knows a lot more about AI than I do. Why is he afraid and what am I missing because I don’t feel that fear?
And I think what it was, was for years and years, when we trained deep-learning models, trained AI models, we built them and trained them to perform a specific task, for example, to translate English into French or something like that. When OpenAI rolled out ChatGPT and its predecessor, GPT-3, it was really the first time I’m aware of where the model was able to do things it had not been trained to do. Some of its capabilities surprised to even the data scientist who built and trained it. And I think that’s probably what gives people pause. It’s not that the model is sentient.
It’s not. ChatGPT doesn’t know what it’s saying. It doesn’t know what you’re asking. It is just generating a response one word at a time based on the question you asked or the prompt you gave it, and what it has generated so far. But it is able to do things that they didn’t anticipate it would be able to do.
So, I can see there why, you know, we should think about this. But as far as AGI, Artificial General Intelligence, computer models that can think like a human, we’re not even close. It’s unlikely to happen in my lifetime or your lifetime. You’re younger than I am. It’s unlikely to happen. If and when that happens, yeah, we need to think about this. But we’re not close. ChatGPT is not nearly as smart as it looks.
Chris: I think I can attest to some of that, but I know exactly what [laugh] you’re saying. And that’s a good point. I think, you know, from my perspective, based on everything you’ve said and other things I’ve seen, it’s like, especially like people in the marketing world, I feel like the big thing they fear is, like, losing a job to copywr—like, as a copywriter to this technology. And I think you’re right. I think it’s leveraging the technology to your advantage and not seeing it as a threat is the real challenge.
Even with developers, it’s like, okay, well, I don’t have to write basic HTML anymore. I can just have this do it and build a whole webpage for me. Great. Now, I can go and tend to all the little tedious things in there and clean it up. So, there’s all kinds of advantages to that. So, that’s really cool. Well, we’re getting close to the end of our time, so just wanted to ask you a couple more questions. First question is, you mentioned in your bio that you love the smell of jet fuel in the morning. Why?
Jeff: I do. So, now that I’m no longer an engineer—well, I guess I am; I’m an AI engineer, but I’m not a mechanical aerospace engineer—I love things that fly and I miss that. So, I have a rather unusual hobby. I build large radio-controlled jets with real jet engines that run on real jet fuel. And it’s true, I do love the smell of jet fuel.
So yeah, one of the challenges to doing what I do, I mean, you can buy the engines; there are a dozen or so companies around the world that makes them. I’ve got four or five jet engines in my shop in the other room there. You have to build the models, build the jets. For the most part, they come as kits and you have to build them out. But buying jet fuel, I can tell you from experience, you don’t walk up to into an airport and say, “Yes, I’d like to buy 20 gallons of Jet A.”
They frown upon that and might even be tempted to put you on some kind of watch list. So, the challenge is getting jet fuel. You can either make it from K1 kerosene, if it’s pure enough, or if you have a friend at a small airport that is small but big enough to host business jets, you can you know, call on a friend to sell you ten gallons or so. But [laugh]—
Chris: Nice.
Jeff: But I do love the smell of jet fuel and I love things that fly.
Chris: That’s awesome. I mean, one of the—going back to the DOS stuff and being a kid, like, I grew up playing Microsoft Flight Simulator. I have always been a flight sim kid. Growing up in Knoxville, specifically, I—what was it—I’ve been out to the [Mountain Hill RC Airpark 01:11:50].
Jeff: I’m a member there.
Chris: I had a feeling you were [laugh].
Jeff: You and I have probably seen each other out there, yes.
Chris: Maybe. Maybe sometime a long time ago. I never flew out there, but I was always curious. And then a year or two ago, I took my son out there. So, we may have seen each other because I took my son out there to go look at all the planes and all that because he was really curious. We were out and about.
So, that’s really cool. It’s a neat thing. It’s something—RC planes is something I’ve always wanted to get into. And especially now, as you can have, essentially VR planes and things like that, you know, it’s becoming a really, really interesting—I mean, it already was interesting, but it’s like even more interesting as a hobby now. So, yeah. That or just become a real pilot, which is my dream. But—
Jeff: Yeah, that or become a real pilot. You know, people have often asked me, why don’t I have a pilot’s license? I mean, I’ve you know, flown in airplanes with friends, we’ve all done that, but my answer is, I love jets, I love speed, and I’ve always felt that I would get bored putting around the skies in a Cessna. But let me tell you, if I could buy an F-16 and fly that thing around, I’d be there in a heartbeat.
Chris: Yeah. Yeah, I don’t blame you. I think it’s cool that we have Cirrus Aircraft in town, too, you know, and they’ve got their little jet. It’s not quite as fun as flying an F-16, but [laugh]—
Jeff: But I would take it. If they offered me one, I’d say, “Yes. Let me go get my pilot’s license first and then I’ll fly the heck out of this jet for you.” [laugh].
Chris: I don’t know if you know this, you don’t even have to do that because they provide the lessons with the purchase of the plane.
Jeff: Wow. Well, I need last year’s book to sell a little bit better than it is and maybe I’ll go by that jet [laugh].
Chris: [laugh]. Yeah. Yeah, that and Atmosera take off and you’re good to go.
Jeff: Exactly. Yeah.
Chris: That’s awesome. Well, cool. And then next question for you is—and this is always the big question we ask on the show—is like, what brand are you really a fan of right now? Or what brand are you crushing on right now?
Jeff: Two answers for you. One, if we’re talking about RC jets, my favorite brand is BVM Bob Violett Models. They’re based down in Florida. They’re one of the few manufacturers of large radio control jets in this country. The founder, Bob Violett, unfortunately, passed away a couple of years ago.
He was a retired Navy fighter pilot and I got to know him well over the years and I just love his company and his brand because they’re all about quality. No one makes a better radio control jet than [laugh] Bob Violett Models, in my opinion. In the tech world, it’s going to have to be OpenAI. OpenAI has changed the world. They’re not the only ones that build large language models, but they were the ones with ChatGPT that really got out there and generated the brand awareness and the awareness of the technology.
So, they have been masterful at marketing. It’s going to be interesting to see what happens long term. I’m not sure that their model right now is economically viable. One reason that it is viable, at least in the short term, is Microsoft has pumped in billions of dollars. So, we’re going to see, but I’m sure they can adapt. It’s, I think, the number-one brand in the tech industry today. So, let’s see what happens there. They are changing the world, one LLM at a time.
Chris: That’s really cool. And yeah, I definitely agree. They’re a fascinating model and I’m a proud paying customer, so that’s really cool. Excellent. So last, last question for you is, is there anything that you would like to promote?
I mean, obviously, we need to promote Devmoot, which is coming up and you’re speaking there. So, of course, we’ll be promoting that and of course, this whole episode will be covering that. But like, is there anything else you’d like to promote while we’ve got you?
Jeff: Well, obviously Devmoot. Really looking forward to the conference. I speak at conferences all over the world and it’s really special to be able to speak at one in my hometown. So, I’m not as plugged into the dev and AI community here as I am in other places, so I’m really looking forward to the conference. And I hope if you’re listening to this and you aren’t planning to go to Devmoot, give it a look. I guarantee you we’re going to have fun.
And not only me, but a lot of other speakers are going to share some really exciting stuff. Obviously, my book, it’s an O’Reilly book, Applied Machine Learning and AI for Engineers, give it a look. And now you know who to yell at if you don’t like it [laugh] but hopefully you will like it.
And, you know, finally, my company, Atmosera. We employ some amazing engineers, we build amazing solutions for customers, not just around AI, but around Azure and the cloud in general. So, you know, if you have needs along these lines, we would love for you to give us a look. I’d love for you to reach out to me, and you know, I’m not going to charge you. We can just chat and who knows, maybe I’ll give you an idea that will help solve a problem your company is having.
Chris: Excellent. Well, Jeff, thank you so much for coming on the podcast.
Jeff: Thank you. It’s an honor.
Chris: Absolutely. And with that, we will hope to see you all at Devmoot on Friday, September 15th.