Dustin Stout- Unlocking the Power of AI

Narrator: [00:00:04] Welcome to Compound Ideas, hosted by Ken Majmudar of Ridgewood Investments, this podcast will feature exceptional individuals to uncover deep insights into business, entrepreneurship, personal growth, investing, and multidisciplinary thinking so that you can learn how to improve your finances, find better investments, and pursue authentic lifelong growth, wisdom and happiness. Learn more and stay up to date at compoundideashow.com.

Ken Majmudar: [00:00:35] Hi. My podcast today is featuring Dustin Stout. Dustin is a very interesting entrepreneur. He started out in Western Pennsylvania with his first objective, which was to become an actor. But organically, as he learned about marketing, he ended up developing a marketing agency, which led him to his current role as the founder and chief executive of Magai. That's M-A-G-A-I, which is an AI based SaaS product that effectively allows you to access multiple LLM's, LLM's being large language models all from one interface, and it solves a lot of the pain points that people have of using directly models themselves. Please enjoy my conversation with Dustin Stout.

Ken Majmudar: [00:01:21] Dustin, thanks for joining the podcast. Great to have you.

Dustin Stout: [00:01:24] Great to be here.

Ken Majmudar: [00:01:25] So, Dustin, let's start with your background. Where did you grow up? What was that like growing up?

Dustin Stout: [00:01:30] Well, I grew up in a small town in Pennsylvania, a little place called Sharon. It is the home of the Quaker Steak and Lube, and Daffin's candy are our two claims to fame. From an early age, I was aimed at leaving that small town. Ever since I could remember, I wanted to be an actor. I wanted to be in film and TV. So when I graduated high school, I packed up, moved out to California. My dad already lived out there, so it was an easy transition for me. Went to school, a prestigious acting conservatory, and before too long realized that it's really hard to break into the acting world and actually make a living. So I ended up waiting tables like all the other actors, and I didn't really like that, so realized that I really wanted to be the master of my destiny. I didn't want anybody else to be able to tell me what I can do when I can do it. I wanted to work on the things that I was passionate about, with the people that I was passionate about. So that led me to eventually starting my own digital marketing agency.

Ken Majmudar: [00:02:25] Got it. Quickly about the small town. I don't think I've heard of the town it's called Sharon.

Dustin Stout: [00:02:29] Yeah, a little town called Sharon. Nobody's heard of it.

Ken Majmudar: [00:02:31] So what's it near? What's the biggest town nearby?

Dustin Stout: [00:02:35] I would say it's right on the Ohio border. So as far west in Pennsylvania as you can get, it's about halfway between Pittsburgh and Cleveland, right at the midpoint there.

Ken Majmudar: [00:02:45] Got it. And what's that town area? The bigger area around that town.

Dustin Stout: [00:02:49] It's called the Shenango Valley.

Ken Majmudar: [00:02:51] Shenango. So what's that like culturally or just give a slice of life out there?

Dustin Stout: [00:02:55] I would say our area is more a mix. It really was a mix of things. You got the more small city vibe, and then you've got the outlying country area and then you've got the more urban areas. So I'd say Sharon is a mix of all of those three types of communities. We also had a lot of Amish down there. We frequently go to Walmart or the mall and see buggies parked outside and Amish people walking around. That was a normal occurrence for us, but culturally not very diverse. When I moved out to California, I was super excited to be just completely enveloped in every culture you could iMagaine. Whereas growing up in small town Sharon, there weren't a whole lot of different people that didn't look like me.

Ken Majmudar: [00:03:41] Definitely. And you went out to Los Angeles sounds like.

Dustin Stout: [00:03:44] Yes. Well, I moved out to a place called Granada Hills, and it's in the valley. You hear about the Valley girls and so forth. That's right in the...

Ken Majmudar: [00:03:53] The San Fernando Valley.

Dustin Stout: [00:03:55] The San Fernando Valley. Correct? Yeah. Northridge, Porter Ranch and Granada Hills, the main areas there.

Ken Majmudar: [00:04:01] But you went to high school in Pennsylvania and then went out.

Dustin Stout: [00:04:04] Yep.

Ken Majmudar: [00:04:05] Okay. And what was acting school like?

Dustin Stout: [00:04:07] It was weird. It's not like a traditional academic place where you're listening to lectures and doing book reports or essays, so to speak. It was very artsy, very artistic, very experiential and very hands on. The instructors, there are all seasoned actors and know their craft very well. So it was a very alternative style of higher education, I would say.

Ken Majmudar: [00:04:33] Did that suit you?

Dustin Stout: [00:04:34] It did. Yeah. I did not do well in the academic system. I was not a straight-A student. I was not top of the class by any stretch of the iMagaination. I figured out quickly in high school that if I just paid attention really well during class, I never had to study and I could be a C student. And that was good enough for me. I wasn't looking to get into any Ivory, Harvard, or any of the big colleges. I knew I was going to be an actor, and I knew that all of this classroom work that I was doing was not going to help me become a better actor. So I was very focused on learning exactly what I needed to learn in order to get to the next step.

Ken Majmudar: [00:05:12] What was it about acting that captured your iMagaination?

Dustin Stout: [00:05:15] I loved films, I just loved watching movies. I loved a good story. And as a child, I guess I loved being the center of attention and entertaining people, making people smile. So that was really what drove me to just want to tell stories on the big screen.

Ken Majmudar: [00:05:29] Got it. And so how far did you take that?

Dustin Stout: [00:05:32] I wouldn't say too far. Did a handful of short films, some student films, a commercial, and a lot of auditions where it seemed promising. But there were 300 other tall white guys that looked like me trying to get that same role, and many of them with more connections in the industry and more experience than I had. So it was difficult.

Ken Majmudar: [00:05:52] What were the years when you were doing acting school and then trying to make a career of acting?

Dustin Stout: [00:05:56] So I graduated high school in 2003, and was at the American Academy of Dramatic Arts from 2003 to 2005. While you're in the Academy, they don't permit you to do any acting work. You have to focus on their program, and rightfully so. It was a Monday through Friday, 8 a.m. to afternoon-ish to two-ish o'clock program, and it was very intensive. So I would say from about 2005, I had an agent fairly quickly, thanks to some connections that I had at my local church. So I had an agent was going out for auditions fairly regularly, several times a week from 2005 till about, I would say 2007 is when I put it on the back burner.

Ken Majmudar: [00:06:38] And then how did you come up with the idea to do? I think your next thing was a marketing agency?

Dustin Stout: [00:06:44] Yeah. So I actually started working as a youth director at a local nonprofit and sat in a conference for nonprofits, and one of the guys that was teaching in this little session was a guy named Michael Hyatt. Michael Hyatt, at the time, famously was one of the biggest bloggers in the world, and he was teaching people how social media and blogging can help you build a platform and reach an audience and grow that audience. And I took his words to heart. It resonated with me. I'm an extrovert. I have this proclivity for technology. I love technology and playing with new things, and the online world was just really starting to bloom in terms of social media. It was peak Facebook for teenagers. It was still cool then. So my audience was teenagers. So I wanted to reach these local teenagers. The best place to do that was social media. And so I started a blog, started learning social media, how to use it to reach my audience. And on that blog that I started, I just started sharing what I was learning, the tactics and techniques to grow your audience, things that I learned from Michael Hyatt and was exploring and expounding upon. So that naturally led to people discovering my blog, reading the things that I had to say. I taught myself web design in that process because I didn't want my blog to look ugly. I wanted it to look good and attractive when people came there. So naturally, people just started asking, hey, seems like you know a lot about this. Could you help my business with this? Hey, who designed your website? Well I did, oh well, could you design ours as well? We need a new one for our business. Sure. And before too long, had enough of those requests, come in and was able to start a small little freelance agency.

Ken Majmudar: [00:08:15] What was that original blog called?

Dustin Stout: [00:08:18] It's still my personal blog. DustinStout.com.

Ken Majmudar: [00:08:21] Oh, okay. So you still have it.

Dustin Stout: [00:08:22] Still there to this day.

Ken Majmudar: [00:08:24] You still write on it and everything?

Dustin Stout: [00:08:25] Not really. I don't think I've published something new in several years. I did a pretty good job at SEO for a long time, and I haven't really published anything new for, I think, 3 or 4 years now, but I have updated a few that lend themselves to being updated, and Google continues to send me traffic, so I'm not mad at that.

Ken Majmudar: [00:08:45] That's neat. Now, have things changed so much that most of those lessons wouldn't be relevant today, or some of those lessons still applicable?

Dustin Stout: [00:08:54] No. I think at the core of everything, I try to always break things down to the fundamentals, and I've seen tons of social platforms and marketing channels come and go, but the core of it is always the same. It's bringing value to your target audience in a way that they can easily consume it and want more of it, and that doesn't change. The medium might change, or the presentation of that content might change, but at the end of the day, it's about serving your target audience with content that helps them, entertains them, or informs them about something they need to know.

Ken Majmudar: [00:09:28] So taking that idea, because I'm a big believer in fundamentals, I do work in investing and we have fundamentals too. It's just different ones. Although I'm trying to do a little more to educate people or to inform them in part to build a following and influence and maybe attract the right type of clients when it's relevant. So I've been learning on a process of asking, even out of curiosity, what you just said makes a ton of sense, but then how do you take it to the next level and actually figure out based on, okay, there's a lot of people that are trying to say, in your case, reach teenagers, maybe less so back then. So how do you create content that informs them or entertains them or whatever and adds value? How do you go about figuring that out?

Dustin Stout: [00:10:10] Yeah, I mean, one of the best ways is obviously to talk to your target audience as a product creator for the last year, that's pretty much all I've been doing is just listening to my audience, talking with them, meeting with them, sending Facebook messages back and forth, communicating with them as much as possible. What are the things you're struggling with? What are some problems you have? What are some things that are entertaining you right now? What's on your Netflix list? Silly little things that seem almost like you would ask just a friend. What are some of the things that are sparking your interest these days? But apart from that, we have access to these same exact networks and tools and places that they're consuming this content. And it's very easy to see what's trending, what are people liking in general, and maybe seeing who is interacting with that content. I'm a big fan. I'm a people watcher. As an actor, observing human nature was key to my success. How well can I actually observe human nature and understand how they think and operate? And the only way to do that is to observe and be quiet. So I live in comments and see what people are saying, and go on Google Trends and see where are the trends going.

Dustin Stout: [00:11:13] So there's a lot of tools out there that you can use to really see what's trending and where culture is spending their time, and you just have to spend the time and effort to care enough about it to go look and see. Because I think a real big problem with not just social media or any content, but also like in business and even investing, we have these ideas about what we think is right and what we think is happening and where we think things are going. But if we're completely disconnected from our target audience and we're just leaning on our own instincts, I think we can make a lot of mistakes and miss the mark. I experienced this very strongly in one of my previous businesses, where my partners became very detached from our target audience, and I was very connected. And so when I wanted to make a decision based on what the target audience is telling me, my partners say, no, we don't think that's right. We want to go this other direction instead. And unfortunately, that led to hundreds of thousands of dollars in loss and ultimately led me towards leaving that company altogether.

Ken Majmudar: [00:12:12] Are there any either tools or books that you would recommend to somebody who's trying to learn the things that you've learned, or be effective in spotting those kinds of opportunities, trends, or needs out there in the marketplace?

Dustin Stout: [00:12:26] Yeah, absolutely. I think several of my favorite books, number one Platform by Michael Hyatt, I think is still highly relevant. It really talks about the fundamentals of building an audience. Jonah Berger has a few different books, and I'd have to root through my library right now to find them. But Jonah Berger is insanely good at communicating how to understand people and trends and reach people and get attention. And one of my favorite all time is by an author named Jay Baer. Jay is a friend of mine back from my social media marketing World Conference days who taught me a lot, and his book Youtility spelled with a Y-O-utility, is extremely good. I mean, it really captures this idea of when you're creating content for your audience, it shouldn't just be content for the sake of content. It should be something that actually gives them a utilitarian value, something that they can use to improve whatever it is they need to improve, whether that's their business or their personal life. It should provide some form of utility. And that book had a big impact on me. So if I had to suggest any, check out Michael Hyatt, check out Jonah Berger and Jay Bear. Jay Baer has written several books since then, all fantastic in the realm of content marketing.

Ken Majmudar: [00:13:37] Awesome. Thanks for sharing that. So you didn't start out when you left acting. You learned the blogging thing; you started creating content that you thought would be adding value. And that led to people coming to you saying, can you build my website? Whatever. So how did that turn into an agency?

Dustin Stout: [00:13:52] So it really taught me that organic content and bringing value to the table via organic content was really powerful. And so I really was just replicating what was working for me. And I was doing that for clients. And I got to the point where I had so many clients coming in that I had to hire a little bit of help. And a lot of us in the digital marketing space, I think, accidentally stumbled into being marketing consultants, because the world that we live in emerged so quickly that we all just fell into it. And that's the story of how my agency started, and I've been doing that for now, almost 15 years.

Ken Majmudar: [00:14:25] So you still have the agency side?

Dustin Stout: [00:14:27] No, I mean, I still have it. But back in March, when I launched Magai, my newest project, I decided I can't take clients anymore. This thing's going to take off. I know it is. So I can't afford the bandwidth to work with clients anymore. And at that time, I didn't have a whole lot of clients because my previous business before that was taking up a lot of my time. And so I only had 2 or 3 different clients that I was working with on the side. So at that point, it was really easy to offload them to other agencies that I was friends with. So yeah, I haven't been doing any consulting or agency work.

Ken Majmudar: [00:14:58] What was the scope of the agency at its largest?

Dustin Stout: [00:15:01] See I also subcontracted for a lot of larger agencies. But me personally, I think the most we've ever had at a given time is 7 or 8 clients at once. And that's also why I needed to leave consulting. I wasn't good at scaling that kind of a business. I'm too much of a perfectionist. I'm too much of a hands-on kind of person, and I suck at delegating the finer details of things. And that's why digital products, I found, were a much better fit for my unique capabilities and talents. Because I could focus on one product, and I could put all my perfectionism into that one product, and it suits me much better.

Ken Majmudar: [00:15:36] Yeah, that makes sense. I also was going to ask before we move on to this next phase, you're not coming from a traditional business background, or not even in a background where you were like, I'm going to start a business and be an entrepreneur or something. So how did you manage the business side of things, whether it was learning them or even basic things like doing books and hiring and taxes and formation of companies and things like that?

Dustin Stout: [00:15:59] Yeah. So it goes back to how I grew up. Like I said, I wasn't a fan of the academic world. Not that I wasn't a fan of it. I am a fan. I think it does a really good job. But I wasn't built for it.

Ken Majmudar: [00:16:10] Yeah, it wasn't for you. Not that you didn't recognize the value.

Dustin Stout: [00:16:13] It wasn't for me. The way that it's set up just didn't suit the way that I experience and learn. I have to be really passionate about what I'm learning. I have to know why I'm learning it for real reasons. And on top of that, growing up, my stepdad, who worked very hard for our family, who basically raised me, he worked very hard but had several instances where he was out of work because the economy was going up and down. He was in the steel industry. Steel industry was very big back in our part of Pennsylvania and that part of the East Coast, and it had some really hard times. And so we fell through many hard times where he was out of work due to nothing of his own fault, just because somebody else decided that he wasn't worth keeping around and he had to be laid off. So with that background, I always had this drive to I'm going to control my own destiny. I don't want somebody else being in charge of it for me, and I'm going to do whatever it takes, learn whatever I need to learn in order to make that happen. And so that combination of drive and wanting to control my destiny, it was obvious that I knew exactly what I needed to learn. I needed to learn how to run a business, needed to learn how to do all the accounting and stuff, and set up the LLCs or the S-Corps or whatever. And thankfully, I started doing that at a time where the internet was full of information.

Ken Majmudar: [00:17:30] Oh, Okay.

Dustin Stout: [00:17:31] So everything I learned, I learned through YouTube videos or blog posts or online courses, and I'm pretty much self-taught in everything that earns me a living.

Ken Majmudar: [00:17:41] So I could iMagaine during the business, you were also say maintaining the books in QuickBooks or something on your own.

Dustin Stout: [00:17:47] Yeah, I mean, there are different times. Once the agency got to the point where I could afford to outsource it, I did.

Ken Majmudar: [00:17:54] Have a bookkeeper.

Dustin Stout: [00:17:55] I had some bookkeeping software, and I was using Bench.co for a long time.

Ken Majmudar: [00:18:01] Oh yeah, sure. I've heard of bench. They're always advertising.

Dustin Stout: [00:18:04] Bench is great. So they pretty much handled everything for me. The investment was, I think I was paying $120 a month, and that was more than reasonable for all the accounting to be handled, but I was at least good enough at delegating for stuff like that, stuff that I absolutely hated. I could outsource that kind of stuff.

Ken Majmudar: [00:18:22] So at what point did you have the idea, oh, instead of doing services like in an agency, I'm going to create SaaS products or information products. I don't know which one you did first, but.

Dustin Stout: [00:18:34] So I tried information products. So as a blogger it's a natural progression to create a paid course or stuff. So I created some paid courses and made a few thousand dollars here and there, but nothing huge. But it came about just randomly one day because I was building websites that was part of our agency. We were actually making websites for people, developing them on WordPress, and I had a couple of friends who were also WordPress developers, and we all had this idea of, oh man, social sharing is such an important part of bloggers getting more viewers and getting their content shared. So why aren't there any good social sharing plugins? So we had this idea to build our own and fix our own need, or fill our own need, so to speak, and we built a social sharing plugin that was a lot faster than all the other ones on the market, a lot smarter because we put our social media marketing prowess into how can we optimize this and make when somebody hits that tweet button that it tweets exactly what we want it to tweet, and when they hit in button, it shares the right image, all these little optimization type things built into it. And nobody else had done that yet. So we created this product, this WordPress plugin, sold it, and it did fairly well really quickly. And that was my first taste of a real SaaS product. And I was like, oh yeah, this is so much easier.

Ken Majmudar: [00:19:48] What was that called?

Dustin Stout: [00:19:49] Product is called Social Warfare okay.

Ken Majmudar: [00:19:52] And it's still around.

Dustin Stout: [00:19:53] It is still around. Yes. I'm not sure what the status or state of it is after I left in 2020, but it is still out there. I do come across the off website that I'm like, oh, that's Social Warfare. I recognize those buttons.

Ken Majmudar: [00:20:06] So how long was that time that you were involved with that company?

Dustin Stout: [00:20:09] We founded? The company launched the product in 2014, and it's in 2020 that I silently exited.

Ken Majmudar: [00:20:16] Got it. Okay. And what was next after that?

Dustin Stout: [00:20:19] After that, I had two what I thought to be really great product ideas. The unfortunate part of that was these ideas were massive in scope, fairly technically complex, and would probably have cost over $1 million to build. So I myself am not a developer. I know enough about code to be dangerous, but I'm more of a front-end designer, so UI and UX is more my strength. So I needed to outsource the development of these products. So because I didn't have over $1 million to just hand over to a development team, I basically had to whittle the product down. What is the minimum amount of things that I can build that will hopefully allow people to get some value, want to pay for it, and fund the future development of these two products that I had an idea for, I had a couple investors who helped me get those off the ground financially, and unfortunately, those MVP's were just not MVP enough.

Ken Majmudar: [00:21:16] Okay. Got it.

Dustin Stout: [00:21:18] They did not find the product market fit. We could not get a value proposition that was compelling enough for people to pay for it. So those two products were an absolute failure.

Ken Majmudar: [00:21:28] Got it. And this will take us from 2020, roughly to?

Dustin Stout: [00:21:32] To take us from about 2020 to about the near end of 2022, when I was at my lowest and thinking I had nothing left to give. I was applying for jobs, something I hadn't done in over ten years. Writing a resume. I was like, this is it. I got to go find a job because nothing I'm doing is working anymore.

Ken Majmudar: [00:21:52] Okay, so now without giving it away, we met through me finding out about Magai, which is your current thing and it's in the AI space. Obviously it's Magai. M-A-G-A-I. So tell us about what Magai is and where you got the idea for it and what's going on with Magai.

Dustin Stout: [00:22:14] So it was the end of 2022. It was November. And I remember this because it was the day before Thanksgiving that I was desperately trying to figure out how can I build some more features myself, because I had no more cash to pay developers? How can I build features myself that will add more value to this dying product that nobody is buying? And it seemed like AI was catching on a little bit. There was some buzz starting to generate, and I'd been using AI products for years now. The Jasper's and the copy AI, and all these copywriting tools. As a marketer.

Ken Majmudar: [00:22:47] Those things were there before ChatGPT.

Dustin Stout: [00:22:50] Yeah, they were there well before ChatGPT. But I always had issues with them, and my issues with them were they were all predicated on templates, blog post template, a Facebook ad template, and they've got hundreds and thousands of templates and you always had to, number one, figure out all the templates, which one is going to suit my needs. And then you had to fit your idea into somebody else's way of thinking, because these templates are based on how these people thought about these particular pieces of content. And so it never seemed to work for me. And I think that's primarily why these companies never hit mass market. They were successful in their own right, but they never really took off in the mainstream. They were still the marketers best kept secret. But as I'm building this product, I'm like, well, OpenAI has this API which all these copywriting tools are utilizing. That's how they get it. So maybe I can build some generative text AI into this product of mine and that'll be more attractive.

Ken Majmudar: [00:23:41] So since you mentioned it, what was the product that didn't quite make it that you morphed into this than it sounds like that's what happened.

Dustin Stout: [00:23:49] It's called So Visual. Sovisual.co. So the idea was it was a social media graphics generator, and all you had to do was fill out a form and it would create all the optimized graphics for the different social platforms. A template library, you pick a template, type your words, and out pops 4 different sized images. But people weren't leaving Canva. They were comfortable with Canva or Photoshop, and that worked for them, and it was hard to pull them away from that workflow. So I thought, well, if they're creating social media graphics, maybe if I have like a little chat box with an AI that you can just ask it, say like create some Facebook posts for me and it would give you some Facebook posts. Maybe that would help as I'm learning and building this little widget for So Visual. That's when ChatGPT hit the scene. And ChatGPT, the minute that I saw it, I went and I signed up and I'm like, well, this is interesting. And I typed in my first prompt and I realized that you could just tell this AI what you want in normal language, just like you're talking to a human and it would give you that thing. It would produce that thing that you want to produce. The light bulb went on. I was like, okay, this is going to change the entire world. All these copywriting tools that I've been using, the problem I've been having with them is they've had these barriers to entry, these it's like the Netflix problem. You log into Netflix, there's a thousand things to do and you don't know which one to pick. And maybe it's going to be a good output. Maybe it's not. But with ChatGPT all of that was wiped away. Tell the AI what you want. If you don't like it, tell it and then it'll do it again.

Dustin Stout: [00:25:19] So it was this complete unlock for me. So I started using it prolifically and very quickly as the entrepreneur in me, I noticed, oh, this is annoying. Or why doesn't it do that? Or I wish it did this other thing. And this is frustrating that I have to go through these other two steps to get this over here. So these little quality of life, things started to pile up and I was like, you know what? They have an API. I just learned how to build a simple AI integration. Maybe I can make a better version of ChatGPT to solve some of these pain points that I was encountering and that other peers were voicing as well. So that's where the bug started. So I started building this product, this better version of ChatGPT. And as I started to build it, it really snowballed into something much bigger than just a better version of ChatGPT. I realized that all these other AI tools were coming out, the image generators, and I was like, well, ChatGPT doesn't do images, so maybe I can build an image generator in here as well, so people don't have to have multiple subscriptions to get multiple AI tools, and it just kept building from there. And then Anthropic became a household name within the AI community, and they're competing with ChatGPT. And I'm like, well, ChatGPT will never give people access to a competing model, but I can I can give people that choice because Anthropic had some advantages over OpenAI and then Google and Meta, They all started joining, and I realized very quickly, this is going to be a huge marketplace of tools, AI models, and Magai can really be the catch all, the one platform to rule them all.

Ken Majmudar: [00:26:52] And so when did Magai launch?

Dustin Stout: [00:26:55] I remember it very vividly because it took me eight weeks from start to finish to build it. I had a few beta testers go through, and then finally I was at the point I was like, okay, I think this is good enough to launch. I'm not going to sit here and delay it any further, because this is more than good enough to get out into the marketplace and start getting feedback. And it was March 31st. I was like, okay, well, I'll just launch the public signup URL tomorrow. I was like, oh, wait, tomorrow's April 1st. I can't launch something on April 1st. Everybody will think it's a joke. So I was like, you know what, I'm just gonna tweet it out now. So I did March 31st last year, decided to let people sign up publicly for Magai. And I've been building in public, sharing and teasing it out since then, so my audience was prepped for it at that point.

Ken Majmudar: [00:27:37] Okay, so now tell us about Magai what people are using it for. What's the scale of it right now, what the feedback is, who wants to use it?

Dustin Stout: [00:27:46] Magai is the all in one AI tool. It allows you to access the best AI models from companies like OpenAI. You get ChatGPT3.5, ChatGPT4. You get access to Anthropic, their biggest competitor whose AI models, Claude Instant and Claude 2.1, are extremely capable and better in some ways than OpenAI's ChatGPT. We also have other models in there from meta, from some other lesser known companies, but there are at this point about eight different selections to choose from. So you can compare the models and use the right one for your specific needs. And we also have an auto model where if you don't know which model is the right model for what you want, check auto and auto will, based on your prompt, it'll pick the right AI model for whatever it is you need. On top of that, we have image generators. So we have at the moment 4 different image generators that allow you to generate visual content right inside the app without having to have a subscription to all these different products. So you can have access to all the world's best AI models and invite your team in. Because we do have team accounts where you can invite team members all under the same plan from one access point, one beautiful user interface, and it solves a lot of the pain points that pretty much everybody who uses ChatGPT expresses.

Dustin Stout: [00:29:02] I can't search for things, I can't put things into folders. I can't separate by clients or brands. We have all of this stuff built in because it was made specifically for business owners, for agency owners. And surprisingly, the real big surprise for me as the product creator was, I never realized how helpful it would be in an academic setting. Not having that experience in a traditional academic setting. I realized one day that a lot of universities started to adopt Magai into what they were doing, and I got an email one day from a man named John Behrens, who is the director of Digital Things. I don't know his official title at the University of Notre Dame, and he is using it to help him teach about generative AI. So all these universities who have programs and classes aimed at generative AI, can use Magai to give access to their students so that their students can explore the different AI models in one interface, and the teacher can manage all of that in one easy to use way. And I had never realized that. So I've been working very closely with John and some other universities who are deploying it in their settings as well. So yeah, it's doing fairly well. I would say in our nine little months of existence, I've got close to 2000 users and financially doing pretty well as the 30,000 foot driven achiever that I am.

Dustin Stout: [00:30:22] It's never good enough, but I think we can do better. And we've done all of this with almost zero marketing, despite being a marketer for the last 15 years. I've been so immersed in building the product and making sure that the product is as good as it can be, and listening to users that I really haven't had, the energy and the time to do any of that traditional marketing stuff that I would do for clients like SEO and social media content and lead magnets and email marketing. I've been really terrible at it, but that's because I just I wanted to give everything to the product and make sure it was as good as it can be before, and hopefully hire some people that can take that over. And we just were able to. So in the last two months, I was able to hire two part time developers and a support person a couple of weeks ago to help me do customer support. So now that those roles are filled, once we get them like a well-oiled machine, then I can kick hard into the marketing gear and really let this thing take off.

Ken Majmudar: [00:31:13] So what's the model for Magai like if somebody's listening and they want to subscribe, what does it cost? I guess it sounds like you don't have to then subscribe to get the OpenAI paid version.

Dustin Stout: [00:31:24] Absolutely right. So we have various plans and all the plans are based on three factors. How many words do you need to generate, how many team members do you need, and how many workspaces or brands do you need to separate now? Workspace or a brand is how we silo off the app. So if you have one client that you're generating content for and then your own business that you're generating content for, you don't always want those to intermingle. So you can have different workspaces to silo those off. So depending on how many of those you need, you might pick a higher plan that has higher number. And depending on how many AI words that you want to generate in a given month, you might have a higher plan. So the bottom plan at the moment it's going away soon is $9 a month. We're going to replace that soon with a $19 a month plan. Start you off at 20,000 words, but you don't get access to any image generating. And that's the personal one person plan. From there, you can go up to the $29 a month plan that gives you 65,000 generated words. And if you pick one of the lower end models, one of the not so advanced models, you can actually get up to 650,000 words out of it. But then you could also invite five team members for no additional cost and share things between the team and collaborate in that way. And then it just goes up there based on team members and brands and words. I think our highest plan at the moment, nobody's asked for more just yet, but our highest plan is only $249 a month, and that gets you over a million words, over 100 users, and pretty much nobody has maxed that out. University of Notre Dame is getting close, but they haven't quite yet.

Ken Majmudar: [00:32:52] Now let's because this is helpful for somebody just learning about AI or starting to use it. So most people are familiar with ChatGPT. Can you give a very high level what you've learned about what ChatGPT is better at versus Claude, versus maybe some of the other models and when to use what?

Dustin Stout: [00:33:09] So it's always a big question, because who can be expected to keep up with all these things, especially with all the advancements being made. So I love sharing some of my observations. Now, there's a lot of benchmark studies out there. And the thing with benchmark studies that compare these models is they can be biased and they can be unreflective of practical everyday use. And so oftentimes what I tell people is this is my experience with these models and talking to other users with the models as well. What I have found is that the GPT models, let's say GPT 3.5 or GPT 4, they tend to be better at creative tasks. So coming up with novel ideas or new information or creating something from scratch, it's much better at that phase of just the creation ideas coming up with new stuff. But where it seems to struggle is when you get an established foundation of information, and you have ChatGPT try to do stuff with that information. Sometimes what it does is, technical term called hallucination, so it will make stuff up that is not true. You've probably heard a lot of these news stories where lawyers use ChatGPT to write up these fake cases, and they thought they were real cases because they thought ChatGPT was just spitting out information like Google. But that's not how these AI models work.

Dustin Stout: [00:34:30] They are crafting words due to a very complex set of algorithms, and try to place words together that make sense, not necessarily searching a database and pulling information. That's not how they work. So ChatGPT tends to be a little more creative, and that can get it into trouble when you have pre-established information. Whereas Claude by a company called Anthropic, Claude tends to be a little safer, a little more conservative, and its hallucination, when you feed it information or like a foundational set of info, so you upload a PDF or some document. It is less likely, in my experience, and the people I've talked to, to hallucinate things that weren't there to begin with or to change details about things. So in practicality, it's great to be able to say, start with ChatGPT when you're coming up with ideas, bullet points, and building your foundation of, say, a blog post or an article or a landing page. But once you have everything there that you want to be there, it's good to be able to switch over to something like Claude. Where Claude is then going to continue that conversation, but it's not going to change any of the details. It's not going to make stuff up unless you ask it to, or at least it's going to be less likely to hallucinate things.

Dustin Stout: [00:35:41] So those are really the main differences. Claude has a greater memory. So many people don't know that AI's have a limited memory because ChatGPT doesn't tell you this. But for most people, using ChatGPT, the conversation as soon as it gets longer than 6000 words, ChatGPT is actively forgetting the earlier messages in the conversation because it doesn't have the ability to process more than 6000 words per conversation, so it's removing stuff from the memory earlier in the conversation, which is why some people have experienced this. Oh wait, why did you say that? Didn't I tell you something differently up here? And yes, it did forget. Whereas Claude, from the day that it was released had a, so as opposed to ChatGPT forced 6000 word memory, Claude had a 96,000 word memory. And now ChatGPT has upped their game a little bit, and so has Claude. Claude has now doubled its memory so it can remember almost 200,000 words in a conversation. I don't know anybody who's going to have that deep a conversation with AI, but it often is the best one to choose when you're dealing with a long conversation, a deep conversation, or have a foundation of data that you want it to work with. Now, the other models that we have are usually, they're mostly in there because they have a lower cost, so they don't cost as much to use for us and for the user because we don't charge them as much from their monthly usage.

Dustin Stout: [00:37:02] But they're also unmoderated. So OpenAI and Anthropic have taken a stance of, here's the things that you can use our AI models for, and here's the things you can't. And sometimes that involves things that are not necessarily nefarious or have malicious intent. Let's say an author, a fiction author is writing a battle sequence and an orc gets stabbed with a spear. Both OpenAI and Anthropic will refuse to work with that content because it violates their content policy. They don't want any sort of violent content whatsoever. They don't want anything dangerous or harmful. So if you have these needs to work with content that is maybe outside of the guidelines of their acceptable use, then you would need an unmoderated AI model to help you do that. And that's why we have currently PaLM 2 by Google, Llama 2 by Meta. And we're just getting ready to introduce Google Gemini Pro into Magai, which is a big deal. It's going to replace PaLM 2 and our interface. But these models are unmoderated. They will pretty much work with whatever content you have and whatever needs you have.

Ken Majmudar: [00:38:07] You hear less about Gemini and Llama, so give us a comparison of those two versus Claude and ChatGPT.

Dustin Stout: [00:38:15] So Llama, I would say by Meta is gaining some steam. Obviously, there's some very smart people who work at Meta in terms of capabilities and how advanced it is. I would say it's about equal to GPT 3.5. So the very first version of ChatGPT that the world was introduced to, the Meta Llama 2 is about as good as that.

Ken Majmudar: [00:38:35] And I think that open sourced it. Right?

Dustin Stout: [00:38:37] Yeah, it's completely open source. So anybody can use it, download it, install it in their local environment if they understand how to do that from a development standpoint.

Ken Majmudar: [00:38:45] Which is pretty awesome.

Dustin Stout: [00:38:46] Yeah, it's pretty great.

Ken Majmudar: [00:38:48] There's a model as good as ChatGPT 3 available and you can just have it for free.

Dustin Stout: [00:38:51] Yeah. And so the open source community has taken that PaLM 2. Also an open-source model from Google I would say about equal to Meta in terms of capabilities. But Gemini Pro actually made a very big splash a few weeks ago. I don't know if you saw the rounds.

Ken Majmudar: [00:39:07] I did see something about that. Yes.

Dustin Stout: [00:39:10] All Google's hype in this video that was blowing people's minds of how powerful it can be.

Ken Majmudar: [00:39:15] I think I heard some criticism, though, that they may be exaggerated.

Dustin Stout: [00:39:19] Turns out that video was mostly fake. It was more marketing and hopeful thinking about where it will eventually go than where it currently is now. But I can say this after testing it for the last several weeks now, it's been opened to developers. I've been testing it inside of Magai by myself and some of my team members, and I can honestly say it's about as good as GPT 4. It's at that level of GPT 4. Now, they have different benchmark studies, like I said, and the benchmark studies, Google will tell you that Gemini did better than GPT 4.

Ken Majmudar: [00:39:50] Yeah, you can pick probably certain things and whatever. But I mean.

Dustin Stout: [00:39:54] Of course they're going to say that, but I have used it in practical applications for practical types of content that I see frequently at my users needing help with. And I think it is as good, if not better than GPT 4 in many different common tasks. And not only that, it's one tenth the cost. So for our users, I'm super excited to be able to introduce that because they're going to get a lot more content for their buck, and they're going to get a capable model like Gemini Pro, backed by a company that if anybody has the ability to beat OpenAI at their game, it's probably Google, so.

Ken Majmudar: [00:40:30] Yeah, I mean, I don't think a lot of people know, I do know, or I mean, I have a very good idea of how the technicalities of the models work and stuff. I have a computer science background, and it seems like there's going to be a arms race now at the model level.

Dustin Stout: [00:40:45] Oh yeah, there truly is. And what I love about Magai is that it benefits. Doesn't matter who wins or who comes ahead, or the more that these companies push each other to be better. Every Magai user benefits because they always will get the best of the best.

Ken Majmudar: [00:40:59] What about the image side of things? Which ones are you giving access to and how does that work?

Dustin Stout: [00:41:05] So currently we have inside of Magai Image Editor. We have Dall-E 3 which is OpenAI's most advanced AI image generator. Very good model. Doesn't do photorealism extremely well, but I can only assume that they're going to get much better. We have Leonardo AI, which is a company built upon basically fine-tuned versions of Stable Diffusion. Stable Diffusion is the big open-source image model out there by a company called Stability AI. And we do have a raw version of Stable Diffusion XL, their most advanced version of Stable Diffusion. But Leonardo is actually my favorite AI image company right now because they have built a platform that is conducive to people fine tuning these Stable Diffusion models for very specific use cases, and they're very community oriented and they're very developer friendly. So, they have an API, and that's why we were able to integrate with them and their ability to just crowdsource basically better versions of the models and make these models better and better is extremely fun to watch. So we have access to their platform, and I'm doing some work this week to get some deeper integration, because they're doing a lot of interesting things with image to image, and they've got motion, they've got some video, image to video that they're working on and some really cool things that their team is building and they're building in a way that is developer friendly. And that's what I love about them. Now we do have Midjourney integration, but it seems like Midjourney is a little hostile towards developers and doesn't really want anyone integrating it into their platform, so we probably will be removing it soon. Unfortunately, it is probably the best image generator in the world right now, as far as quality output goes. The trouble is you have to do it in discord.

Ken Majmudar: [00:42:46] Yeah, I didn't even understand that. But it's really funny, I've tried playing a little bit with the images. It was so simple what I wanted. So I was creating a social media post of a restaurant review. At the end I wanted to say this is my score. I gave it a 9.5 or whatever. I said, make me a seal that says final score or something and it kept messing it up. The 9.5 had multiple dots, and I kept prompting it to fix whatever I saw, and it would create other errors. It's such a simple thing of a seal with a score in the middle, and maybe something that says final score in the name of the restaurant or something, and it couldn't do it after many prompts.

Dustin Stout: [00:43:22] Yeah. So image AI is not quite great at text on images yet to creating actual text. Now Dall-E 3 is pretty good at it. You can actually generate some pretty good stuff. Midjourney version six will generate some text. You have to try it a lot and give it a lot of chances. Eventually it'll get something right. And there is one brand new company that I recently saw that is actually making some real headway in putting text on images through AI. I can't remember the name right now. They don't have an API. They're not ready for public yet, but it's definitely one I'm keeping an eye on. I think once Dall-E started doing it and giving the world a taste of AI can put text on images. The demand has really spiked for that because before that there was not even a chance you were going to get text on an image through AI.

Ken Majmudar: [00:44:09] What's the one that does work, do you think?

Dustin Stout: [00:44:12] I'm trying to remember what it was. I came across it the other day. I want to say, not Lensa. Man, you know, I'll remember it in the middle of the night. I'll wake up in a cold sweat. Sorry, I just remembered it. It's called LensGo.

Ken Majmudar: [00:44:24] LensGo.

Dustin Stout: [00:44:26] L-E-N-S-G-O. I think, dot AI. They're doing some interesting things.

Ken Majmudar: [00:44:31] So where do you see things going now underlying models that you guys use? And then with Magai what's the roadmap?

Dustin Stout: [00:44:38] Our roadmap is to really focus on refining a lot of the tools and feature sets in there. We have a very aggressive roadmap for this year. It's hard to plan too far ahead because you don't know quite where the industry is going to go. But high up on my list personally is getting some video tools in there. There are two companies that have been asking me to integrate with them, their AI video generation. So I want to work that in for sure. Really heavily focused. I put my two developers there. Task for this month is to make Magai multimodal. If I can pull out my Baptist preacher side.

Ken Majmudar: [00:45:12] Oh yeah, because I guess OpenAI said they're multimodal now. Multimodal meaning in one prompt you can have images and text and maybe audio, I don't know, video, audio.

Dustin Stout: [00:45:23] Yeah. So the AI will be able to both interpret images. So you can upload an image and the AI will see it in a sense. Will be able to interpret the visual, pull text from it, pick out subjects, colors, everything it'll be able to in all sense of the word, see the image and work with it, but also it'll be able to output. So right within the chat you can say produce an image of blah blah blah blah blah. And it'll actually generate an image right inside the chat. Instead of having two different sections of the app, you can do everything from one place. You can generate images, you can upload images and work with visual content as well right from inside the chat.

Ken Majmudar: [00:45:56] What about the underlying this battle between all these companies? I know you said you'll benefit, but I mean, they'll all need a lot of funding, I guess. And some of them are big companies versus companies that might need to raise money. I guess on the Upcomer side is the OpenAI and the Claude Anthropic. And then you have Google and Facebook or Meta funding these big models. Any thoughts about where things are heading? If we fast forward six months, a year, two years, I mean, things are going so fast. My head is spinning. I feel like I haven't capacity to even handle it. There's just every day you turn. I mean, there's a site called Future Tools that I look at sometimes, and there's thousands and thousands of applications that folks have created, I guess, and more and more coming every day, every week.

Dustin Stout: [00:46:38] Like I said, I think it's hard to judge too far in advance, but I do know OpenAI has been pretty open about their goal to produce AGI. So artificial general intelligence, basically a self-aware AI that can do anything and everything. So I think they're all chasing that dream, but I don't think it's going to be within the next two or even three years. I think it might be much further down the road.

Ken Majmudar: [00:47:02] Yeah, I'd be surprised.

Dustin Stout: [00:47:04] So I think in the short terms, what we can see is more llms that are multimodal. I think that's the future. Everyone wants to chase that multimodal. They don't want to have multiple different apps and things to do, both images and text. I think that's definitely the short-term future is they're all going to chase the multimodal aspect of these llms. And I think they're also going to work on context length, understanding context and having access to real time information, because that's also a very high in demand. People want not just old data that it's been trained on, but they want to be able to pull real time internet access stats. And one of our models, actually two of our models inside of Magai are by a company called Perplexity, and their AI models do have internet access. So you can ask it about what football game is on today and who's winning. And it can do a real time web search and find that data and information and use it in real time. So I think a lot more models are going to go that direction. And this year is going to be a really fascinating to see who does the best job at getting there.

Ken Majmudar: [00:48:05] And from your perspective, as someone who runs a small business and is in marketing and say there's other people like myself also wear similar hats, have you found some really powerful things that, whether you're using Magai or one of these models directly, I don't know whether they might be prompts or just ideas or things where you're like, wow, this has just been a game changer for me as an entrepreneur or as a marketer.

Dustin Stout: [00:48:28] Yeah, I think the multimodal aspect, being able to upload not just images, but also PDFs. I mean, that's something that Magai has had for months, the ability to upload a PDF, but the ability to upload, say, a CSV of information. I do this frequently now as I want to monitor the pulse of my business. I'm trying to monitor a bunch of different data factors. And I don't know about you, but I'm not a huge fan of sitting and looking at spreadsheet for long periods of time. But with these AI models, you can upload a spreadsheet and it can view all the information on the spreadsheet and give you real time analysis of that information. And so I think that's extremely helpful for someone like me who wants to analyze big numbers of how people are subscribing and what the churn rate is, and revenue and all that fun stuff, and then be able to offer insights on top of that and help me extract data. And I think one of the cool stories that I got from Magai user that led me down this path of not just uploading content, but doing multiple things with that content. So I had a director, he's a more of a theater director who's using Magai, and he said, I just got to tell you about this thing that I did the other day. It was so cool. He uploaded a script to a play that he would be directing, and he was using Claude because it had the biggest context length. He had the AI read the whole script and he said, I want you to make me a list of all the characters ranked by how many words they speak throughout the play, and it did create a quick list of all the characters, ranked them by number of words they speak in the play.

Dustin Stout: [00:49:55] He said, now I want you to create a props list that I can hand off to my props department so they can get all the props ready. And it did. It has the whole play, pulled out every single prop mention, and he had a nice formatted list that he hands off to his props department and they said, okay, give me a character analysis for each one of the characters with a brief description so that I know which traits to look for in an actor to play that character. And instantly he had that. So he knew exactly what to look for in his actors. And then he said he gave the AI the first day of rehearsal, and then the opening night, the dates. And he said, can you create a rehearsal schedule based on this timeline, so that we schedule all the actors and all the scenes that they need to get with sufficient time to cover it before opening night. And it did. He said it did a perfect job of creating that initial schedule, and I was like, wow, like, talk about diversity in capabilities, starting with just this little simple task and another one a little more complex, another one is more complex. And he did this all within an hour. He said this work would have taken me a week to do on my own, but now I was able to use your AI and all of these things got done in an hour.

Ken Majmudar: [00:51:04] Yeah, and this is why I expect that there's going to be a productivity boom. I've presented on this from a financial investment context. Normally in the world productivity goes up half a percent or a percent a year. But if it starts going up several percent a year or more, it's going to be the greatest period of wealth creation maybe we've ever seen in the history of human times.

Dustin Stout: [00:51:25] Absolutely.

Dustin Stout: [00:51:26] And another quick story that just fascinates me. We have one woman who is, she's a fantastic teacher, she's a digital marketer and a digital marketing trainer. But she jumped into AI and she's using Magai. And as an experiment, she decided to use Magai to create a business plan for her and to come up with an idea for a business. She started an Etsy shop, and then she's using the image generator in Magai to produce these products for her Etsy shop. And they're like handbags and t shirts and mugs and notebooks. And she created an entire business plan, marketing content and product graphics to print on these products, all with AI. And she was able to do it a fraction of the time that she would have been able to do it otherwise. And without having to hire any designers or anything, she was able to do all of that through Magai and create another side income source, just in a matter of a few hours. It's fantastic.

Ken Majmudar: [00:52:19] That's neat. Well, when we put the episode out in the show notes, we'll put a link for those that want to explore or sign up for Magai. And Dustin, any last words? Maybe any new books that you're reading or any new shows or movies? You're a movie buff, so anything that you'd want to recommend to listeners.

Dustin Stout: [00:52:36] So I'm actually not much of a reader. So when I do read, you know that it's serious. And I literally just got this book in the mail.

Ken Majmudar: [00:52:44] Oh yes.

Dustin Stout: [00:52:45] Yesterday.

Ken Majmudar: [00:52:46] I heard Ben Hardy speak, so I'm trying to ten-x my firm. Part of the idea is because of that book title.

Dustin Stout: [00:52:51] Yeah, so I'm looking forward to diving into that.

Ken Majmudar: [00:52:54] Just for those that are, because we're not going to have a video podcast. It's called "10X is Easier Than 2X" by?

Dustin Stout: [00:53:00] Dan Sullivan.

Ken Majmudar: [00:53:01] Dan Sullivan and Ben Hardy.

Dustin Stout: [00:53:04] Looking forward to that book. But one of the things that I always try to leave people with is our whole culture is going through this shift and learning how to use these AI tools, and no matter where you're at on the spectrum of brand new, haven't started using AI yet, to someone like me and you, it's every day, all day for us.

Ken Majmudar: [00:53:23] I'm not, actually. I still need to. I almost think what'll happen is that companies will probably end up having just one person that's the AI expert evangelist, because I think it takes some effort to go into every function, process, whatever, and be like, okay, how can I apply AI going forward and make it part of our process so that it's done much more efficiently or much better or whatever? It's not obvious, but like I know a guy, he's an entrepreneur, his company had hundreds of people, and I think his plan over time he's going to try to keep growing the company, but he feels like he'll be able to maybe double or triple his revenue with less people than he has now. And it would have probably meant he would have doubled his workforce instead. It'll actually go down a little bit, but they'll be able to do more.

Dustin Stout: [00:54:07] Yeah, I think that's totally possible. But one thing that I like to emphasize is that the only way we're going to really figure out the maximum potential of this technology for all of us is to actually just go play with it, go use it. When Google first came around, many people didn't really understand what they were going to use Google for.

Ken Majmudar: [00:54:26] As hard as it may be to imagine today for people.

Dustin Stout: [00:54:29] Yeah, it's hard to imagine. Everything's a Google search now. But before Google came out, I was around when that happened. Before Google, we didn't know why we would use it. Why would I go Google something when I can go open up a telephone book, it's right here. Or whether that's looking up a business or looking up maps, directions. We went through this whole cultural learning curve with how do we use search for our daily lives in Google, for our daily lives and the ramifications and side effects of that? And I think we're going through an even bigger shift in culture and in human behavior, and how we operate day to day with AI. And the only way to shorten your learning curve is to just go in and use it. The more you use it, the more you'll realize you can use it for. So just go in and play.

Ken Majmudar: [00:55:15] Great. Well, I'll make that the last word. Thanks for being on the podcast and it's been great having you. I enjoyed the conversation.

Dustin Stout: [00:55:22] Likewise, Ken.

Ken Majmudar: [00:55:24] For more episodes of Compound Ideas, visit our website at CompoundIdeasShow.com for more insights like these and to contribute to the conversation, go to my firm website at Ridgewoodinvestments.com and click on the link to insights at the top of the page. Also, please follow me on social media. I'm under Ken Majumder. On LinkedIn @KenMajmudar, Twitter @KMajmudar, Instagram @KenMajmudar, and on YouTube we have a new YouTube channel Investing with Ken Majmudar.

Narrator: [00:56:08] Ken Majumdar is the founder of Ridgewood Investments and several other affiliated companies. All opinions expressed by Ken and podcast guests are solely their own opinions and do not reflect the opinion of Ridgewood Investments or any of its affiliates. This podcast is for informational purposes only and should not be relied upon as basis for investment decisions. Clients of Ridgewood Investments and its affiliates may maintain positions in the securities discussed in this podcast.