My Biggest Regret: I Prioritized My Startups Over My Kids

Sean Weisbrot: Keith Smith is the co-founder and CEO of me for you, a company that helps celebrities, creators to create authorized AI chatbots. And this was a really interesting conversation because we talked about what it means to be a creator and what it means to use AI and how AI can bridge the gap between yourself and your audience in a way that allows them to feel like they know you. Without you needing to spend all of the time necessary to have a conversation with them so that you can maximize the benefits for the community while maximizing the revenue for your company. It was a very interesting conversation, and I know you're going to like it. So why don't we get to it now. What made you want to become an entrepreneur?

Keith Smith: So my entrepreneurial journey goes back, uh, really to when I was about seven years old. Uh, my parents had friends that were, uh, that were in my view, entrepreneurs, and they, and they, they were, but they had started a clothing store and, uh, these were very good family friends. They would come over every Thursday night. Play Rook with my parents. And as a little kid I would, I would anxiously await for Robin and Lauren to come over 'cause they would tell stories about what it was like to be starting this company of theirs. And then they would continue to expand on this and talked about how when they opened a second store and a third store and how they would go off to market and figure out what was the new fashion trend of the, of the year and be able to try to make sure they bought the right kinds of, uh, of, of fashion. Uh, and what would happen if they didn't buy the right fashion and it didn't sell. That sort of thing. I was just absolutely. Fascinated. So I took a bit of a detour. Uh, when I was in high school. I decided that I was going to be a, uh, a pastor. Um, and, uh, and so I, I went to bible college. I was planning to be a pastor. And then when I graduated college, or when I almost graduated college, I should say, uh, as a side note, I flunked out of college, which is a badge of honor at this point in time. I had three weeks to go. In Bible college and I decided this was not the path for me. I'm going back to plan A. And uh, and I think that combined with the fact that I, I figured out pretty early on in my twenties that I'm highly unemployable. And, and that made me, uh, realize that there was kind of no other option other than to go out and start starting companies. So I started doing that in my early twenties and, uh, currently on my sixth startup now.

Sean Weisbrot: What gave you the idea that getting into artificial intelligence was the right thing to do for you at this time?

Keith Smith: I, I've been thinking about AI for a long time. Uh, it's been being talked about obviously for, for decades. Uh, it, it feels, uh, a little bit, it always felt like throughout my career, a little bit like cold fusion that it was, it was always just 10 years away and, and it was gonna fundamentally change society as we know it. And, and so I've been, I've been paying attention. And, and started using ML extensively at my last startup. And we used it in order to be able to develop a, a really innovative, uh, and, and, and new risk model in order to be able to risk assess very, very small businesses that are, that are online sellers. And that worked incredibly well. And so that was really my first foray where I spent nine years, uh, working with our development team in order to be able to build out those ML models and seeing how. Absolutely amazingly good. They were, uh, not just by themselves and not just automatically, but with a lot of work. And they were frankly better at the humans at making those kinds of decisions. And, uh, and so that was my first real practical application of, of, of machine learning. And then, uh, seeing in early 2022 when I started to really dig into. What the latest developments were around the latest AI models and specifically, uh, with with GPT and, uh, and, and and seeing what open AI had developed as well as some others. But open AI was clearly head and shoulders above the rest, even in early 2022. That was clear. And so that was the moment where, where the light bulb went off for me, and I said, the, that, that that decade away moment has actually arrived and AI is actually going to be a real thing and it is going to fundamentally change how we as humans. Interact with computers and I need to make a change. And at the time I was running a FinTech company and uh, and I asked myself the question, if I continue to innovate and do you know, really innovative, interesting things in this particular company, in this particular sector that I'm in, will I make a positive impact in the world? And, and I didn't like the answer that I got back. And so that was a moment where I said, it is time for me to start on a new journey. And that's when I started doing. Customer explorers started in the stupid phase like we all do when we're, when we're starting a founding journey, started asking a lot of questions and meeting with a lot of folks, uh, that are potential customers to be able to get a little bit smarter.

Sean Weisbrot: Yeah, I like that you did that. I also recognized that AI was really going to be important. I would say probably a year ago. Um, I, I try my best to get to a trend before. The wave is like ready to start moving where like I got involved in blockchain in 2015 and some people might go, oh, that's like, you know, five years late. But actually it was, you know, it's still early. Um, so I was quite early on that. Yeah. And I saw how. Companies were looking to use automations and integrations. You see companies like Zapier and all that. And I saw AI and my own startup had fallen apart. And so I started to have the time and the energy to look at, well, what can I do next? Um, 'cause originally we were trying to build a team, enterprise team collaboration platform. And we had thought of using ai, but we had also been pulled into like the metaverse because of our investors and. All sorts of stuff. So when that company fell apart, I was then free to explore what's actually going on in the world. 'cause I had spent four and a half years kind of head down looking at what was in front of me and what we could build. And the AI that we were thinking of putting into our system at that time was years away. Because there was no open AI at the time. I mean, yes, there's open ai Yeah. But they were working privately. So there, there was no product to, to, you know, work with, and we thought we'd need to raise a hundred million dollars to build out our training models and, and get our data and cleaned up and, and be able to serve our customers with it. And then chat to D came along literally like six months after the company fell apart, and I was like, oh. We could probably have just used Chatt, PT and, and not needed the a hundred million dollars. I mean, obviously we didn't raise that kind of money. If we had, you know, I'd probably still be running the company. But, um, but it was quite interesting to kind of reflect on the difference between before Chatt PT and after chat TPT came out, how much easier and faster and cheaper it would've been to get us to a really important part in our business. But it was too late, unfortunately. So I was able to see, okay, there's something really interesting here, and that was literally a year ago. If you think about it, Chatt BT came out, I think it was November of last year. Yeah,

Keith Smith: November of last year, exactly. So

Sean Weisbrot: I'd say the, the minute CHATT came out, so, so like I, I realized in 2018 that AI was important, but it wasn't until a year ago that chat TT was, was out and therefore AI was tangible. And so I started thinking this is going to be the next wave. It's going to last probably 10 or 15 years. Very few people understand AI to begin with. Chatt PT will maybe make it easier for people to get to understand what it is, but what can I do to make myself a place in, in the, the wave? And so I thought, I'm gonna provide services to companies that wanna use ai. And so I have, uh, three companies, two that I've invested in, one that I've, that's my own. That are all working towards making it possible for companies to have the extra liquidity to think about automations and integrations, and then to actually hire AI talent or implement AI through consulting and outsourcing. Um, so those are the three companies that I'm focused on. Now, obviously you have more experience than me in starting and running companies, so what made it that? Like creating an AI for a celebrity was the thing that got you passionate about ai.

Keith Smith: So I am, I am, uh, obsessed with ai fascinated by creators and, and the creator economy, generally speaking. And I think that that creators have this really interesting place in the ecosystem. There's been a lot of. Of companies that have been started over the course of the last five to seven years that have really been targeted at the creator economy. And I've been watching that with, with great interest. And, and I think there's, there's kind of three ways to be able to look at creators within the creator economy, in my view. Uh, you can look at them as customers. I. You can look at them as suppliers or you can look at them as distributors. Um, I think the companies that have, have not done very well is those that look at creators as, as a customer base. Uh, and they say, okay, well look, there's 200 million creators worldwide. There's 50 million that are, that are full-time creators. That's a huge customer segment that we can go after. But the reality is, is that most of, of, most creators are, are, are really like a small business in many ways, in the, in the sense that they need to generate additional revenue. They're unlike typical small businesses in the sense that, that there is usually a person who is the personality behind that, that creator. And they are their brand, and their brand is them. And, and being able to, and you just can't separate that person from that small business. And so what that ends up meaning is that, is that that person and the personality there becomes the constraint or the bottleneck to growth in that business. And so it needs a very different type of. Solution in order for it to be able to really grow. Um, and at the same time, you know, they, they are not like a small business, uh, that, that wants to invest a whole bunch of money in the hopes that they're gonna be able to generate revenue over the course of the next two or three years. They, they, they just don't have the liquidity to be able to be a really good long-term customer like that. So our view is, is that if, if, if we can have a solution where we can partner with creators as both suppliers. Supplying their genius, their content, uh, their ingenuity, their personality for the purposes of training and ai. And we all know the value of training sets as it, as it relates to ai, and then also partner with them from a distribution perspective because I. They are the influencers. And, and so the idea is let the influencers influence and influencers influence about 60% of the buying decisions in the world these days. And so they have a massive outsized impact on, on what we decide is cool, what we decide we wanna buy, what we, what we decide that we need next, or how we're gonna spend our time. And so that is, that, that is kind of the, the business model perspective that I took into that, that. Those interactions and those early conversations to say, how can I bring AI to be able to benefit the creator economy, but not from the standpoint of, let's see if we can get creators to pull out their credit card and pay for something. Instead, let's figure out how to be able to give them something valuable so that we can partner with them as both suppliers of their content as well as distributors so that they will distribute. Uh, a product to their user base. And as it turns out, the product that we've developed is an AI clone that allows any celebrity creator to be able to have their own AI clone that can then talk to their fans. And so it allows super fans to be able to scale the kind of engagement that they actually want with their celebrities that they follow, or the creators that they follow. And, and in doing so. We then give a strong financial incentive to our creator partners that we work with because we share 50% of our net revenue back with them. But in this case, now they are actually promoting themselves in the form of ai. So they're not having to go out there and promote our brand or our company. They are really promoting themselves in the form of ai. Has a new modality of engagement between them and their super fans, and that really scratches a really critical itch and it solves a major pain.

Sean Weisbrot: I definitely understand that. I've been thinking about this for a long time. Actually. The first version of my, uh, last startup was supposed to be an end user messaging platform, and the goal was to, uh, create a blockchain wallet. That allowed people to deposit their different cryptocurrencies because one of the things that people had been touting for years was Blockchain is going to enable us to take control of our finances, but nobody was making it possible for people to be able to spend the coins that they had. So I thought if I create my own messaging platform and I have a blockchain wallet that supports all of those coins that people already have, and I make it possible for them to spend those or to earn them inside the platform through various, uh, services and features. One being a paywall for a, a private, uh, live chat. Which Discord didn't do until like a year ago. And I was talking about this in early 2018, and we actually had the blockchain working and we had the wallet testing. Um, so we, we were planning on doing that and it would've been an incredible use case. So I've been thinking about creators for years and now as a creator, it's not my my business, but. I mean, of course if I can monetize the content, I might as well. Right? Obviously having consulting clients is a different story, but being able to monetize the podcast itself is, uh, definitely desirable. But when you don't have millions of followers, it's very difficult to monetize that. Um, so it's. It's definitely something that can be highly profitable. The first instance I heard of someone kind of profiting off of an AI likeness of themselves was this woman, I can't remember her name, but she created a kind of virtual girlfriend experience where there could be, uh, sexualized content with the, with the person, and you would pay per minute to con converse with her. And I think in the first month she did like $5 million. So you definitely have a business model for sure.

Keith Smith: And I think that, and there's a really interesting dynamic that goes on there because, uh, you know, you have this with, with with, with that particular example. And, and there's been a lot of examples that that have come out since, since that one. But that one made a lot of press for sure. Uh, but you have this dynamic where. Yeah, there are girlfriend and boyfriend apps and companion apps that are AI based that, that have been out for a while. Uh, and, and, and those exist. And they exist in the app store and they've been there for a while and they have varying degrees of, of, of, uh, functionality there. I, I've tested a lot of them just as customer explorer, and I can tell you that most of 'em probably are not very good. Um, at least the le that the level that my expectation would be to have a virtual girlfriend. Uh, um, and, um, but at the same time there's this, there's this interesting thing that happens when you're not just. Creating out of whole cloth an AI that is, you know, is, is is something out of kind of a, you know, a weird science, uh, um, book or a, a movie. I don't know if you, if you're, if you're old enough to remember that, that movie from I think the eighties, um, but that was in my, my high school days that that movie came out. And it was about a couple of teenagers who kind of. Developed in their fever dream, this perfect woman. And so that's kind of what a lot of these ai, you know, girlfriend type apps are. There's a big difference between that and just this totally fictional person and somebody who's actually grounded in reality. This is a real person. We see them and we have this kind of parasocial interaction with them. This idea that like they're a celebrity, I feel like I kind of know them because I see their life. I get to see what they post on social media, et cetera, and now all of a sudden I can actually have a conversation. Directly with the AI representative of them and somebody that, that, that speaks like them, talks like them, uh, uh, eventually will sound and look like them. And we'll be able to have a video call just like this today. It is text-based chat, but in the very near future it'll be full video with full audio and synchronous. Um, and that fundamentally changes. The, the, the psychological kind of interaction that starts happening there because, you know, you're interacting with a virtual being, but you know that that virtual being is grounded in reality and in a real person, and that does change the tenor of the relationship. And we already see that in our data.

Sean Weisbrot: So how do you train something like that to make it so that you know when the person's having that conversation that. It feels like them.

Keith Smith: Yeah, that's a great question. And, and so what's interesting is that when we first, when we first started this process, our, our thesis was that. What fans really wanted to do was to get behind the scenes and to see the real person, not the version that is the Instagram version or the YouTube version, but get behind the scenes. What would it be like if I slid into their dms and I could have a DM kind of conversation with them? And, and so that was the original thesis and we started doing a lot of explore around that. We even did an initial, uh, um, uh, just, just kind of proof of concept around that. As an interesting anecdote, uh, in the early startup and the early founding days, it was, it was funny that we wanted to do some training and so we needed, uh, we needed some people who were on Instagram who had a decent amount of, of DM data, and most men that I know don't have a lot of, uh, of dms other than the ones that they're sending out. Uh, and so I needed. Frankly, two women who get a lot of inbound dms in order to be able to train, uh, and, and see if we can build it based on their DM data. And, and since we've moved away from this and we don't train on DM data, and I'll tell you the reason why, but, uh, but, but I started with my, my girlfriend and my daughter and I asked both of them to share all of their Instagram dms, going back to as long as they've been on Instagram. And if you wanna see. The definition of trust. Ask your girlfriend and or your daughter to share all of their Instagram DM data. I'm happy to say they both did. Uh, but, uh, but it was, it was kind of a funny moment and we used that to be able to, to do training. But what we, what we figured out quickly was that the, the relationship that we as fans. Want with a celebrity or with a creator, is actually really with the public persona that we see of this particular creator, creator. And the creators confirmed that for us as well. They said, look, they, they may want to know who I am behind the scenes, but really the person that they've fallen in love with, or the person that they respect, or the person that they idolize is the person that they see publicly. And so, so that led us to training on primarily the public data and the public information and the public content that is already created by that creator. As well as oftentimes some exclusive content. So sometimes there will be like, you know, training, uh, or uh, uh, there will be, uh, courses that are, that are, uh, behind a paywall and they will open up some of that for our training data. So we do get access to some non-public kinds of, of content. I. Or, or stuff that it would otherwise be, be subscription based. And, and so all of that then goes into, into the training set. Uh, and then, and then we also then, uh, spend a lot of time with creator in order to be able to make sure that we are going to deliver an experience that is on brand and is authentic to their brand. And that was the first big question we had to answer for ourselves was, can we actually deliver an AI clone that is true to the brand of this creator, that the creator will have a conversation with themselves. And that's very interesting and very weird the first time you do that and you have a conversation with yourself, but very fun, uh, will they actually look at this and say, yes, this is a good representation of my brand. Uh, and so I'm happy to report that we've been able to, to pass that test. Uh, the six creators that we launched with, uh, and in our beta, uh, a little over a month ago, they all resoundingly gave us the two thumbs up. That, that they were amazed at how. While we were able to train the AI to be able to respond like them, give answers like they would, uh, in, in much the same way that they would if they were responding themselves.

Sean Weisbrot: Hey, just gimme 10 seconds of your time. I really appreciate you listening to the episode so far, and I hope you're loving it. And if you are. I would love to ask you to subscribe to the channel because what we do is a lot of work and every week we bring you a new guest and a new story. And what we do requires so much love so that we can bring you something amazing. And every week we're trying really hard to get better guests that have better stories and improve our ability to tell their stories. So your subscription lets the algorithm know that what we're doing is fantastic. And no commitment. It's free to do. And if you don't like what we're doing later on, you can always unsubscribe. And either way, we would love a, like if you don't feel like subscribing at this time. Thank you very much and we'll take you back to the show now. Well, that sounds really interesting and I, I like how it, it's recognized that there's a public facing persona and the real person because like I know doing the podcast that. There's things that I say and do, and I try to be as, as authentic on the, my public and private self as possible. Um, so that you don't really see much of a difference. But if you knew me well off air, you might think that I skew negative in the way I think and talk. Where when I am doing the podcast, you may feel that I skew positive in the way I think I and I talk. So,

Keith Smith: so this is the fluffy version. We're getting a view then, huh?

Sean Weisbrot: Because people like to listen to people that are positive and they like to be around people that are positive. Yeah. And so I, I don't hide the reality of my experience from the audience. But I definitely filter myself be because my daily existence may not be as great as my public existence because when I'm talking to you, I'm thinking about what you're talking about. I'm asking you questions based on yourself. And so even though I may share something personal. It's not the, the, I, I, I'm, I'm not, I don't know how to say this. It's difficult to explain. Basically, I try to be as real as I can, but when you see someone for 40 minutes a week versus being with them for hours a day, it's a different, you, you see more of them, right? So it's like when you're, you know, let's say I was married before, before COVID, we would only spend. An hour or two in the morning and an hour or two in the evening together. The rest of the time she was at work and I was at home working during COVID. There were periods of time where her gym was shut down, where she was working, and we might spend the entire day together. Well, when you spend the entire day with someone, you see all of that person. Yep. And what a lot of people saw during COVID was they didn't really like the entire thing that they saw that they thought before was not a big deal because they weren't with them all the time. So. That's, that's my point. You see a part of me and I try to give you everyone, everything about me that I can, but you don't see all of me, and therefore, who I am outside may be slightly different or, or there may be like you're, maybe you're not seeing the edges. I. Of myself.

Keith Smith: Yeah, that's, and that's part of the fantasy and part of the seduction. And I don't mean that in a, in a sexual way at all. I mean it as a, that's part of what, what we look at from these public figures, we go, wow, it's really aspirational. It's really nice, but we don't have to see them, you know, brushing their teeth or, you know, and they just wake up in the morning or any of those kinds of things. Uh, we get to see the, you know, the kind of the, the, the cleansed version and, and, and what we've created then. By being able to take the personality of a creator and put them into an ai, into a chat bot, is to be able to kind of take the best versions, uh, that, that otherwise pinhole view that we get into somebody's life and try to expand that into an entire character. So you are still getting in some regards a character 'cause you're still getting the best version, uh, of this person through the form of ai. The difference is, is that the AI doesn't get tired. The AI doesn't get bored. It doesn't get annoyed when you ask it the same question over and over again. Uh, you know, it doesn't get annoyed when I leave my underwear on the floor. And so it allows, you know, you to be able to have still a bit of that, that that fantasy kind of relationship, but a much deeper version of it and over a much more extended period of time. And the kind of thing that frankly a lot of, of, of, of super fans want and a lot of, a lot of people even that wouldn't necessarily consider themselves to be super fans. It's the kind of engagement. That is just it. It, it creates an entirely new modality of engagement that we just never really considered was possible before. Because, you know, as soon as you get to a certain size and scale for a celebrity, the kind of of engagement that these super fans want from that celebrity just doesn't scale. And so there's no way for the celebrity to be able to give their super fans all of the things that they actually want and the engagement that they want, the, the personalization, the access, the relationship, all of those things. Well, now for the first time in human history, because of ai, we can defy the laws of time and space and allow, I. A celebrity to be able to be in more than one place at a time, and to be able to have these authentic one-on-one interactions and conversations and relationships with their fans and scale those at a level that just has never been feasible before with any sort of technology platform.

Sean Weisbrot: So Chat two PT or Sam Altman came out a few days ago and said, we're finally updating the 4.0 data so that it'll be able to talk to you about anything up to April, 2023. Bard and, uh, grok. This, uh, Elon Musk's new AI are meant to train on live data from the internet. Obviously, your celebrities are not static. They're moving around, they're giving new speeches, they're creating new shorts. They're constantly creating. How do you. Make it so that their AI is able to keep up with their live person so that when someone wants to talk with the ai. The AI is capable of knowing what they're talking about because it has access to that information or it's been updated, like how do you manage that?

Keith Smith: Yeah, it's a great question and there, and there's two really critical things that need to happen for any sort of companionship or relationship to be built. And, and the first is memory. Uh, if, if you and I talk every day, but I, I subsequently forget everything we talked about before we talk next time, we're just not gonna have a relationship. Uh, that is, it is impossible to build a relationship on top of that. And, and that is a little bit like what it is to chat, talk with chat GPT at least today. Uh, and a lot of the other ais, they just don't have good long-term memory built in. And so, so the first thing we did was build in, was build in, uh, uh, long-term memory. But the second thing is that you have to have up to date. Information because if I'm going to be having a conversation with a celebrity and they just posted something that, uh, you know, that was very relevant to their life and they just posted that this morning, and, and, and the training data for this AI does not go up through this morning, then all of a sudden I'm gonna be having conversations with an irrelevant, you know, and an outdated version of this person. And that does not feel very authentic. And so, uh, so that is something that. We are still building a lot of the pipes for that, but we have, that was one of the first early features that we built was this ability to be able to not only pull in that data for a, for the first time training, but then to be able to pull that in on an ongoing basis and have that flow. So there is a constant update and one of the things that we're trying to figure out then on a creator by creator basis, it depends a little bit. In some cases, you need that information to be updated by the hour. In some cases you may only need it to be updated by the week. Uh, but it depends a little bit on, on the creator and, uh, and so those are things that, that we are, that we are fine tuning at this point in time. But, uh, but yeah, absolutely having that relevant up-to-date information is absolutely critical in order for there to feel, have any sense of real authenticity that's going on there. And then one of the other important things with that, as a, as a side note is that, is that recency matters in, in our conversations as humans. Recency matters. That's how we can use things like pronouns and that sort of thing. And, and you, you know who I'm referring to because I've just said that person's name and then I use a pronoun to refer to that particular person. And it's like, okay, we all understand that, but if I use a pronoun to refer to somebody that I was talking about yesterday, that's gonna be very, very confusing. And so the same kind of concept applies here, is that, is that if, if there's something that's recent that has happened that needs to have a, a higher priority in the conversation that we are having and, and intuitively humans do that. Ais don't intuitively do that. And so making sure that, that as we continue to train with new, updated information on a daily basis, uh, or an hourly basis, even depending on the creator, that that, that information, this most recent information is the stuff that takes priority. And if so, if I ask an ai, uh, that is, that is on me for you. What is, you know, what is your view on X? And they had a view on X. That was one thing a year ago and another thing yesterday. I'm going to expect that it ne needs to have that updated information and have that updated view on X when I ask it.

Sean Weisbrot: So recency is important obviously, but being able to have a two-way conversation with an AI is something that I think is severely lacking across the board. If you look at chat d, pt, you ask it a question and answers. If you look at Bard, you ask it a question and it answers. You look at Rock, it does the same thing. Yeah. There's only one AI that I've found that will ask, actually actively ask you questions and maintain the memory of the conversations. Let me guess. Can I guess Go for it.

Keith Smith: It, it's, it's pie, absolutely.

Sean Weisbrot: To date. Yeah, they're good. It is my favorite AI because it wants to keep asking you questions and when you respond it wants to ask you follow ups and it actually listens and it processes and it sounds like a human friend. Do you support that or is that something you're working on? Because you know, let's say for example, your system knows that it's my birthday today, it's not. But example November 7th, it's my birthday. If I have established a relationship with Neil deGrasse Tyson through your platform. Is Neil gonna wish me a happy birthday and ask me what I'm gonna do? Is he going to, you know, share something with me that he thinks would be fascinating for me? Right. Do these AI copies reach out and communicate like a friend? Or is it just one sided me to you?

Keith Smith: Absolutely, and, and it's, and it's such a critical point. By the way, Neil deGrasse Tyson is a perfect example. And so Neil, if you're watching call me, uh, um, we want you on the platform really badly. Uh, um, but, but yes, absolutely. And, and it's interesting in the very first iteration that we launched. We found that the, the AI was trying to be very, very helpful. So you would, you know, you would ask it a question or you would say something, it would give you an answer and it would almost sound like it was ending the conversation. It would be like, thanks, I hope that's super helpful. Have a great day, and you're like, and it's being polite. But that's not how we talk as humans. Can you imagine having a conversation with a human that way? And every single time they were like giving you a, you know, a final, you know, goodbye. And, hey, thanks for coming anyway. It was great to see you. And you'd be like, wow, I feel like I'm being rushed out the door. And so we found that very quickly. So we, we, we made significant changes to, to how it interacts and in order to be able to try to mimic human interactions and, uh, and so it's trying to mimic. The turns of phrases and, and the, the type of speech that that particular creator is using. But, but what it doesn't necessarily have a lot of, uh, access to a lot of, in, in most cases with training data is it doesn't have access to a lot of conversational data. Um, that is a little bit different with interviewers that, that have a lot of like podcasts and YouTube. In that case, there's a lot of training data where there's, there's interaction and it's a, it's, it's a, it's a tennis game going back and forth, and both players are hitting the ball back and forth. Uh, but without that. By default, the AI is kind of this terrible tennis player where you lo the ball over and it, you know, it pockets the ball and then you're like, okay, great. If I'm gonna start again, I have to, I have to hit another ball over. And so now what it does, uh, and, and this took a lot of coding and a lot of work in order to get it to this point, is much more conversational. Continues to ask questions, talks, and it remembers things and it classifies things as I tell it. And so, so you're not just filling out this form where I'm saying, Hey, you know, I'm a 52-year-old. I live in New York City, I have two kids. I like to ride a one wheel. You know, these are my hobbies. This is what I like to do on the weekend. Like, I don't have to fill out a form that, that says that those things just kind of happen naturally in conversation as I'm con, you know, having conversation with the AI and the AI remembers all of that stuff. And then it brings it back up and it talks about it and it asks me questions about it and, Hey, how was, how was one wheel riding? I know you were really excited about going one wheel riding this weekend, Keith, you know, how was it? Was it as good as you thought? Uh, you know, and so those are the kinds of interactions that we end up having, and they feel much more like authentic. Real human conversations and we continue to make improvements there and we'll continue to make improvements. But yeah, you're hitting on some key points there that, that chat, GBT is very much like, uh, like, uh, you know, talking to a, a socially unaware first year analyst who just kind of answers questions and is done and it's not a very entertaining and interesting dialogue. Uh, and you'll find a very different experience when you talk to, you know, ais like PI or any of our meet for use on,

Sean Weisbrot: so two questions just came to mind. The first question is if, let's say I'm, we'll continue with Neil deGrasse Tyson. Let's say I have a conversation with Neil, and Neil learns some personal information about me. Does your system. Store that information so that if I start having a conversation with another ai, that AI knows that information about me and can automatically do that, or do I have to rebuild that knowledge base with that new ai?

Keith Smith: Great question, and we, we've debated about this a lot internally as it stands now. That is a siloed, uh, uh, walled garden within that particular conversation that you've had with that particular ai. So what you tell them is what is, is what they know about you. And what it does create though, is it creates the, the ability for you to have a little bit of a different interaction and a different kind of personality. With different ais and, and so, and you get to know one in one context, you get to know another one in another one, another context. And what's interesting about that too is that 'cause the shape of these conversations change over time, you may be having a conversation with the exact same me for you that I am. Our conversations will go in drastically different, different directions. 'cause I'm telling and talking about a particular thing and we're, and we start forming a relationship and a connection around the things that I'm talking about. Whereas you're gonna talk about the things that are important to you and you're gonna form connections about that. And, and, and it is going to remember those things about you. And this is gonna steer the conversation that way. And you're gonna have. What seemed like very, very different types of conversations and relationships that I would have with the exact same me for you that you would be having.

Sean Weisbrot: So that next question I had was, is it possible to do a kind of top down push where let's say Neil has a new book coming out, or he has an idea for a book and he comes to you and he says, Hey, I've got a hundred thousand people on your platform that are actively engaging with my a, with my ai. Can you run a survey for me and tell me what percentage of them say they would be willing to buy this book before I even write it?

Keith Smith: Yes. And so you can think about as, as a creator, you can think about every single conversation that happens with, with your, your fans and your me. For you AI as a customer explorer, uh, it is the ability to be able to understand on an anonymized basis. What your fans care about. And so one of the things that we do is, as is we, we, we analyze those conversations. One from, uh, and first and foremost from a compliance standpoint. And we actually do that real time. And so, so one of the things that's interesting about AI is that it is, it is better as a tattletale than it is at oba. And so you can say, here's the rules. Only do this. And then it spits back something that doesn't obey the rules. And you say, did that obey the rules? And it goes, oh no, it actually didn't. Sorry. And, and so it, it can know, it can identify and it can look at that conversation and know that it didn't obey the rules, but it doesn't necessarily obey in the, in the middle of it. And so, just like how my brain works, I probably spend 20%. Of my brain power thinking about should I actually say that thing that just popped into my mind or should I not say that thing that just popped into my mind. And so we have a little bit of that same kinda logic. There's a processes that is babysitting and going, no, that does not comply with the, with the parameters that we've set in order to be able to have an authentic conversation per what this creator wants in terms of their brand identity. And so, so there's that real time monitoring going on. And then we actually do that as a follow up as well on an anonymized basis for these conversations. And we're also trying to make sure that they. They are that these conversations are meeting the requirements and the goals that are set by, by the creator, uh, in order to be able to make sure that, that, uh, that, that there are, uh, that there is kind of under overall kind of compliance. But then with that we then pull out insights. What are the things that they are telling that their fans are saying to the creator that are gonna be interesting and, and, and things that are gonna need to be known. We also do this, by the way, for our own product. So we have developed. What we call Keith Bot, uh, which is based on me, uh, um, and Keith Bot happens to have a very one track mind. He's one of the me for use that's on our site. Uh, but he just wants to talk about the product. He just wants to ask you, you know, what was your experience with the product? What did you like, what did you not like? What worked, what didn't work? What should we fix? You know, are there new creators that we should add? And it's all about, it's a wide ranging conversation. With my personality, my, my training data, but with a very one track mind about, let's not talk about anything other than the Me for You project, which is probably not too unlike me in real life actually. In, in, in reality, when I stop and think about it, my, my friends and family would tell me that anyway. Uh, and so what happens is out of that then we get. This incredible, every single week actionable insights that come from those conversations that feeds right into our product lifecycle and into our product, uh, into our product roadmap so that we can make real time improvements. And so we are constantly making improvements in order to be able to make sure that we can meet our, what our customers want, uh, as well as meet the goals that are set by the creators in terms of, of what they want these conversations to be doing. And one of the things that, that just shocked us, uh, like crazy was. The, the way that we were able to in the first four weeks from launch, just by having that feedback loop, that rapid feedback loop, we were able to make improvements and, and see engagement increase massively just over the first four weeks. And so when, when we think about what our company looks like 52 weeks from now, I. It, it tells a very, very optimistic story that, that that constant iteration, that constant feedback loop will continue to make us better than yesterday. And that is, you know, that's, that's the whole entrepreneurial journey right there, is figuring out how can you get a really tight customer feedback loop, make that actionable, feed that back into your product and make constant improvements. So I've taken the calling this entrepreneurial crack because I am truly addicted to it. It is a feedback loop that I have been looking for for the 30 years that I've been starting companies. Um, and AI enables all of it, and it is absolutely magical.

Sean Weisbrot: It's very interesting. I think a lot of companies would benefit from having this, 'cause right now a lot of companies just have like a chat bot and, or like they've got some text-based FAQ and I've always known that being the, like the person that goes out there as the founder. That you go out there and you tell people about your product or your service, you're gonna be able to answer those questions way better than any sales person or any text-based thing that you've written down, which is probably boring and corporate. So by having a bot with your own personality that knows everything about the product and basically acts as a a product manager, um. Is pretty cool. I like that a lot.

Keith Smith: Yeah. Yeah. And, and I think, you know, as, as businesses, we've all thought about this from the, from the reverse perspective, which is how can I, how can I figure out how to be able to not pay as many customer support people? Or, or maybe it's even more altruistic than that and it's, how can I be more responsive my customers, by being able to give them the answers immediately by having a chat bot that, that, that, that fills the void of that customer support person. And so we think about it in the context of. Let me just answer questions and give information to my customers. But it is incredibly, and it is very good at that, but it is also incredibly effective at then being able to pull that information out. Well, what's, what is the problem? What are the issues? Being able to categorize that and be able to say, okay, this is, these are all the various things that folks are asking for. This is these, you know, we can, we can start to now stack rank these because we can start to see some similarities between these various conversations. Feedback that back into your, into your software, uh, development lifecycle. And all of a sudden you have this constant improvement chain that is really, really,

Sean Weisbrot: I want to talk to you more about this off air, actually. I just had an idea that I don't wanna mention on air. Um,

Keith Smith: oh, this sounds useful. Hear all about it. Good.

Sean Weisbrot: So what's the most important thing you've learned in life so far?

Keith Smith: So, when I look back, um, I've, I've, I've made a lot of mistakes, so I've been able to learn a lot of things, uh, which is the best way, I guess, to, to learn things, at least for me. Um. The thing that, that if I look back and I have regrets, which is an indication that I now, now learned something, uh, I, I can't say I have a lot of regrets, but the regrets I do have are that at various key points, uh, in, in my children's upbringing, I prioritized work over them. And I was a single dad, uh, and I, you know, had started five different companies prior to this. So it was always me. And my two kids and a startup. And so it was always this very, uh, a bit hectic. I, I, uh, my kids and I now, now that they're adults, we joke that, that they were raised by Wolf, uh, singular and, uh, and, and so it was, it was a bit of an odd upbringing to be, you know, to be, to be raised in a house with a. Uh, with a, with a, you know, founder or startup founder as, as a father who was constantly running around and starting new companies. Um, but, but the regret that I have, and the lesson that I learned more than anything else is that, is that at those key moments when you need to show up for your kids or for your family, or for your network that you care about, that is, that is really the thing that is, that that matters frankly, even more than the stuff that we're building, uh, to prioritize them. Over your company and, and I, I, I've, uh, once I turned about 40, 45, I started to get good at that. I was really bad at that early on, and, and that's a regret that I have, and that's a big lesson.