Influencers Clone Themselves: Connect with EVERY one of Their Fans PERSONALLY

Raymond Dong joins "Behind the Bots" to discuss MyCompanions.ai, an Artificial Intelligence platform that helps influencers better engage with fans. MyCompanions.ai builds AI companions that can have personalized conversations on the influencer's behalf, freeing them up to create more content. The AI is trained on the influencer's voice, tone, interests and more to mimic their style. Fans enjoy feeling like they're getting exclusive behind-the-scenes access. The AI helps identify super fans and maximize revenue potential. 

Raymond explains how MyCompanions.ai leverages ChatGPT and other language models as a "middleware layer," optimizing conversations. He sees great potential in using AI to support creators against the rise of fake AI influencers. MyCompanions.ai is currently Telegram-based but looking to integrate with platforms like Instagram. 


MYCOMPANIONS.AI

https://mycompanions.ai/
https://twitter.com/MyCompanionsAI
https://www.instagram.com/mycompanions.ai/?hl=en
https://www.linkedin.com/company/mycompanions-ai


FRY-AI.COM

https://www.fry-ai.com/subscribe
https://www.youtube.com/@TheFryAI
https://twitter.com/lazukars
https://twitter.com/thefryai

Creators & Guests

Host

Ryan Lazuka

The lighthearted Artificial intelligence Journalist. Building the easiest to read AI Email Newsletter Daily Twitter Threads about AI

What is Behind The Bots?

Join us as we delve into the fascinating world of Artificial Intelligence (AI) by interviewing the brightest minds and exploring cutting-edge projects. From innovative ideas to groundbreaking individuals, we're here to uncover the latest developments and thought-provoking discussions in the AI space.

Raymond Dong: For myself, I grew up in the Midwest post school. I was a consultant at McKinsey and Company, then had a career in the hedge fund world trading public equities, making long-term bets on all these public companies in software, technology, healthcare. I started to dig in on how all these companies become so great in technology, but then also looking earlier at private companies to try to spot the next big innovations. Over the last two and a half years, I've been an investor here at a fund called Boat Capital, our multi-stage investment firm, invest across all different areas in terms of technology-enabled sectors. We have a separate piece where we incubate some businesses as well. Through investing technology, I got really deep into applied AI several years ago, looking at how it applies to different sectors ranging from healthcare to consumer to logistics.

As a user myself, as soon as ChatGPP came out and it was widely available, I've been using it for trying to integrate it with all of my workflows at work and have found a bunch of really interesting use cases. And then now, over the last couple of months, the emergence of my companions basically was able to work with a couple of folks, a CTO and then my other co-founder, the three of us, saw a big hole in terms of a gap in the market where the creator economy has just been taking off now for a number of years and there's no signs of slowing growth. But when it comes to AI-enabled tools to help creators, there weren't many out there. So we strive to be the preeminent AI-enabled creator company, drawing upon a lot of our past experiences working in technology, but also in the creator space.

Hunter Kallay: And then you do have a project out called MyCompanions.ai. Can you just tell me a little bit about that? What's your role and then what's for the average person? What's the project look like?

Raymond Dong: Yeah, absolutely. So I helped launch this company with two other friends of mine and we're about two, three months in now. So it's very, we haven't been working on it a long time, but we've seen immense traction already, which has been super exciting for us. And basically what it is is it's AI-enabled creator co-pilot. Our initial product is a AI-enabled messenger. So we're tackling this issue where influencers, especially larger ones, have hundreds of thousands, millions of followers.

It's very hard to have one-to-one conversations. And newer platforms like Patreon have helped try to address that by putting paywalls, but also it's still very difficult. So what we have created is we use AI to amplify these conversations to where through our platform, which is currently telegram-based, creators can go on and basically message their fans one-to-one, but then the AI can come in and out seamlessly to amplify the reach of their conversations. To where at any given point, a creator can be messaging thousands of their fans in a very personalized way. And that's based on how we train the models to basically mimic their likeness, their voice, their interests, and we're constantly improving the model based on the topics that matter to the creator and what they want to share with their fan.

Hunter Kallay: Interesting. So what does the conversation look like with one of these AI influencers? What kind of stuff can we talk about? Can you just walk me through? What does the conversation look like? Is it like I'm talking to kind of like a chatbot?

Raymond Dong: Yeah, for sure. So these are all, I would say, real influencers. So we basically don't have any full AI influencers. Most of our influencers are fully real people that want to just talk to more of their fans. I'd say when it comes to conversation, we are very, very focused on the creator influencer in terms of topics that they want to talk about, topics that are totally off limits.

As part of the onboarding process, we train the model to basically censor anything that want to censor. So the typical conversation is kind of what you would expect. So it's fans talking about a creator's life, their work, photo shoots, travel, food. Like these are very popular topics that users like to ask creators about to get a snapshot into their life. Creators able to send photos from like a photo shoot, for example, behind the scenes stuff to their fans as well. We have a tipping feature, it's all automated using AI. And yeah, fans have a lot of fun with it. And they're constantly getting kind of new information that isn't anywhere else.

Ryan Lazuka: Awesome. So say if Taylor Swift want to go on here and talk to have an AI Taylor Swift, is that sort of how it is? It's an alternate AI personality based on their background and experience and anything that they've done in the past. And AI learns about them and then talks like they're the celebrity. Is that how it works?

Raymond Dong: Yeah, the goal is to be as close to a mimic of in your case Taylor Swift as possible, whether it's a voice, how they respond to messages, even their tones and intonations and how they like to type in a chat, all of the above, right? And it's constantly getting better. We're constantly feeding in new experiences from the influencer. But like, you know, a decent percentage of the time is the actual influencer going into messaging with people.

It just, and that adds to the context, right? But it just so happens that they can turn AI on and off. And then when they turn AI on, AI can just carry out the rest of the conversation. So that way the influencer gets a lot more leverage in terms of how much they can amplify their conversations with the time that's very valuable for them.

Ryan Lazuka: Would the end user know when that's turned on and off? No, it's very seamless.

Raymond Dong: And that's a proprietary to our company. We've developed our AI models such that it's a seamless transition between the human and the AI. And we fight that, right? We say it's an enabled companion. So you're talking to the person or to the person's AI that is in exact replica of the person.

Ryan Lazuka: Okay, cool. So if you're talking to Taylor Swift, the real Taylor Swift, will you know that you're talking to the real one or the AI one? It sounds like you can flip a switch if you're the influencer, but does the end user know if they're talking to the real person or the AI person? Sometimes.

Raymond Dong: It's a little complicated, but there are ways to tell. But I think to the traditional user, they care less about whether it's AI or person as long as they know that they are learning about their favorite influencer or creator.

Ryan Lazuka: Yeah, I don't even think it really matters. It's just interesting to see. The same topic keeps coming up and up every time we talk to someone new is, people are really going to care if they're talking to the real person or the AI person in terms of anything. Influencers, customer service, as long as the quality of the content is good, that's really what people care most about.

So it doesn't even matter if they're talking to the real Taylor Swift in this example or the AI version, as long as the content is engaging. That's all that really matters in the long run.

Raymond Dong: Yeah, we agree with that. And I think it's, we've done a lot of work to make the kind of how we push notifications and messages and photos to make it as engaging as possible. Because what we've actually found amongst our users in this market is there's a lot of passive folks that are kind of more introverted than extroverted. So as a result, in our products, we do a lot of work to try to proactively push conversations.

Hunter Kallay: Very cool. I was wondering, just from like an ethical standpoint, if I was like an influencer, and I'm trying to think like, I'm burning myself in their shoes using this platform, I'm thinking, do you have any people who are a little concerned because maybe their AI companion might say something that they wouldn't want them to say, you know, some like ethical concerns that maybe somebody screenshots, oh, look what this celebrity said, and they're like, well, actually, I didn't say that sort of thing. Is there those sort of concerns? And how do you deal with that?

Raymond Dong: Yeah, so that's a concern to think very seriously. And it is very much at the forefront of our business. Because at the end of the day, we're an influencer led business.

So if influencers are happy, then we don't have a business. So the way we address that is we are very sensitive during the onboarding process to understand kind of topics that are like topics that are off topic. And every influencer gives us a list, we have a suggested list of a bunch of things, and they kind of check through what is kind of okay versus not okay. But to just give you a sense, like several of our influencers don't want their bot to talk about politics, religion, like sensitive topics like that. And we train the model such that if anyone asks a question, like, oh, what do you think about religion, the bot automatically just says, sorry, you know, I can't talk about that kind of thing.

Instead, let's talk about how your day is going. Like it fits very seamlessly. And, you know, we're constantly monitoring the AI to ensure that it doesn't ever go out of bounds.

Ryan Lazuka: It feels like it's going to be that that's going to be the new thing is like if you posted something on Twitter, you didn't like before x.com now, you could just say, oh, my account was hacked. Well, maybe the new things just be like, well, my AI bot set it for my X.

Raymond Dong: Yeah, for sure. I mean, that's why AI security is I think going to be a huge deal. And I think a lot of these cybersecurity companies have a huge tail end. It like, like, for example, even we haven't gone to photos yet, but like AI generated photos, which we, you know, we have on the roadmap for a lot of our real influencers, but we haven't integrated yet. But like, it's, it's, it's staying very hard to tell something's real or fake, right? So you have to monitor that.

Hunter Kallay: For sure. How are you looking to integrate the photos? Like, are they going to be posting like, like, let's say like Snapchat, like you're, you're sending them snapshots of like what they're doing and stuff like that. Is that what you're envisioning? Yeah.

Raymond Dong: So we already do that. You know, as part of the proactive work we do every couple of days, we gather from the creator a sense of what their life is, what they're doing. And then we will proactively send a photo through the interface through telegram to their fans. And it's often times a free photo.

Sometimes it'll be like, okay, tip, and then you can see the photo or you buy a photo. But we integrate that pretty seamlessly. And typically we paint a narrative around each photo. And that's like the power of AI, right? Because if you just have like a paywall platform that doesn't have AI, you can send a photo, people can buy the photo. But what we're able to do is paint a picture, a narrative around each photo. So if you have, for example, one of our influencers, you know, has a runway photo shoot, she's a model, and we can send a back behind the scenes photo of that photo shoot. And then we can paint a narrative around how that photo shoot went, what was it, what were the challenges. And then it's like a very engaging conversation led by that photo. And that's where the power of AI comes in.

Hunter Kallay: And how does the conversation end? Do they just continually talk to the influencer or is there an ending point to it?

Raymond Dong: Yeah, so we have users that joined the first or second week that are still going strong 10, 12 weeks later as a subscriber. And, you know, how we have it set up is there are many users we see that are talking on our platform, our plus a day.

And they're always coming back trying to find new topics to talk about helping kind of discover new areas, frankly, that that within our model. So I would say there's a strong cohort of people that are continuously coming back and interacting with the influencer. Influencers love that, because, you know, that's their most loyal fans, they want to understand what they care about. So they can post more of that content. Oftentimes, those are also the high spending people.

Ryan Lazuka: And then how does it look like, say if you're an influencer, you go on my companions and you, you have people talk to your AI personality. Like, obviously, the influencer is not doing that, they're not talking to them. But they can, like you said, but is there a time at the end of the day if the influencer wants to come back and see all those conversations? Is that something that's private for the end user or can, you know, the influencer see all those conversations throughout the day of all their fans?

Raymond Dong: Our mission from influencer co pilot is that like, it's the influencers tool, we're just creating the software for them, right? We use RAI health monitor privacy and all of that. But it's them talking to their fans, whether they are themselves, so they'll be able to access their own dashboards and their conversations. And they can come in at message or they can leave it's AI and kind of see how that's going.

Ryan Lazuka: So they could like, if they find a really cool conversation, you know, in the midst of all the all the conversations throughout the day, can they go on and message that person directly and say, Hey, I saw my AI bot was talking to you, this is really cool conversation. I'd like to take it a little farther.

Raymond Dong: Yeah, absolutely. And the AI, like I mentioned, there's a very seamless transition back and forth. And oftentimes, it's like a really pleasant surprise for a fan if they can get a sense, like, like get a sense that, Oh, wow, this is like, you know, feels like the person's really pushing it forward. Like the user loves that too. So it's really a win-win.

Hunter Kallay: Yeah, you talked about pushing it forward. What are some ways that they do that? I was just like asking the user about their experiences and their day or what their opinions are on different things.

Raymond Dong: Yeah. So oftentimes, it'll be like adding color to a conversation that they I didn't know before. So the example of someone influencer doing photo shoot, you know, we would try our best to amplify that conversation that they would have with the fan about how the photo shoot went, what were some challenges were they doing after, but then the influencer can pop in and just be like, Hey, you know, this is what happened in the photo shoot. This is how I felt. And I'd love to hear what you think about this given, you know, XYZ that I know about you based on my conversation previously.

Hunter Kallay: So they kind of learn about the fans too in some way. Correct.

Raymond Dong: And I think that's a big challenge today without like before AI is that like, whether it's Instagram or TikTok, like people get very little data from their fans. That's what their fans want to see. It's very much like a retrospective that they might do.

But what we are able to do is flip it on the head and make it a full proactive experience and say, Hey, these are live conversations. This is what people want. You know, this is how you should think about what is what depends on your objective. You just want greater engagement. These are topics that people care about. If you want monetization, this is what people are willing to pay for.

Hunter Kallay: So what kind of integrations do you have right now? What kind of platforms can they use this on?

Raymond Dong: Yeah, sure. So, you know, we're only a couple months old. So our main platform now is Telegram. And it's a Telegram based spot. And we have tools that we've built out proprietary on the back end to allow the co-pilot features for the influencer to come in. So we have like dashboards, we have mess like our internal out messaging apps where people can come in and turn AI on and off.

But for the user right now, it all shows through Telegram. Actively in the roadmap, we're adding kind of additional integrations. So all the big platforms like Patreon, we're working on. And then we're trying to navigate through like Instagram, etc.

Even SMS texting. But you know, Telegram we started with because it's a massive community. And frankly, it's a relevant community for what we're doing. And also just technology is great in terms of building our application on top of them. Awesome.

Ryan Lazuka: And right now there is a there's a demo on your website that people can go and check out that we'll take you to Telegram. And you can check out this tool right now.

Raymond Dong: Yeah, absolutely. I mean, we've been live for two months. We're like not just a demo. Like we have, you know, 15, 20 influencers live that many thousands of users already. It's, you know, very much live and well.

Ryan Lazuka: How did you get these influencers? I'm just curious. Like, did you market or do you guys know people through the finance side of things?

Raymond Dong: Most of it's our personal network. So my co-founder, Maggie, she's kind of an influencer herself. And has been for many years. So it just makes it a lot easier for her to initially onboard her friend. And then from there, it's been spreading very fast word of mouth. So an influencer comes on has been experiencing stable recurring revenue, and then they'll typically refer their friends. And we launched this referral program. So basically any influencer or anyone, if they refer someone that joins our platform gets 5% of their earnings into perpetuity. So it's a great deal that we're offering to anyone to try to capture this network. Awesome.

Ryan Lazuka: And I think it like people just think of influencers, but you can also use this and correct me if I'm wrong, but as a tool, like as a mentorship, right?

Like if you're an experienced carpenter or like an experienced computer programmer and you want to consult people, well, you could sort of use this as a consulting device where you don't have to teach people, but your AI bot could be there for you teaching for you. Absolutely.

Raymond Dong: So you're absolutely right. There's so many different use cases and we've been thinking about this a lot, right? So like right now, a lot of it just conversation based, and we have kind of this niche of Asian female focused influencers, we're actively onboarding other folks like food influencers, talked about different food and we actually trained the model to have personalized recommendations by city and all these different things. Wine influencers, adding in, you know, like self love type influencers. So not to the extent like full-on therapy, but just like someone that you can talk to in the morning that's like pumps you up and is like, like Hunter Ryan, you guys are amazing. Like let's have a killer day, like go crush

Ryan Lazuka: it, think everybody needs that, especially in the morning.

Hunter Kallay: I was going to ask you a question to take things a little bit behind the scenes. Obviously, it seems like there's a lot of, I mean, not maybe not in duration, but there's a lot of AI technology behind the training system. It seems like that's very deliberate. Can you talk to us a little bit about what that training looks like for the influencer when they're getting onboarded?

Raymond Dong: Yeah, sure. So basically how we do it is we break it into voice and kind of textual context. When it comes to the voice that we get as much audio of the influencers as possible. And oftentimes, we have our own kind of proprietary scripts that we ask influencers to say, that capture all the intonations and tones and accents and things like that. And then when it comes to the contextual stuff, we have, again, a script of hundreds of questions that we have influencers fill out that we create a narrative around to try to make it as close to the personality as possible. And then the process of actually training the model takes a long time for each influencer. It's a week, two weeks of effort on our engineering team to go train the model, to sound like the person, to feel like the person, to respond how those like the person would, and have the contextual details that the influencer wants, but then the ones that they don't want.

So that's a little bit behind the training process. Now, I think what's really interesting is that all these AI models have come out like OpenAI, Mistral, like there's so many of these now where we'd like to position ourselves as kind of be like a middleware layer, where we can work across all the different existing foundational models, depending on the topic and the use case and kind of the situation. Because it's a balance of like, which model is a better context and can kind of deliver better results. But also there's actually big price differences between the models. So we try to optimize for that.

Ryan Lazuka: So do you sort of pick which model you're going to use during the conversations, or is it depending on the influencer that, you know, the topic they're bringing up?

Raymond Dong: Yeah, so it's a little bit of both. Like we typically have like a base model that's like a, like the default for each and both that we put out, based on kind of situation and conversations. But then we've been adding in this feature where you can flex between models based on cost and quality. Cool.

Ryan Lazuka: And right now, is it like an influencer wants to, you know, sign up, give you guys a try? What's the process like? Is it something, it sounds like it's more of a consultation right now between a live person, or is it something they can go on the website and do that right now and fill out all these questions and answers online?

Raymond Dong: Yeah, so sign up is actually very seamless and we're making increasing more seamless for an influencer, you know, they can do their end of it in as little as like 20, 30 minutes.

Okay. Basically, go on the website, submit a request, we'll send you kind of a onboarding pack. And then all you need to give us is one, a clip of your voice saying the script that we'll send you, and then two, to fill out the questionnaire, right? And then of course give us approval to basically use your digital likeness as well as photos and whatnot.

And then that's it, right? And then we take a bit of our time for our engineering team to fully vet and create the bot, we send it back to the influencer for testing. Usually they'll test it for a couple hours or a day, give us their approval, and then we do a joint launch. And then once it's launched, the influencers will control over everything because it's a co-pilot.

Ryan Lazuka: And it's, there is a price for this, I'm sure, right? How does that pricing work? Is it based on like a monthly fee for the end user to talk to these influencers and can the influencers set the price as well? How does pricing work? For sure.

Raymond Dong: So for the user, there's two monetization mechanisms. One is a subscription. So right now it's about $250 a week. So what is that $10 a month to talk to the influencer basically, you know, post unlimited messages per week, avoid tax, they'll receive some free photos. So that's one. Two is photos slash tipping. You know, we see a lot of revenue from users asking for basically pushing photos back behind the scenes type photos of influencers that they tip to receive. So that's the monetization model.

It's worked very well for us so far. We have a free tier where basically for free you can talk to these influencers and you get a set number of messages. And then, yeah, on the influencer side, it costs nothing to get started. It's very easy onboarding. You get kind of recurring revenue paid out. And then we just do a rev share.

Ryan Lazuka: I think that I think this is really going to take off. We do, I do YouTube videos on interesting stuff as well in the AI space. And, you know, when I do my research, it's always I always go on YouTube and find out what's the most trending thing and things like AI girlfriends, AI companions are always like super hot topics on YouTube.

So it feels like this niche is really going to start taking off. What do you feel about companions and AI girlfriends and things like that? Because it really seems we're being pushed in that direction where instead of talking to someone human, you might just have your best friend AI person that you talk to instead because, yeah, they know more about you and you actually like, like speaking with AI rather than a human being.

Raymond Dong: Our company position is very much like we want to enable real people to amplify conversations with their the same kind of basis, which is a very, like quite, which is complimentary, but actually very different from a lot of these like AI fully fake avatars that people create, AI girlfriend, and companion. And the reason for us actually is we believe that an image of a real person will always feel more authentic and there'll be more meaningful conversations than like these AI models can get to, especially in the near comedian term. And the ability for a real person to have a personality emotions, identifying to connect with the user to share authentic images, even if it's augmented by AI, we just believe has enormous growth potential.

And frankly is much less competitive for us, right? So like if you just talk about the avatars, it's a low, much lower barrier to entry. Does anyone can just spin off an anime person and push it to a user base on whatever forums or do paid ads, but to actually create these profiles of real life influencers and to have in our opinion, this co-pilot technology, that's like massive and is a much bigger market right now. Cause we're basically taking share from all the guys like Patreon.

Ryan Lazuka: And the cool thing is like you said, you mentioned you're gonna eventually be able to do text messaging and direct messaging and stuff like that. So once that kicks in, that'll be really, that will feel real to the end user, more than I'm not saying telegram is bad. It's a great, great platform. But once you get to that point, it's gonna be really cool.

Hunter Kallay: It's gonna be very interesting to see what happens to influencers themselves, the human influencer because the trend that we see with a lot of AI projects is that it's taking care of kind of the mundane tasks or the tasks that people don't really have time to get to. And in this case, it's kind of spending more time talking to fans. And so I wonder what this is gonna free the human influencers up to do, what they're gonna be able to do next.

Raymond Dong: I mean, I think from our, and we've been working with a bunch of these influencers now that have, we've freed up a lot of their time. And I think it comes down to doing what they love doing which is creating content, looking at brand partnerships, all these other functions that an influencer has.

And then they create a cycle. Because the more content they create, the more we can feed through AI, create narratives around, push to their fans, get feedback from their fans and what they wanna see. And then the influencer has more time to actually create that content. And then it's just like a great flywheel.

Ryan Lazuka: What's the feedback you get from the influencers that they give you positively that they say, hey, this tool, I didn't even think it was gonna be able to do this, but it's done this for me. Is there anything in particular that you guys didn't foresee happening with these influencers that your tools help them with? Yeah, for sure.

Raymond Dong: I think like coming in, I knew that these tools would help better engage your user base. But I think what's been pleasantly surprising is that we've been able to extract a base of super fans.

And if you look at relevant industries, whether it's video games or dating apps or casinos, entertainment, you always have this like whale effect where like five or 10% of people drive like 80% of revenue for a lot of these companies. What we've been able to do with AI is very much finally tease out, understand and satisfy the super fans in a way that basically wasn't possible before.

Hunter Kallay: That's an incredible insight. Because you're totally right. I mean, if you think about it, these super fans are driving everything for these influencers. I mean, like for me, I'm not a big celebrity person. When I see celebrities, I'm just like, oh, you know, that's cool, whatever.

But then there are these people that it's like, they got the posters on the wall. Anything that comes out about this celebrity, anything that comes out about this influencer, I'm all over it. I'm gonna buy all their content. So if you're able to create more content and please those sort of people, it's gonna be a lot more lucrative for you. So, I mean, that's just, that's an incredible insight.

Raymond Dong: Yeah, no, we totally agree, right? And then we have our technology signals to where, if it's a super VIP engaged user, we give them like white glove service. So the influencer always prioritize them, right? They get a lot more live conversations, most behind the scenes stuff and anything in between.

Ryan Lazuka: So is it possible for like, say a super fans or any fans that are using my companion AI, my companion's AI, is it possible for the influencer to be like, well, in certain situations, post a link for me to my new product, you know, for sale. So you can sort of sell in the telegram chats as well without the, and use, what's it feeling natural? Yeah, totally.

Raymond Dong: So actually the feature that we've been working on to do exactly that is kind of gifting. So it'll actually link up to products that an influencer will have on their wish list or brand that they wanna push. So right now, in any conversation, you can buy a photo, get a whole narrative around the photo. But what we're pushing out shortly is the ability to say, hey, we're at a restaurant together and it's this restaurant, it's my favorite restaurant.

This is what I love ordering from here, right? And a fan would get that, the brand gets the exposure or the influencer could say, this is my favorite hair product. And then we could push out a direct link to that product or even the fan could tip such that the influencer can buy the product themselves.

Ryan Lazuka: Even if it's not their product, it could be like an affiliate link or something like that where they're making videos.

Raymond Dong: Yeah, yeah, yeah, exactly. And just think about like the power of AI to push that. Like what's really changed the game with AI is the ability to push all of that stuff in context where intent is much higher.

Ryan Lazuka: For sure, for sure. Like if you're, if you know you're just, if you're, if you know you're being sold to, you're probably not gonna buy it. But if it just feels natural, like heck yeah.

I mean, the sales rate's probably gonna go through the roof. So that's awesome. Exactly, exactly. And what about the, like, share as much as you are willing to, but your tech stack, like how does it work for someone that's more technical? Like are you, you mentioned that you're talking to LLMs or using chat GBT and Mistral. And like what's the behind the scenes and how is it working? You also mentioned you have a engineering team working on, you know, training the models. Can you give us a little insight on how that's done?

Raymond Dong: Yeah, sure. So I mean, we run everything through kind of Azure. Like we love Microsoft as a stack. So we use all of their tools, especially on the database side of things, on the foundational models. Like I said, we've used all of them, OpenAI, Mistro, et cetera. We use different ones for different bots.

We kind of are able to alternate between them depending on situations. You know, the front end is mostly telegram based with Stripe integration for payments. And then when it comes to the engineering team training and refining these models, a lot of it's like prompt engineering. So we're very good at crafting for each of these foundational models, the right prompts to develop the right types of conversations we want, but also fine tuning. So we take the data that comes out that we see and get feedback from for the influencers and consistently fine tune the model, that better conversations and the stickier conversation.

Ryan Lazuka: Got it, cool. And in terms of like programming languages that are using Python or things like that, or what's that look like?

Raymond Dong: Yeah, for sure. So mostly Python based right now. And then, you know, there's little kind of front end mostly telegram based products. So, you know, our site is like a standard, you know, WordPress kind of thing.

Ryan Lazuka: Are you looking to have a chat on a website as well one day or no?

Raymond Dong: Yeah, we've thought about it. I think our priority right now is to do integrations with these big platforms that have a lot of traffic already, whether it's Instagram or Patreon or any of these others, as opposed to having this like invented chat feature in our site.

Ryan Lazuka: I think we covered a lot about your product. Back to how you got started in this, you have a financial background. That's a whole nother realm in AI and finance. But what made you get into this world? Is it just because you guys were investing in companies and sort of saw an opening here to do this? Or how did you get into this specifically?

Raymond Dong: Yeah, I think that's right. Like we've looked at a lot of creator companies to consider investing in. You know, I personally should be very interested in this trend now. We've invested in a series of applied AI companies across industries. And, you know, my former consultant brand, I'm always trying to think of, okay, like what industries is AI going to disrupt next? And in a way that's like actually actionable and fast, because I think most industries can take a long time for stuff to propagate through.

But then through my work and just thinking, talking to a lot of experts, it was very evident to me that there were real pain points in this like influencer economy that AI can't address today based on where the state of the technology is with like GPT-4 and XYZ models out there.

Ryan Lazuka: Sure. Yeah. Sometimes people get ahead of themselves and think chat GPT or GPT-4 or whatever LLM model out there can do anything and everything. And they can do a lot of stuff, but you know, some things it's not good at like video, you know, it does video and people promote that it does video, but it's not great right now. So it's like, if you can stick with something that's, that AI sort of or chat GPT or these LLMs are made to do, like what you guys are doing, you know, it can be implemented right now.

Raymond Dong: Yeah, no, we agree. And I've, you know, to have some context, like I've looked at other industries, like for example, AI for finance, like just go give him my head fun background, right? That was one of my initial thoughts, like can AI help me do stock research? Can AI help me pick stocks? And I think that's just an industry where the standard in terms of how your level of accuracy that you need is like basically a hundred percent. Like, like you can't just get a wrong like stock price.

Can't get a little wrong, you know, management team for a given company, right? Like it's just such a high bar that it just takes a little bit longer to adopt and requires a lot more upfront work. For us, right, the generate AI aspect is great because we just want relevant conversation. It doesn't have to be the same thing every time. And it's, you know, if it's accurate 95 or 99% of the time, that's fine. It doesn't have to be a hundred percent accurate kind of thing. So that's why we found this to be an amazing use case. And then we're just building all of our proprietary technology around it.

Ryan Lazuka: Yeah, you bring up a good point. Like if you, if you, if your AI influencer, you know, tells your fan that they're dumb or something by accident through an hallucination, it's not the end of the world. I mean, you're not losing $10,000 on a stock trade.

So like stuff like that might happen, but it's not, you know, you're not going to lose money off of it. So very cool. What, what are your thoughts? I'm fascinated by AI and finance because most people don't talk about it. But what are your thoughts on AI being able to depict stocks and help people with our finances?

I know you mentioned that, you know, you sort of have to be right a lot more than wrong, but how come we're not seeing a bunch of AI apps out there for finance or AI stock pickers? It seems like that is sort of the next thing that hasn't been, Yeah. It hasn't been really like used yet. For sure.

Raymond Dong: So, you know, I've spent a lot of time thinking about this, exploring, building different products in this space, actually even putting out like data products actually, just giving my background. And my view for this is there's a whole cohort of people that are trying to use like real time AI to pick stocks and give just information.

That is highly difficult because the second that an investment analyst sees something wrong and if it repeats a few times, they're not going to use a product. Like, why do people love Bloomberg? Because it's like always right. And it might not be, it's the price data is real time, but a lot of the more like qualities it is not real time.

It's been like QAQC by someone. So I think the, where AI will actually penetrate finance faster, is to help with fundamental stock research to start. Where you can use AI to basically run through 10 years of earnings transcripts and earnings calls and 10 Ks and 10 Qs and give like summaries on topics you care about, such as for the software company, where were the drivers of ARR deceleration or acceleration? And they can pull out all of that qualitative context that you can then map the quantitative numbers to, to give you to basically accelerate your learning process from days to minutes. And yeah, so I think that's where AI is going to be a doctor of person finance, like fundamental research and just extracting and summarizing like large amounts of information in a relevant way. I think the more like AI actually recommends stocks or to give you like definitive answers real time, like take some time.

Ryan Lazuka: Yeah, but do you think, do you think that's going to happen? It just seems to me as an end user, you know, I've got a computer programming background, but I'm not a stock analyst at all. It seems like there's going to be a time where you're just going to be able to upload, you know, hundreds of thousands of documents on financial data for stocks in the NASDAQ 100 per se, and say analyze all this data, look at any kind of patterns and then predict what the next day is going to be.

It seems like that we're going to get to that point at some time. Do you think that's possible for AI to look at charts, for example, and say, find these patterns and then sort of predict what the next day or week is going to be?

Raymond Dong: Yeah, so I think a lot of that actually exists just like behind closed doors and mainstream yet. Like these quant funds that a lot of my friends used to work at, like a Renaissance rent tag. They've been embedded, whether it's predictive or generative AI into their processes for years.

And that's exactly what they do, right? Like, if this stock, if this barrier, or this number, then what are the chances, based on everything else you've fed into the model that it goes up versus down? Like these all signals and kind of repetition that computers do automate a lot today. I just think that for it to get mainstream, it takes a little bit of time. And then even when it does get mainstream, I think there's a large number of investors that just won't ever feel comfortable putting money into a computer to do messings.

Ryan Lazuka: Sure, when you put in $100,000 your savings and you wake up the next day and it's a 10,000 because AI did it, it's not gonna look too good.

Raymond Dong: Yeah, yeah, exactly, exactly. But for it to help you learn about companies and stocks, like that's fantastic.

Ryan Lazuka: Yeah, that might be a great gateway for people getting into AI and finance still like to research. Because the research is gonna be phenomenal. Yeah.

Hunter Kallay: Yeah, that's a good point. I wanted to ask one more question about, back to the influencers at least. We've seen some companies reporting $8,000 a month like AI influencers. And when I say AI influencers in this way, I'm talking like no humans involved. I mean, just a purely AI account posting deep fakes and fake activities, whatever that the AI is. And they're generating like crazy revenue. What do you think this, I'm just gonna leave it open, I guess. What are your thoughts on this sort of thing?

Raymond Dong: Yeah, so it's really interesting. Like I've seen a lot of those. I mean, I've tracked all the major ones like there's that Spanish agency out there now that just makes like a deep base of basically tricks that they put out to try to get Instagram following some guys and then et cetera, et cetera. I think like there's segments of the market that'll make sense for. And I think that in my opinion, like these numbers seem inflated. And I think part of the reason why the numbers are there is because there's like an innovative kind of like a wow factor that's driving a lot of press and basically money.

But then you think about like what happens down the road. Well, for someone to sustain an audience, they have to know about their audience, like have real personality, emotion, right? Like videos have to get better. I think like photos alone, which is what a lot of these like AI influencers do today won't last beyond this like near term pop.

And that's why I like to frown. So we're so focused on like real interaction, real engagement because in all of these businesses, what you'll see behind closed doors is that like you've launched something like an AI influencer. You see some pop from a novelty factor and then they typically like peter out. Like even if you look at, I've done many deep dives on meta with Facebook, like they launched these, you know, AI influencers for them is like real people, but it's not a co-pilot like we are. And you saw like a one-time peak and then like engage into Peters out.

Hunter Kallay: Yeah, I think there is a fascination with like, this is an AI personality account. And people are just really in awe of that right now. I think that you're right to say that there is going to be some sort of wave that comes crashing in a bit to taper down people's awe of this AI buzzword and these AI personalities and things like that. Agreed, agreed.

Raymond Dong: But we're still in this like high cycle band, which is in a way really good for us because we get to talk with awesome people like you guys and share the word and just, there's a still novelty factor.

Ryan Lazuka: Yeah, we'll probably hit, you know, some kind of downturn in AI where everyone jumps off the bandwagon at some point. But if you can get through that, I mean, right now it seems like that's impossible because the hype is so hot and real.

But eventually it'll be a downturn and, you know, as long as builders like you guys keep building, I think, you know, out the other end is going to come really awesome products so it'll last a long time. And, you know, I sort of compare the hype to, like when member one, you know, Instagram threads came out, like they were supposed to take over Twitter. It's like sort of the same thing. It gets so hyped up for the influencer, I'm talking in comparison to the influencers, which Hunter just brought up. But those things get so much hype for a week or two and then they die off and, you know, the real products are the ones that last. So, but one question I have for you is, if AI influencers only sort of gain more traction and they become a real thing, that's something you guys can implement in your product too if you wanted to, right? Is that something that you're looking into?

Raymond Dong: Yeah, no, absolutely. I mean, it's, our product is fully modular now where we could push out a bunch of AI influencers. In fact, a bunch of these AI influencer agencies that reached out to us and like, hey, can we just like white live with your tech for us and we can use your co-pilot with our AI influencers and the people that are the personalities behind us.

Ryan Lazuka: You're having an AI influencer have an AI chat with somebody else? Like, is there?

Raymond Dong: Well, you have an AI influencer and the person behind that influencer that created it have the chat with their fan base, right? Like you could do that on our platform. But I just think that, and I actually have written a couple of blogs about this on our website that you can read, but I think this dynamic actually accelerates the adoption of our business because influencers are now feeling concerned because they see the rise of AI influencers and they're like, okay, are they gonna steal my fan base or are they going to steal my revenue? And we're saying like, yeah, it's real phenomenon. So as a real life influencer, real person, you need to enable, like you need to embrace the AI tools that are there to help you kind of fight back in a way.

Ryan Lazuka: Yeah, because that could be it. That's a differentiator between an AI influencer and a human influencer is a human influencer can do both. They can be AI and human to the end user work. Yeah, very cool.

Hunter Kallay: Absolutely. I've been saying this over and over again. I'll keep harping this. Whatever industry you're in, need to find how AI is influencing that industry. And if you're not leveraging AI in that industry, you're already way behind everybody else.

Raymond Dong: Correct, correct. That's so true. And I mean, that's why I'm telling all my friends, like just play around with chat GPT, right? Try that integrated into your daily workflow. Look at product hunts in the GPT store and just get a sense of what are the top trending products out there that you could see some value in.

Ryan Lazuka: Is there any cool AI tools that maybe someone doesn't know about that you use every day, Ray?

Raymond Dong: I mean, I use chat GPT to write a lot of content. Like if I'm drafting a partnership email with an agency or stuff like that, but that's all like pretty standard. There's this new tool that I've been exploring a bunch that is like AI, like short form video generation, which I don't know, you guys might find interesting. Like I used to use like Final Cut Pro to generate my own videos. And this thing like basically automates all of that. So you just write your script and then it basically layers on voice, text, and even like music and like video to create like a 30 second short form video.

Ryan Lazuka: Awesome. What's the name of the tool? Do you know, Off-Amp? Called InVideo. Yeah, I've heard of that. I think they have a GPT store on chat GPT.

Raymond Dong: It's on the GPT store, but you can also use it directly on their website. So yeah, it seems really cool. I've made a few videos off of it. I'm still trying to get comfortable around the tool, but as my opinion, we push out a lot of like content, just to front of Robert Brand, talk about all the product work, talks to influencers about how the AI tools are relevant in their lives. And I'm trying to see if I can use the same video tool to like help streamline those.

Ryan Lazuka: For sure. So you could use that video tool within my companions maybe, like through an API call or something. Yeah, yeah, yeah, yeah. Totally like help edit videos of our influencers. Oh, that's awesome. Yeah, it's crazy. Like even on YouTube now, for anything, if you go on, there's a lot of AI created content that you can just tell it's AI created, but a lot of it's really good. It's more...

Raymond Dong: Yeah, yeah. My thing is if it's

Ryan Lazuka: good, it's like a watch it, you know? It's like, who cares? You made it, you know? But is there anything else that you want to promote, Rape, in terms of any other projects you have or links or whatever, now's the time to do it?

Raymond Dong: Cool, no, I mean, I think my main message is like, any influencers out there that are interested in just learning about AI, I'm very much happy to personally chat, whether it's, you know, there's my companion's product out there now, or if it's just purely brainstorming, feel free to reach out to me and love to have conversations. I think we're seeing a lot of early adopter creators start to just like bomb bar in boxes as we've been getting out there now more publicly. And so, yeah, even if our tool today isn't what you're looking for, we're adding a bunch of stuff coming down the road. So like AI, in terms of helping you negotiate brand partnerships, right?

AI to create collaborations with other influencers and to find those influencers. There's so much out there and we have like an amazing tech stack and tech team, so we're just constantly innovating.

Hunter Kallay: Yeah, so that is at mycompanions.ai, you can check all that stuff out. And then be sure to subscribe to Ryan and I's weekday newsletter for i-ai .com. We've got the latest news updates, tools, community engagement, as well as the mystery link every single day. Then on Sunday, we're diving into developments and developers and some really interesting projects like this one. And then be sure to subscribe to Behind the Bots, where you can see more interesting conversations like these.