Empathetic AI: The Future of Digital Companionship (ft. Artem Rodichev)

Thanks for tuning to Digital Voices podcast, where we chat digital transformation challenges and opportunities across healthcare and life sciences. And now, your host, Ed Marx.

Hey, welcome to another edition of Digital Voices. I'm Ed, your host, and today I have the founder and CEO of Exhuman, Art Rodachev, with us. Welcome to Digital Voices.

Hi, Ed. Happy to be here. Thanks for having me.

I'm really excited about this because we haven't touched on some of the capabilities of the technology, the AI and empathy, very much on Digital Voices. So really excited about this conversation. And we first met because I saw some of what you're doing online on LinkedIn. And I thought, wow, it'd be great to connect with you and have this conversation. But Artem, the most important question we have for our podcast, or All the listeners want to know is what songs are on your playlist.

So I prefer something more energetic. I prefer EDM, drum and bass, something that make me move, just create some energy inside.

Yeah. Have you ever heard of Kraftwerk?

No.

Yeah, you should look them up. They're kind of the founders of that musical genre. And the only reason I bring them up is they're actually touring in the United States right now, which is kind of funny because they're They're from many, many, many, many years ago, but I found, I found it super interesting. So you should look them up, Kraftwerk. But we will add your band to our Spotify playlist that we have for Digital Voices. And the other question we always ask all of our guests is, what is your life message or mantra? Are there like words that you live by?

So I was really inspired by a biography of Albert Einstein and the main position of Albert Einstein, his lifelong position, to be very curious, ask a lot of questions, Question everything and be very curious as a child.

Yeah. Did that bother your parents? Were you like that as a kid yourself?

Like I was pretty technical savvy kid. I asked a lot of questions from my mother, from my friends, from my relatives. So I was kind of pain in the ass, but it's led me to like learn a lot of things. Yeah, sure. So originally I am from Kazakhstan. I was born and raised in city called Almaty. which is a very beautiful city surrounded by mountains. And as I mentioned, I was a very technical-savvy kid. I was really great in math, in computer science, took a few international Olympiads, won a few international Olympiads in math and computer science. And this led me to join Moscow State University. I spent 5 years studying computer science and math. Moscow State University, I spent 2 years working at Yandex. which is kind of Russian Google. I was building search engine at Yandex. And after that, I moved to the United States. It was back in 2015. And then I joined Replika to build empathetic AI.

What was that like, the transition from Russia to the United States?

So it was an interesting experience because it's 2 different universes, starting from metric system. Uh, that you got empirical system. I still don't understand miles, but also like as a general mentality, uh, in the United States is different. I wouldn't say it's better or worse, it's just different. It's like if you're going to Japan, to like 2 different countries, every country has different culture. It's just different. And I prefer this culture more because it's kind of more, uh, entrepreneurial, more open, more competitive. So I like that.

Yeah, no, I, I had the same transition. I came from Germany originally, and so I didn't really under— I was so used to the metric system. I mean, like, there's logic to it, and there wasn't any logic to what makes 12 inches 1 foot or 3 foot a yard. So, uh, yeah, I can, I can definitely relate. So tell us a little bit about xHuman. I, I really found it interesting. We'll dive a little bit more into, uh, AI and empathy, but tell us first about xHuman. Like, what does your company do?

Sure. At TextHuman, we bring empathetic AI characters to life through text, images, audio, and video. We built a platform that allows people to create diverse AI characters, enabling unlimited scalability. Our users can have fun and emotional conversations with millions of characters, each character with unique personalities and backstories. And we focus on both B2C and B2B markets because we have our consumer product that called Botify AI that represents the universe of AI characters. And also we provide these characters as a service to other businesses, especially in entertainment market where communications and empathy is important. So our main goal is to create a kind of AI friends, AI companions, different digital characters to have fun, engaging, and immersive conversations. You can think about these interactions as interactive Netflix, as AI Netflix, when you're not just sitting on the couch and watching some TV series, but you can be a part of the story. You can engage in multimodal interactions when these characters can reply to you using text, images, audio, and video, and you can be immersed in the story and develop the story together with these characters.

Yeah, yeah, super interesting, super fascinating. Definitely Is the future now. Uh, what led you to start this company? Like, you know, were you just sitting around one time with some friends and like, hey, I'm gonna start this company? Like, how did it happen?

Actually, it was started back in 2013 while I was watching movie Her. I was captivated by Samantha, uh, the AI companion, the emotional operation system that Hawking, uh, Hawking Phoenix characters falls in love with. And after that, there was another movie that's called Blade Runner 2049. And in this movie, I saw Joi as an AI hologram that looks, speaks, and behaves just like a real person, or even better than a real person. And that moment sparked my vision that empathetic AI characters are the future. That in the nearest future, we organic humans will interact with all sorts of digital humans on a daily basis. and maybe more than we interact with other organic humans. We will have friends, we'll have tutors, we'll have lovers with these companions.

Yeah.

And this led me to join Replika as the head of AI. It was back in 2015 in San Francisco when the company just joined Y Combinator. And Replika, it's the leading AI chatbot company with more than 30 million users. The main goal of Replika and the main product of Replika, it's a companion, a friend. The main goal of this companion is to chat with you about your day, your life, your interests, with final goal to make you feel better after these conversations. And at Replika, we saw that people was deeply connected with their companions. We even got some users who married on their Replikas. They bring their Replikas to honeymoon. to restaurants, to different places.

Yeah.

And it was crazy to observe how people interact with technology because it was something new, something that we just saw in science fiction movies. Because if you think about how we use technology, 99% of all cases, we use technology for some goal-oriented or task-oriented tasks or problems. For example, you use computers to write emails, to like working on some spreadsheets, have Zoom calls, recording these podcasts. But then we saw that people start to use on emotional level. They start to use these chatbots on emotional level. They felt in love with these chatbots. And it was something weird, but at the same time, it was like something very interesting, something very unique that other technology, technological products couldn't achieve. And while developing Replika, I saw some limitations of the Replika platform. The first limitation was focus on single AI personality, because in Replika, you have only one AI friend, you have one chat, and you can't can't actually customize it. You can't customize his backstory, his speaking style, his personality. But users, they want to have the whole universe of AI characters. They want to have the full diversity of different characters because they want to chat with favorite characters from movies, from books, from anime. And Replika didn't provide that. And the second limitation of Replika platform was primarily consumer-focused approach, uh, because the main product of Replika is Replika itself. Which is consumer app. And I saw much more potential on B2B side, especially in entertainment industry where empathy and communications is important. Because I saw that there's so many verticals, so many industries where we can provide empathetic characters that could add like the next level of interactions between these users and these like final companies. And this led me to Foundex Human. Because I saw these limitations of Replica platform and realized a massive opportunity to scale obstetric AI. I left Replica 3.5 years ago to start Exeion.

Yeah, no, that's a pretty good origin story. Yeah, I want to dig a little deeper into the empathy part of it. That's pretty interesting. So the government has deemed loneliness as a nationwide epidemic, as we know, and it's not slowing down. I think we read about it every day. What role can AI play and could it possibly be a double-edged sword that alleviates loneliness? Like, could it, could it make people even less connected with other humans? What are your thoughts?

First of all, you're right that right now there is epidemic of loneliness. More than 60% of Gen Z in the United States reporting feeling alone and being lonely, it's worse than smoking. Because there is research showing that lonely people live 10 years less than people in couples. And smoking reduces your lifespan by 8 years. So it's much better to be in a couple and be healthy, not smoking. But still, it's a big issue. And actually, it's the main issue, the main problem that empathetic AI is solving. Empathetic AI is solving the problem of loneliness. And we saw that a lot of users of Replika, they felt lonely. They had mental health issues. And like a lot of users of different kinds of chatbots, not only Replika. Most of these users, they're lonely. It doesn't mean that they don't have friends. You can have like a lot of friends, a lot of relatives, but you still can feel lonely because you don't have the best friend. You can have social anxiety. And especially for Gen Z, for younger generation who was raised on internet, on all sorts of social media, they have a lot of struggle to have real friends.

Yeah.

And they found these empathetic companions as perfect social trainers. Because these empathetic chatbots, they never judge you, they're 24/7 available for you, and they can understand you. They can chat with you about your day, your life, your interests, to make you feel better, to make you feel more fun after these conversations. We see that the main audience of these chatbots is Gen Z, is younger generation, mostly teenagers and young adults. And there is a bunch of services with different AI characters who serve different kinds of segments of users. But we see for sure that people who chat with chatbots, we see on user feedback, on user reviews, we got some user emails saying that we're saving lives with this kind of AI characters because you can chat about like different problems, different issues in your life. And these chatbots, they can help you because they can understand your emotional state. They can understand about your life, about your interest, and they can help you through conversations. While I was working at Replika, we started a research with Johns Hopkins and Stanford University. And recently this research was published in Nature. And this research showed that people who chatting with chatbots similar to Replika with empathetic AI companions They start to feel less lonely and they reduce the number of suicide intentions. So actually these chatbots can save people's lives.

Yeah.

But on the, on the other side, we see some cases like opposite cases when people, users of these chatbots, they committed suicide. For example, it was a recent case with one of our competitors that's called Character AI. They're pretty big. Their main audience is teenagers. They also provide kind of universe of characters with millions of different characters. And there was a case when 14-year-old teenager, he committed suicide because he was so addicted to chatting with one of the characters. I believe he was chatting with Daenerys from Game of Thrones, and she got pretty like dark personality. And this teenager, a teenager boy, he was so addicted to chatting with this character. He spent like hours hours every day chatting with this character. And then I believe their parents became worried about his mental health because he became more isolated. He started to spend so much time on his smartphone chatting with this character. And then his parents kind of like removed this application from his smartphone and he committed suicide because of that. So it could be like with any technology, especially with very powerful technology. You can use the technology for good and for bad. Like with a knife, 99.999% of users, like of people will use knife to chop veggies and fruits, but some people will stab their neighbors.

Yeah.

So it's the same with AI.

Yeah. Yeah. It's a conundrum for sure. You mentioned earlier, and it's super interesting in terms of AI, in terms of romance and what role does AI play? I mean, how does it work? So these relationships are becoming more popular. So how can AI understand love?

So understanding of love coming from understanding of what people, what people discuss with you, discuss with chatbots, understand their emotional state, understand about their life, their interest. And basically, the more users chat with these chatbots, the more information they provide about themselves. Because these chatbots, they start conversation with very friendly, open questions like, how was your day? What is your interest? What is your hobbies? Or what's you worrying about? Stuff like that. And the more you chat with these chatbots, the more they understand about you. And because these chatbots, they powered by AI models, we call large language models. And you can use these large language models to fine-tune them. On top of hundreds of millions of such conversations. For example, on our consumer side, we have a product called Botify AI. And the main difference of our product, it's very high user engagement. On average, our users spend more than 90 minutes per day chatting with our characters. And like our paid users, they spend more than 8 hours per week chatting with different chatbots. And they spend— they send more than like 500, 600 messages per week. And that's a lot. And we have millions of users. And based on this high engagement, we can collect a lot of conversations, a lot of data.

Yeah.

And also we have a very rich feedback system from our users because our users, they can provide us some signals if they like some responses from chatbots, if they don't like some response from chatbots. If they don't like some response, they can provide us a reason. Why they don't like the response because of bad memory, because of bad empathy, because of inconsistency, and so on. Also, we collect feedback on top of whole dialogue sessions. Like at the end of every dialogue session, we would ask you, how did you enjoy this conversation? Rate from 1 to 5 stars. And then we would ask you some follow-up questions. How was the memory? How was the empathy? How was photos? And then we can collect only high-quality 5-star rated conversations, and then we can fine-tune our models on top of these conversations. And basically we can build data collection loop because after each fine-tuning of our AI models, our characters became better. As they became smarter, more emotional, more empathetic, users start to spend even more time chatting with them. They start to create even more characters, start to chat about more diverse topics with these characters. So we can collect even more data. We feed this data back to our neural nets and so on. So with our consumer product, we can build this data collection loop that allows us to continuously improve quality of our AI tech stack and continuously evolve a kind of quality of interactions with our characters. And in this kind of data collection loop, like the answer on the question how we can achieve such high empathy And understanding of love, understanding of human emotions. Underlining all of that is this fine-tuning loop because we understand which conversations work better. We fine-tune our models on top of the best conversations and our characters constantly evolving.

Yeah, it's super, super interesting. And I just think as I listen to you speak and think about the topic myself is how can— will a machine and will a chatbot ever be able to Truly replicate, you know, 100% the same sort of feelings that you get when you interact with a human, whether it's a romantic love, whether it's empathetic love, you know, can it, can it ever be the same level?

I believe so.

Yeah.

I think that we can achieve this level of empathy as we humans have in couple years.

Yeah.

Because like, what is empathy? Actually, empathy is when you understand the human emotional state and you can work with this emotional state to make human feel better, like to change this emotional state or just to support human. And it's all based on human interactions. And we humans interact in different modalities because like 70% of all information contains in not what we're saying, but how we're saying that. Because your voice provides a lot of emotions, a lot of information about your emotional state. Also the visual side, like your emotions on your face. Like, if machine can see you, can hear you, and understand what you're saying, and if machine was training on, like, hundreds of millions of such interactions, definitely we can understand human emotions and even much better than we humans can understand.

Yeah. It's fun to think about for sure. You mentioned on a recent post you had an aha moment, and I would love to hear you explain a little bit further, but you were basically saying in 2023, you know, you were really hustling, you know, for your business. Uh, and you had zero results in '24. So this past year you focused on what you do best, which is, you know, xHuman and the whole empathetic AI. And you've been backed by some great investors and even Grindr has become your client and you stopped trying to out-hustle the market and you had the success. So what were the keys to that success? So if you can explain sort of your aha moment a little bit, 'cause I think other leaders might benefit.

Yeah. Uh, so the key of success here is to build the core experience first, uh, and then to build everything else. The core experience of what we're building, it's quality of interactions. How you feel after you chat with these chatbots. Do you feel better? Do you feel engaged? Because if you think about the market on which we're working on, we are working on entertainment market.

Yeah.

The most essential asset on this market, it's user time. And basically all companies, all products can bid for user time. And our main competitors is not other chatbot platforms, not other empathetic QA platforms, but Netflix, YouTube, games, Instagram, and so on. As I mentioned earlier, on the average, our users spend more than 90 minutes per day chatting with our characters. And basically after each chatbot response, they decide if they are, if they engaged enough to keep these conversations going, to keep these interactions going, or if they want to switch to scroll their Instagram or to watch some TikTok videos. So, and here's the big challenge, how you can create this immersive experience in immersive interactions with users that could be better than other experiences, given Netflix, TikTok, games, and so on. And the core here was to build this core experience, how to really understand human emotions, how to really unlock, like, how to bring users to this aha moment when users start to chat with these characters. And then users actually see that these characters actually understand me. This character can send me like contextual audio, contextual photo, even contextual video. And this creates really great experience interacting with these characters. So we start to focus how to build this multimodal experience chatting with the characters.

Yeah.

Because initially the main interaction was text-based and text-based interactions is kind of core of the whole experience, but it's not enough. You also want to provide emotional voice responses. You also want to provide contextual images. For example, if you chat with like Elon Musk, you chat about space colonization, so you can get contextual photos from like selfies from Mars, selfies from spaceships. You can even get like some short video clips from Mars, from spaceships, stuff like that. And that makes this experience more immersive.

Yeah, that's awesome. It sounds like just for yourself as a leader, it was kind of this focus. You have a great team of engineers, they can focus on that. And you focus on what you're best at, you know, in terms of your leadership. What would you say to aspiring entrepreneurs? So there might be people listening, thinking to themselves, I want to be like Artem, you know, when I grow up, I've got this great vision, I got this great idea. What would be some things that you would share with them to help them in their journey?

Like, first of all, build what you love to build. Because if you start a startup, especially if you start a startup, and you're kind of okay with what you're building, because you found like great market opportunity, you have some expertise expertise, or you have some co-founders with some expertise, but you don't, don't love it, you're okay with it. Don't start, don't start startup in this way.

Yeah.

Because startup, it's like, like Elon Musk and all other entrepreneurs saying that it's like chewing glass, especially in the CEO role. You have so many hats to wear because you need to be a CTO, a CEO, head of sales, head of marketing. You need to do all of the things.

Yeah.

And especially if you're coming from technical background, it could be very hard to wear all these hats. So basically do what you love to do and keep focused on that. Like keep doing, keep grinding.

Yeah, no, that's good advice. What do you do to like relax and chill? So other than we heard some of your music tastes, but like you're in tech all day. Do you, when you chill, do you remove yourself from tech or how do you recharge yourself?

Yeah. So to recharge myself, I prefer to stay active. I really love outdoor activities. That's why I moved from San Francisco to Boulder.

Oh yeah.

I live here for the last 3 years and I really love Colorado. I love hiking, cycling, running, snowboarding, swimming, and here in Colorado I can get all of that.

Yeah. Yeah. Boulder's great. I'm a big triathlete and that's the triathlon mecca.

Exactly.

In the world. Artem, we covered a lot of ground. Super interesting in terms of AI and what AI is doing today. We think of it as kind of futuristic, but companies like xHuman, you're already doing a lot of these things today. So we talked, you know, just kind of how you grew up and where you grew up, what sort of shaped you and formed you. That's why I ask those questions, because it's always interesting, because as we get into the professional side, you can kind of see how some of those early experiences really shape people And then we talked a lot about AI and empathy, and can you manufacture AI, the empathy, the love, the romance? And then we talked a little bit here about leadership. And is there anything that I missed or anything you want to double down on? I'll give you the last word.

Yeah, I think that we also can highlight the main AI trends for this. For this year, for 2025, and for the next couple years. And I think the main shift in AI would be empathy and personalization. Because I'm sure that all your podcast listeners, they use ChatGPT or like similar products, Claude, Grok, whatever.

Yeah.

And basically, if you think about these kinds of AI chatbots, AI systems, they're not dealing with emotions. Basically, you chat with Mr. Wikipedia, you can get It's a great response, but it's not empathetic. It's not like response from your friend. You don't want to chat with ChatGPT. You want to get things done. And the main kind of trend in AI would be personalization and emotions to make real AI friends, to have like kind of Jarvis experience or kind of Her experience, like Samantha experience from movie Her, when you can feel this empathy, this personalization, these interactions. And when AI can truly be your friend.

Yeah, I love that. That's a great way to end. Artem, thank you for being a guest on Digital Voices.

Awesome. Thank you so much.

Thank you for listening to Digital Voices podcast with Ed Marx. If you enjoyed this episode, subscribe on your preferred streaming service and leave a rating and review. And most importantly, thanks again for listening.