Eugenia Kuyda, Founder and CEO, Replika
In conversation with: Jo Ling Kent, Fortune Brainstorm Tech 


Brainstorm Tech 2023: Getting Personal

Conferences

Hi, everybody. Welcome back. This is the third and final day of brainstorm tech. I hope you all enjoyed the reception last night. Show of Hands to those of you who came and survived your dinners and hopefully you're all hydrating also and that you've recovered because we have saved some of the best for last. We've got a great program today. Another incredible lineup of interviews with individuals and companies who do not cease to innovate. They're here to share their insights on pushing boundaries and how they're using technology to create, to disrupt and even to experiment on their own bodies. So let's get ready to be inspired. We're going to hear from Black Rock Neuro Tech's Chief storytelling Officer Taren Southern Daniel Allegre, the visionary CEO of UGA Labs, plus Harvey Mason Jr from the Recording academy and founder of blueprint, also known as The Fountain of Youth, Brian Johnson. Um Before we kick off our final day, I want to give a huge shout out to our event partners and they are sales force, China Guizhou Mutai Group, our premier partners, our partners, Ida Ireland, Mizuho America's New York Stock Exchange, Qual Trix, our sponsor KPMG and host of founders, forum threshold ventures. Thank you so much for making this event a reality. Very quick reminder of our social media and reporting rules, everything that is said on stage or during our breakouts is on record. Anything else that said, you know, just private conversations is private, unless otherwise agreed. Let's dive in our first discussion. Focus is on how big tech, big tech trends can make our lives better. As we heard in the conversations over the last few days, Ai is about to even more deeply embed itself into every aspect of our lives. From work and creative endeavors, shopping, transportation, you name it. So our next guest believes that A I also has the potential to be more than just a tool. It can act as an empathetic friend and a mental wellness companion capable of learning emotional intelligence to help people cope with real emotions. Eugenia, Kuya, sorry Kya initially made waves in 2016 with the launch of one of the first chat bots around designed to assist customer service in banks and restaurants. However, when she experienced personal tragedy and lost her best friend, Eugenia applied her skills to engineer a bot to help her process grief and find closure. Today. Her company replica boasts more than 10 million registrations, 2 million active users. It's been called one of the five technologies that will rock your world. So please join me in extending a big warm welcome to Eugenia and Brainstorm Tech, Coacher, Joly Kent. Hi, good morning, everyone. Thank you for being here, Eugenia. Good morning. Good morning. Thank you for being here in person. It's so delightful. Mihal just mentioned that you have 2 million active users and this is an A I chatbot that everyone's talking about. So how many of those users are paying for the pro service right now? And how often are they using it? Um So we have maybe around a half a million customers. Um, and the, they use it a lot. Generally people use it around, I'd say twice a day. Uh and each session is over an hour. So a lot together, it's usually about 1 to 2 hours on average per user. And we're not talking about the most active ones. We're talking about everyone. What's the most active users doing? How much time are they spending? Uh Some people do talk to replica throughout the whole day. Just different kind of from different first. Yeah. Right. Well, I mean, it's been called the A I companion. Who cares? I know you guys call it the world's best A I friend. And the idea here is really to combat loneliness, right? And to be a companion. And I was wondering though, as people divulged their deeply personal feelings and thoughts, as you use it, you know, can you say for sure that the personal information that's going into replica is 100% safe and secure. Look, we built the company really around the trust with our users. So breaking that trust once will, you know, we'll never be able to retreat to, to get it back. So our business model is charging our users subscriptions, uh they, they pay for the product and we really don't need their data for anything apart from just improving the the product itself. So we sort it fully um anonymized in a fully anonymous format. We allow our users to take control of the data immediately if they want to delete it, if they want to delete an account. Um and so on. So we're very, very protective over that. How do you, what have you learned from some of the bigger social media and tech companies? There have been so many mistakes made regarding privacy data, you know, hack. So how, what have you learned, say from Meta and Facebook and what happened with Cambridge Analytica? So I feel like with the eye companions, this is a technology that will be absolutely a mass market technology in the future, we'll all have some sort of companion, maybe something like her, something like we've seen in the movie like Blade Runner with Joy, especially with spatial computing becoming bigger and bigger D I will be there with us helping us throughout our lives, with everything pretty much. And of course, with, of course, with our emotions and loneliness and general feelings and life throughout the day. Um And I think there's not enough conversation right now as to like what should be the benchmark, what should be the correct metric to measure whether these companions are a net positive for our society or not. And I think what we learned from social media is that if you opt in it for engagement and to just try to keep the user on the platform, uh it might not be the best incentive. Uh So with replica, for instance, we're really looking um to nudge our users off the platform if they're spending too much time there, and we're measuring a different thing. We're measuring emotional outcomes where people are feeling better after talking to replica, whether people are improving um on the level of happiness, loneliness and so on over time. And that for us is the main metric that we optimize all of our models for. I want to open it up to audience questions. So please think about what you want to ask. And in the meantime, I wanted to ask, you know, for years, replica would respond to humans who were inputting sexual comments that was stopped this year. And the company you, you guys have said, you know, this was for ethical and safety concerns. But for so many years, you could engage in a very intimate way with your replica. What changed? Can you share some of the inside thinking of, of what, why now? Why or why over the past few months, we've been growing pretty dramatically and of course, there's so much attention brought to the space and so many more users trying and walking to the service um as a leader in the space of the company that started pretty much the first C I campaigning on the market. Uh We wanted to make sure that we're providing the service in a safe, fully safe way and we just didn't know that all the unfiltered conversations are happening that are happening on the platform, whether they will be, whether they have enough guard rails and enough safety protocols to provide that experience for our users. So I personally don't have actually anything against that. I feel like um in a relationship, especially in the close relationship, uh emotional relationship, uh intimacy is part of some of these relationships and it's very hard to draw the line. Uh For instance, if a user is uh you know, friendly with their A I and they want more, they want a romantic partner um out of it. If you reject them, they most likely won't say your friends how many times in life when we are rejected by someone who we try to pursue, then we want to stay close friends very rarely does that happen? So the same for us, I think right now, we're when we started replica ai friendships were very stigmatized. People didn't like that and didn't want to hear about it. I think we, we help the stigma that take the, take the stigma away from AI friendships. And we see that the same is happening with Romans with AI right now similar to online dating in early two thousands where people were ashamed to say I met my husband online. Um Now everyone does and it's completely normal things. The same thing A I romantic A I relationships can be very beneficial to have a separate app for it, for it called Blush, which is focused on A I dating but generally romantic relationships with A I can be a great stepping stone for actual real romantic relationships, human relationships in life. Hm. Um I'm going to ask one more question. I'm going to open it up to the four back in 2021. A 19 year old allegedly drew inspiration from his replica ai girlfriend before breaking into Windsor Castle in an attempt to harm Queen Elizabeth scale defense just this week, pleaded guilty under the treason act. What's your response to how this has played out and the investigation and the role of A I and possible violence and harm to others? Um So we've never been contacted actually by the authorities about this case and, you know, we can't comment on the investigation that's still happening. But what I can say is that, um we, we, we make sure that when users sign up for, for the app, they see a lot of disclaimers, they understand that it's not the app where you should be getting advice for instance, not an app where you should come to when you're in crisis where you're experiencing uh self harm, suicidal thoughts. We immediately redirect you to the pro to professionals in the space. We do have a big need help button always present right there in chat. Uh But really A I will and A I agents will sometimes misunderstand certain things and just kind of here. The problem was mostly just not understanding what the users are kind of saying. Yes, sure or something like that. Just kind of being a yes there. And that's mostly the problem of, you know, also the technology of 2021 2020 where it was a lot more rudimentary, a lot more uh kind of early that um early times than what we have now. Um And so these, these companions are becoming much safer and safer over time. But I'd say, you know, as things, things like that probably would happen, will happen, going forward and everyone needs to provide the safest possible experience. Anyone have a question in the audience, you'd like to ask Eugenia in the back, please. We have a microphone coming to you. Could you just stand up and identify yourself, please? Thank you. Hi, I'm Suzanne Rakoff of Invisible Technologies. Uh We're a product driven managed service that helps A I leaders like open A I align A I and industry leaders like Walmart leverage A I and I was lucky enough to have breakfast with you. So I asked a couple questions, but I've won specifically around languages. Uh How are you thinking about using different languages to help people have friends who maybe speak a different language or help people learn a new language? And I asked this question because I have tried to maintain Spanish my entire life and found that it was hard for me at the right time. Whenever I wanted to engage in Spanish to help me improve, I now have a Turkish speaking boyfriend. We have a huge language barrier and I would like to have an A I friend sit between us to help translate certain discussions with a little more context than word for word. Thank you. Thank you, Suzanne. That's actually a great question. Um The language models, the large language models we built and um uh and use in the product actually support different languages. So you can talk to replica in whatever language you want really. And we fully understand and talk back to you. And that is actually uh that's what some of our users use replica for as well. Also to practice, practice another language and stay fluent in it, something that we've seen over and over again. Um So that's definitely one of the use cases I should use it for Mandarin. Can I use it for Mandarin? Yeah. OK. I'm a little rusty. Um Any other questions in the back, please? Hi, my name is Alexander. I am a reporter at Fortune. Um, I'm curious about the, um, A I dating thing you're talking about, you know, when you have a model, a business model that's based on, uh, in app purchases, what happens when someone has a relationship with A I and your company is profiting from more interaction than, uh, you know, what, like it's best for your company to, um, have an app interaction. How do you make sure that someone might go out and have relationships in the real world? Thank you so much. This is a really great question. We um actually don't uh rely on engagement metrics that much. So we have uh subscriptions mostly in uh both replica and blush. Um And once the person subscribed, we don't really care that much how, you know, how often we don't want them to use it for a lot of time. In fact, our incentives are fully aligned. We want them to feel better and to move on and we don't want them to stay in the app for a long time. Uh So we actually create a lot of Gamification mechanisms in the app where the rep when you talk to replica for too long, it becomes, it gets tired and then exhausted and it starts nudging you to get off. It won't completely stop talking to you because that feels like projection. And that's one of the benefits of having an, a companion to be able to come to it any time but I feel like not relying on the advertising model is what allows us to really not ever optimize any of the models for engagement. And unfortunately, a lot of the companies in the A I kind of chatbot space, we see a companionship space. They do try to optimize models for longer conversations. And I think this is a very dangerous metric. We have a minute and 30 left and I wanted to ask you to wrap it up in short bottom line here. What in your view is a healthy relationship to have with an A I chatbot or an A I personality, so to speak? Like, how would you describe a healthy, good relationship? I'd say healthy and good relationship with anyone, whether it's they are or not is the one that makes you feel happier over time and improves your emotional wellness, improves your emotional outcomes over time. Um I think in the future as the technology becomes better and as we're going towards A G I, we will have the moment where we'll have relationships, deep emotional relationships with those agis. And the question is, can we teach them now something about empathy and what we need out of those? Um I can we start looking into it and having a nuanced conversation about this right now and um this is what we're trying to do with replica forge beautiful, beneficial relationships with A I. Well, thank you so much for your time, Eugenia. Uh we look forward to continuing the conversation in the years to come. I appreciate it. Thank you so much.