- 3 weeks ago
Karandeep Anand, CEO of Character.AI, explains how the controversial platform is taking responsibility for minors engaging with AI chatbots and adult content
Category
📺
TVTranscript
00:00I'm Justine Miller here at TechCrunch Disrupt in San Francisco, sitting down with Karandeep
00:09Anand. Thank you so much for being here. You are the CEO of Character AI. Tell us a little
00:15bit about what Character AI is for those people who might not know and how long it's been around
00:19for. Indeed. Thanks for having me here. Character AI has been around for over three years now,
00:24and interesting trivia fact, Character AI was born before ChatGBT came out. So the idea
00:31that you could speak to AI, super intelligence, chatbot, and have fun, engaging answers is actually
00:38almost three years old. And today, the platform is the leading AI entertainment platform. Effectively,
00:44people show up to Character AI to role play, to have a little bit of game and fun, and they
00:49create these fun characters. You can create your own characters or characters people have built,
00:53and then they go on these amazing world-building quests, and they're spending hours and hours,
00:58and effectively, this basically replaces the time people used to spend on the TikToks of the world
01:04or Netflix of the world. So this is the new form of entertainment for Gen Z and Gen Alpha.
01:08And what really differentiates you from the many other AI platforms that are out there
01:12that are all doing different things? How is yours different compared to those?
01:16I think when everyone else in the industry was chasing how to solve your personal assistant task,
01:22like how to become the smartest coding assistant, smartest personal assistant. For the last three
01:27years, Character has been singularly focused on how to build the best entertainment platform.
01:31So we are not your tool for getting the recipes or directions or solving math problems. This is a tool
01:38for how do you have entertainment. And we've been singularly focused on creating the most engaging,
01:42fun entertainment experiences, not just through chatbots, but also through multimodal experience
01:46like short videos, short audio books. So in that sense, we didn't change the utility
01:51path, which was the entertainment path. And hence, we are reaping the benefits of being the first
01:57and the biggest one now.
01:59And how are these digital personalities kind of reshaping the human experience with interaction?
02:04It's amazing to see the kind of creativity people are able to bring on something which I would have assumed
02:10to be pretty simple chatbots. But you realize, like, you know, all of us have this inherent creativity
02:16locked into our heads, and we need a safe space to go experiment and express ourselves. That's what we see
02:22with a lot of the users on Character. They come here to explore new worlds, create new worlds, and roleplay.
02:27One of my fun stories is one of the users, she's not a gamer, but her husband plays Call of Duty.
02:34So she actually created Call of Duty characters on Character, engaged with them so that she could have
02:40a more fun conversation with her husband. I'm like, wow, I would have never thought of that use case.
02:44So it basically, you know, gives you some glimpse into when you have this super amazing, super powerful
02:50AI tech, but you can actually make it a lot more human-like. You can explore worlds and situations
02:56and conditions which were impossible to have explored earlier, yet doing it in a very safe and non-judgmental way.
03:01And I think that's how the whole idea of what a human relationship with a machine is very rapidly changing.
03:06And how is that for you, having the experience that you've had coming from Microsoft and Meta and other tech companies?
03:12This is very different than that. How have you kind of brought these sensibilities of what you've learned there
03:16to something like this that's so different?
03:18I think the Meta experience probably was the closest of building something which is very large internet scale,
03:23yet a compelling consumer experience. But at the same time, I think we've also brought some pretty hard lessons from the time at Meta,
03:30because I think when we were building social media at scale, we didn't think about the downstream implications
03:35of the good and the bad that can happen with social media. And that's the reason why I think I get an opportunity to go
03:42build something brand new, like character AI, but this time we're doing it rightfully, you know, with the right constructs from the get-go.
03:49A simple example is, in social media, people got hooked on to doing scrolling. Everyone's scrolling.
03:55And I have a six-year-old, and I do not want her to grow up on that. Luckily, she's super creative,
04:00so she does not want to spend hours consuming content. She really wants to create content.
04:04So I think my biggest learning, to answer your question more precisely, is I think we can build a very, very compelling
04:10consumer-grade experience, like what Meta builds, but do it with the right guardrails in place.
04:15So people are actually growing with the platform and getting creative as opposed to just room scrolling.
04:19So I think the past helps get wiser and get better at doing something which will be a super long-lasting impact of the world.
04:27And it's something that so many people have been so excited to jump onto, right?
04:30Indeed.
04:31You have, I think, 20 million monthly active users, right? So tell me why you think so many people are so excited to give it a try.
04:37I think it's a combination of the very, very novel format, where people are like, wow, I no longer need to just read a book.
04:45I can interact with the characters in the book. If I don't like how the book ends, I can change how the book ends.
04:51I'm listening to an audio book, and I'm like, I really want to pause and ask the character more questions on why to do it.
04:57These things were magical. Kind of goes back to, you know, first time I read Harry Potter.
05:01And Harry Potter is mentioned of these images or photographs of Harry Potter's parents are not static.
05:07They're actually moving images. I'm like, wow, how cool is that?
05:10Now think of that being images are not only moving, you can talk to those characters.
05:15I mean, just sheer creativity that it unlocks, it just brings in a lot of users who are like, wow, this is what the next generation of entertainment will be in the world of AI.
05:24And I think you have had so much success and so many people who've been so excited about it, but then also some controversy as well.
05:30And so tell me a little bit about what that's been like trying to deal with all that and how you're, you know, trying to help make this platform safer and kind of those ethical implications, especially when it comes to kids.
05:41Indeed. In fact, we made some big announcements this morning, and I'll cover that in a second as well.
05:47But I think the overall point is valid, which is when you are on the bleeding edge of technology, there are a lot of things you learn on the job.
05:55And, and trust and safety is one thing which, you know, even though I joined Character only six or seven months ago, there were already some hard debates happening on like, hey, is AI safe or are you not safe for teens?
06:07Uh, and we've been doing a lot of work in making sure that we build a safest under 18 experience.
06:12Uh, and, and what that means is changing very rapidly.
06:14So, so the announcements we made this morning, well, we were shutting down all access to chatbots for under 18 users.
06:20That's a very bold move because that's what the core product was, but, uh, it's happening simply because we want, we want to do what's right for the world.
06:27But also we've learned that there are far better ways to engage users, which are audio and video and short, short image generation, short storytelling, than just giving them access to open-end chatbots, which may or may or may have long term implications that we don't understand yet.
06:40So I think it's a very interesting trade off that you constantly have to break on how do you make something safe and get engaging.
06:47How do you create something which is on the cutting edge, but not get too far ahead of the skis?
06:51I think that is what makes my job fun.
06:52And that's what I was going to ask about next is the balance of all of this.
06:55Yes.
06:56Because you obviously do want to make it safe.
06:58But again, if you take away so many capabilities, then it's not even being used to its full capacity.
07:03So how do you find that balance, especially when the consequences can be life or death?
07:08Indeed.
07:09And that's the reason why we spend a lot of our energy on creating a, you know, completely different experience for under 18 and over 18.
07:15Where with over 18 users, adults, we can treat them a little bit more like adults and give them a lot more leeway in, you know, experimenting and we learn and watch very carefully.
07:24But one other thing, you know, we've been very fortunate as a platform is we have a very large creator community.
07:30So it's not just users.
07:31There are a lot of creators who are on the cutting edge of wanting to explore AI with us.
07:35And these are the users who are trusted testers.
07:37They're part of our creator club.
07:38They're part of our advisory board.
07:40And they help us navigate a lot of these challenges and trade offs up early on pretty well.
07:45So I think the fact that we've been so in tight feedback loop with them and the fact that we have history of like, hey, what went wrong with social media is probably the best way we can today to go beyond the cutting edge, but still not completely break into jail.
08:01And when you think about that new process you have with under 18 and over 18, what kind of verification process or steps do you have to actually determine if a user is under 18 or not?
08:12So that's a space where I think the entire industry is still figuring out, hey, what's the right way to balance users' safety and users' privacy?
08:20Because there are extreme views today on, oh, you should age verify everyone and everyone should be uploading the government IDs to know that, hey, they're over 18.
08:29And I think we can do better.
08:31So what we're doing and I think today's announcements are very critical.
08:34If you're going to create such a different experience for underrating over 18, we need to make sure that people who say they're over 18 are actually over 18.
08:40So we're rolling out a series of age detection algorithms that we've built in-house so we can determine what a user's age could be regardless of what they enter as their age.
08:50But we're also partnering with third party services like Persona, which are, you know, they're industry leading, very, very reputed players who have a lot of signal on what the age might be.
09:00A combination of those to help us detect the age.
09:02And when in doubt, we do fall back onto age verification.
09:05And that's everything from biometric scan to government issued IDs.
09:09But we use that as a last resort because, you know, a lot of people are very sensitive, as they should be, in handing out the government issued IDs to a whole bunch of startups.
09:17I mean, I love the fact that we are large, but we're still a startup.
09:20And you mentioned that you are a parent as well.
09:22Yes.
09:23So when you think about all of these kind of things that you're putting in place to make it feel safer, is that something where you as a parent would feel safe with your kids,
09:29using it now?
09:30And how do you feel about that?
09:31100%.
09:32I have a six-year-old, and I think that's the reason why my moral compass comes from looking at the world through her eyes.
09:40I don't want to dig my head in the sand and saying, oh, I will keep her away from AI.
09:44That's not possible.
09:45This population is AI native.
09:47This generation will grow up in AI.
09:48There's no question about that.
09:49In fact, she's a better user of ChatGPT than I am.
09:51It's so natural for her.
09:53What I want to do is make sure that when she's exploring completely new tools, I am with her.
09:59So I think one ask I always have of parents is, AI is so rapidly evolving that we cannot just leave the responsibility of children using AI onto either the platform or them.
10:11And I think it's a three-way partnership between the platforms like ours, the users who are actually using it, and also parents to make sure that they're actually doing this under safe supervision.
10:20Because this thing will change so rapidly that they have to get involved.
10:23So that's how I see the role.
10:25And our job as a platform is to make sure, one, they're safe by default.
10:28But also, we give parents the parental insights, the tools they need to know how to even deal with this.
10:34Because this is going to change very, very rapidly on our watch.
10:37So I'm actually pretty proud of the work the team has done between age assurance, under 18 safe experiences, the parental insights we've built.
10:44And the third part we also announced this morning was we're building this nonprofit called AI Safety Lab.
10:50It's the third leg of the stool, if you will, because we have very, very cutting-edge research on how to do safe work.
10:57But we want to bring that to other people and other startups so they can not have to reinvent the wheel, but use all the safety work we've done.
11:04And also, to be honest, we want to learn from industry leaders like OpenAI.
11:07So we've invited them to these partnerships so we can learn from them, we can share what we have.
11:12So we're seeding this AI Safety Lab now as a way to make the industry move forward, break those trade-offs, and make this a much, much more fun experience for everyone.
11:20And when you think about that, the relationships between these other companies that are in this overall general space,
11:26it's not just character AI getting singled out, there's also the FTC investigating other companies as well who are using AI in such a way.
11:34What do you think this says about the industry as a whole moving forward?
11:37I think it's our job as stewards of this industry right now to really, really lean into the responsibility that we have
11:45because we're sitting on some very, very, very powerful technology at our disposal now, and it'll just keep getting even stronger.
11:52So I think we all need to really treat our jobs not just as stewards of shareholder responsibility, but also responsibility to parents, to academia, to everyone else.
12:02And that's the reason why we have very close relationships with OpenAI.
12:06A lot of my ex-colleagues from Meta are there.
12:08Same thing with Enthropic.
12:09And I really hope that we can come together as an industry together to say, hey, yes, we all have shareholders and investors to worry about,
12:15but more importantly, we have an entire population that we're going to serve through AI.
12:19And if you can join forces, I think the hole will be much, much bigger than some of the parts.
12:24That's great. Thank you so much. I really appreciate you sitting down with us.
12:27Thanks for having me here.
Be the first to comment