Brian Chesky: the chatbot is the MS-DOS of AI — Airbnb is designing for the next paradigm

Oh my god.

Absolutely. Insane.

Scorcher. Well, uh, we have Brian Shasky, the co-founder and CEO of Airbnb in the Reream waiting room. We're bringing him in the TV show. Brian, great to see you again. How are you doing?

Hey guys, good to see you. Thank you for having me back on.

Of course. You're welcome anytime. It's great to be here with you. Uh first kick us off with the news. You've been acquiring talent, building the roster. You have a new CTO. How did it come about? Tell me about this person.

Yeah, so like you know obviously a couple years ago um I started asking around like who are some of the best leaders in AI and I started meeting people and one of the names that came up over and over was this guy named Ahmed Ald. And he was head of generative AI at Meta. We met and I I just was completely blown away by him. Um, you know, not only was he at Meta and he essentially created uh led the team that created the llama models have been downloaded a billion times. There's been over 60,000 derivative models. Um, but he was also at Apple. He actually joined Apple out of college, worked on the original iPhone in 2005, led a lot of the multi-touch technology. He um led the autonomous uh technology group that was the self-driving car project at Apple. So, he was somebody at the frontier. He really understood both Frontier technology but he also understood that like kind of Apple design craft which I think Airbnb also was really passionate about and we just really hit it off became friends and as my current CTO we were talking about handing the baton to somebody we thought it would be really great.

Yeah. How are you thinking about design in the era of generative AI these apps are amazing and the fun and the models like you see the benchmarks you you use the models and it's it's magical. It's like you know we created fire. Uh, and then one rough edge, one button that doesn't quite work the way you expected it. I was lamenting the fact that I was using clawed and I couldn't get a deep research report to read it out loud to me. That feature is in one app and there's different features there. How do you

Yeah, it feels it feels very notable that even in all the leading LLMs, there's still so many rough edges in the product.

Part of that is like a new new product paradigm. We're kind of figuring out what the the edges are. All right.

Yeah, maybe there's all sorts of metaphors. You know, I I remember when I was a kid,

um I had I I think I got like a compact computer and I was running MS DOSs and we're kind of more in the MS DOSs phase of AI. Um you know, I don't I I think I not to belittle the interfaces, but the chatbot is not the endstate interface. The current chatbot is not the end state interface for AI. I mean, think about your phone in your app. Um do you want to chat to get the weather? or do you want a chat to like do most of the functions? The answer is no. And so I think the the the future AI interface will have a reference to chat, but it's probably going to be much more visual. Most people aren't completely text- based. And I I do think there's been a it's kind of like we've been iterating the engine of the car, but we haven't really redesigned the car. And so the chatbot is really a pre-AI interface that we've plugged onto this incredibly powerful engine. And so I think what we're trying to do is think through like what are where can we take this interface? What is the like multi-touch reference for an age of AI? And we're not going to do this overnight. We're not going to be the only company doing this. But I really think like the models are even more powerful than people realize because we have hardly harnessed them. And if you look at where they're going to be in a year or two, we really want to like skate to where the puck is going and do something really really cool. So I totally agree with you. And I think the other thing I'd say is I'm on the board of Y Cominator and I think the last I checked 87% of the companies are enterprise companies per batch.

People aren't

I know we're we we go and we meet with every batch and uh to have them on the show and and we're always trying

hey every we want every consumer company to come on because often times Yeah. often times they give you

and and the interesting thing is all of them all these founders started doing consumer businesses. You asked them they're doing Minecraft servers and flipping sneakers and stuff. So they actually have some consumer

DNA

but they end up really

So you're you're bringing up a really good point. I joined Ycomir in 2009 and that was not like a two years after the iPhone was introduced and almost every startup was a consumer company and enterprise is awesome and I think it makes complete sense why AI is starting with enterprise cuz the models are not cheap to run. You can basically vertically integrate any business and completely automate the workflow. It's going to revolutionize how we do work. But the point is that the biggest prize is consumer because that is what's going to reach daily life for billions of people. And I think a lot of people are a little bit a lot of entrepreneurs are afraid to go into consumer. That's the biggest thing we've heard. They're afraid. Um partly they're worried that like OpenAI or Google are going to like essentially eat their lunch and vertically integrate. Um the other concern is consumers just hard. You know, it's a little bit more of a hitch business. Distribution's a little more mature because of the app store. There's not really new distribution channels in the same way. At the same time, we are in the midst of a revolution. And more than just a couple chat bots have to be the way that you interface with the world. And I also, this might be self- serving to say, but I also believe this independently.

I don't think we're going to use just one agent.

I don't think we don't have just one person in our life. I don't think we're going to have just one agent in our life. I don't think we'll use thousands, but I think there's going to be some level specialization that's going to help. And we like there's going to be too much specialization where I wouldn't download that agent or use that agent. But I I think even if you want a meta orchestrator whether it's Claude or Gemini or Chetch BT specialization is going to help even within consumer. So we want to do some really cool stuff within traveling and living.

Yeah. What what advice do you give to the 13% of Y cominator fan founders that are going after uh consumer at least for now or at least in at this moment is it hey keep your burn really low run a ton of experiments stick it out because it feels like part of the reason why 87% wind up in enterprise is because businesses are just throwing money at them they're saying yeah you're doing consumer but if you come solve this one really small problem for me and build this API for me I'll give you a contract. I'll be your first customer. It's very easy to get the money flowing on the business side potentially. Uh but what how can you actually make it as a consumer founder when there is a big prize but it might be a harder road.

Yeah. So many thoughts here. I mean one of the things that Paul Graham taught us and Gary Tan teaches which I think is very effective is use other Y cominator companies as your early adopters. Hence why a lot of companies go into enterprise. If you have an artist company don't go to like Boeing to try to use your service. go to a young startup and work your way up to the large companies. And so they've developed this incredible ecosystem where thousands of companies are adopting each other's tools and they bootstrap each other. And that's actually been part of the explanation for why Y Cominator has so successfully created so many enterprise companies. But when I was in Y Cominator, we got everyone to use Airbnb. And so in other words, like we have a huge alumni network. I would tell people that are going into consumer, number one, don't be afraid to do AI consumer. It might fail, but like investors aren't going to hold it against you that a startup failed. They're gonna like a a founder who's failed is more fundable the next time than someone who's never started unless they really blew it. Yeah.

Because you have experience and I think that's what's so great about this. Number two, you got a lot of early adopters within Y Comer who are normal people too. They have daily life. I would agree to keep your burn rate low. Um I would pick a niche. I wouldn't pick a niche that's too narrow. So like maybe like AI to do something incredibly narrow is going to get really really difficult for somebody to want to like change their workflow and behavior. But just think about all the parts of daily life that are kind of annoying. And maybe pay attention to like your sister, your significant other, your children, like whoever's in your life and just look at their life and ask like how could their daily life be a little bit easier? And I would just go for it. I mean and by the way like

the app store is mature. That's a little bit of an issue. At the same time, the top apps in the app store are AI apps. So, if something is truly great and truly breakthrough, people will seek it out. And I do think my prediction is I is that AI is going to revolutionize consumer applications in the coming years. And they might not be apps, they might become agents. Agents might be the new apps. We have to figure out what that means. What's the new platform paradigm? It's probably agents, not apps. But we'll see.

But ultimately, I think that's the big takeaway here. And so what we're going to do is we want to really innovate on the consumer experience. We want to inter integrate we want to interface on like what does consumer interfaces look like in the age of AI. Um and I think they should evolve past the current design paradigm and we'll see where it goes.

Yeah, it would be I'm not expecting to open the Airbnb app and see a blank text box purely like there's so much more that you can do. Have you thought more about uh the Airbnb app is is unique I think because it's not this uh reactive people are opening it 10 times a day. There is an intention and there's probably a lot of value in understanding uh okay this person they like to travel around the holidays. They're opening the app in October. Let's think that through. agents and LLMs can probably work through a customer's history, their preferences, just do better recommendations, what you're surfacing them. How how much have you thought about uh being proactive about reaching the customer with something that's more tailored? Uh what's been working? Where do you see it going?

Yeah, I mean so many things. One of the other problems with consumer have to keep coming back to it is there's a business pro model challenge, right? like these models are not cheap and so there's probably one of three ways you can make money in consumer. You can put ads in your chatbot, you can pay a subscription or you can do essentially transactions. Well, Airbnb luckily has the business model. We have transactions and our transactions are pretty high dollar transactions. So, we actually have the really the business model going for us. We also have a lot of data. Um, we do about 50 billion searches a year on Airbnb. That's nothing compared to Google, but 50 billion. So that's that number of times somebody's typing in something to a Airbnb search box. So we actually do have quite a bit of data about people, their preferences, and we can bring it in. And so and part of what we want to do is if the interface is a little more chat-like, again won't be per se a chatbot exactly like OpenAI, but if it's a little more of a turnbyturn chat-like connection, you're going to get a lot more information from people. And I think you know the one benefit of the current chatbots are they have memory and so you're putting a lot of context in the window. Um and I think memory by the way is something I think we're going to want to technically innovate on because the chat bots still are stateless and so you're essentially the prompt is putting in this kind of you know obviously a lot of your memory. So, I do think there's going to be an an area around how can everybody understand people's preferences, learn more about you, understand what you care about, and just be really thoughtful like what's your bucket list, who are you traveling with? One of the other things every can be really good at is the majority of trips have multiple people. Most chat bots are single player mode. They have collaborative tools, but most people are not chatting with multiple people. Most people in Airbnb are searching for an Airbnb with multiple other people. So I think another thing we really want to do is really be great at like understanding two people they have totally different preferences. How do we help them figure out the right choices for them? So a lot of collaboration tools I think going to be really important. But I think at at our heart what we want to do is really build out these robust profiles deeply understand people and give them what they want in the real world.

How are you thinking about the challenges of generative AI imagery in Airbnb listings? I can imagine there's sort of a risk of uh what is it called? catfishing where someone takes a photo. It is their house, but they said, "Hey, like patch up the walls and you know, let's let's change the lighting and make it look a little bit warmer, a little bit nicer." And there's probably a little bit of that. That's fine. You know, if you took it on a phone, you want it to look nice, but you want it to represent the actual experience. How are you thinking about the challenges with that?

Yeah, I mean like it's really interesting, right? Like I don't know. I saw on Twitter yesterday like the Stranger Things where like you can turn your face into someone else's face. And so what this really means is the like know your customer authentication. A lot of it is like you know you do a selfie and an ID or you do a video. A lot of this stuff now AI can get around. And so I mean we should remember AI stands for artificial

and so we're now starting to live in a world where if you see in a screen it might be artificial. And so I think authentication is going to be critical. And one of the best authentications is in the real world. I mean we're going to be pretty soon living in a world where you don't know something's real unless you've seen it in the real world. unless we have authentication.

One example of the way can solve this for example is we do have a network of professional photographers over 5,000 around the world. In fact, we built an ondemand professional photography network before Uber even launched where you hit a button and if a photographer shows up and takes photos your home and advantage of that is those photos are by Airbnb and so if we took them a customer would know those are authenticated photos from us. That's just an example of something that is possible. But I do think like solving this, is this real? Is this authenticated is really important. So we're gonna have to have something we're gonna have to have technology solutions for it.

Yeah.

What's your latest thinking on integrating Airbnb with various LLMs? Like how how are these conversations going into Yeah. What kind

Yeah. I'm interested. I think I think last time I was on I said they weren't ready.

Yeah. Yeah. Um and I I my view, you know, um we chose not to to launch with ChateBT. They launched with Expedia booking and the reason why wasn't like like I I was a huge fan of that project and um gave Sam some advice on it and stuff and I think it's awesome. I would love to show up on Chat GBT. I would love to show up on Gemini. I'd love to show up on Claude. I don't Here's the thing. I don't want to be the first. I want to be the best. Like we want to have the best integration, not the first integration. There are some limitations. 90% of people book an Airbnb send a message to a host ahead of time. You have to have a verified ID. There's just some things we need in integration for Airbnb to have a really good integration. We'd love that. By the way, the traffic coming from Chat GPT converts higher on Airbnb to a booking than the traffic coming from Google search. Yeah.

So, actually various like e-commerce brands have been reporting this too. Like the traffic is super high intent because people are informed.

Yeah. So I get I get this question is asked to me often as an existential thing as in are you going to partner with them or not and are they going to kill you and is there going to be no app in the future and my answer is there probably are no apps at some point in the future. They're probably agents that's there's probably some platform like paradigm. We're probably designing agents. I don't know for sure but I think that we're going to exist on those platforms but I also think specialization is going to matter. I don't think there's just going to be a few chat bots and they do everything and you are separated from every application. That's going to be true of some apps and that's going to be true if you need like maybe a commodity, but there are going to still be reasons to work with different agents or different applications. I think some specialization will make a difference. what are the big uh what is your super one sec one thing one thing I appreciate about your framing of like we don't want to be first we want to be the best is because I think uh other CEOs in travel that are not tapped into Silicon Valley that aren't on the board of Y Combinator feel this pressure to be like I have to prove that I'm like at least trying to be AI native whereas I feel like you don't have any like there's no ego around like can we build great products with AI it's simply like yeah we're going to but like I don't need to I don't need to prove it by like rushing into something.

I remember I appreciate you saying that. I remember when the metaverse launched I had uh somebody who didn't on my team that was not

metverse. You mean the the idea of the metaverse? [laughter]

Yeah. Yeah. Sorry. When Yes. When it was announced Mark Zuckerberg like what's our metaverse strategy? And I I remember like other companies like we have to have a metaverse strategy and it's like that is definitely like what someone usually would say if they're not really steeped into it. So, you know, I think it's important that you're you do something really really well, but I don't think you want to be first to signal we care about AI. That's kind of vanity. You don't need to do that. And so, when it's ready, we'll do something great and we want to add value to customers. And that's kind of all that really matters.

Uh, if you were running Apple, what would your AI strategy be?

It's a big question.

Oh, wow. [laughter] I would um I I um I would say that like the devices should be completely built from the ground up. AI native, AI first, probably not even AI first, AGI first. We should ask ourselves what is it going to look like in a few years. It's probably going to be fully agentic. These devices are going to be able to work while you sleep and you should work backwards from that. These devices and the paradigms haven't changed since the advent of AI. These are pre-AII devices running AI. So my question is what would it look like if you designed it from the ground up knowing what we know today? It would probably look pretty different.

Yeah.

So less Gen Moji, more [laughter] actual hardware,

more hardware.

We are we are in the midst of a revolution obviously. I mean I don't want to be like saying what everyone already knows.

But here's what I do want to say.

Like

the everyday life of the average person has barely changed in the last 3 years. We're on Chachi a lot. We're on Gemini. We're on Claude. People listening are much more technical than the average person. They're doing unique things, workflows. The average person's still mostly living the life they live three years ago, using the same devices. The world hasn't finally made the shift yet. It hasn't funnel made the shift. It won't make the shift until daily life has changed. And daily life doesn't change until the devices have changed, the operating systems have changed, apps become agents, and we're all living in a completely natively consumer AI world. And we're not using chat bots, at least not today's chat bots. We are still living in the DOSs era of AI and we need to move into the multi-touch era.

The chat is begging us to ask you for an arm routine. Is there a dedicated arm day? Are there any particular machines that you recommend? Free weights. What is the secret to those arms?

Oh, thank you very um [laughter] so well well first of all like genetically I have better arms than shoulders or chest. So that's just one thing. But I actually don't have an arm day. I used to do a 5day split where I do back, legs, chest, shoulders, arms. I now do um arms with shoulders. So, I'll usually do two exercises of biceps, two exercises of triceps, typically a compound freee movement and a machine movement. So maybe like a dumbbell curl followed by like a hammer strength cable with like a a V a rope rope or V bar and then um triceps always doing a push down with a cable and then maybe like a French press or skull crusher depending what we all call it.

Okay, that's a good

I like to call them skull crushers. They sound a lot more hardcore than totally. Yeah. Yeah. You know there's heavy metal press.

We don't know how to name exercises like that.

We don't.

The world's become soft. It really [laughter] has the best. I do think I do think generally if you're like under 40

for sure you should be doing almost entirely compound movements and free weights and dumbbells and barbells. You know, I'm 44, so I think transitioning to some more machines is good just because the tax it puts on your joints, but I still try to put in free weights. Are

are you using any AI in the gym? Are there are you tracking movements or or an asking question? I'm a nutrition and I have a

I have a trainer um he was a former Mr. Universe. He competed in Mr. Olympia so he I let that I let him do that. Um he is he is he is um but I my diet and my supplements are um very much like I put everything into AI and I get like I do I do things I don't know if it sounds eccentric but like I'll get blood work and they'll give you a PDF and you can put the PDF in a chatbot and it'll tell you like to analyze it and it can even tell you what supplements you need based on your blood work. It's probably something everyone should do. It's not that hard and so like I don't get I don't go out enough so I don't need more vitamin D. the blood work said that so I take a vitamin D supplement

because I'm not I'm like inside working a lot so stuff like that.

Yeah. Yeah.

Have you been surprised uh seeing the explosion of peptides?

I wouldn't have imagine a world where you know people are clamoring to do like actual

injection

injections uh and given given how popular it was. I mean, it's in the bodybuilding world, there was always like a divide between how hardcore you were, which was like, are you like natural or are you going to

are you natty or not?

Yeah, natty or not. And now nobody's natty anymore, seemingly.

No one's natty. Um I'm I'm not surprised. I I'm I'm surprised actually a little bit how much the tech community has embraced fitness. I mean, like when I came to Silicon Valley, I was like slightly self-conscious about like wearing long sleeve shirts cuz I thought if I wear a short sleeve shirt, people think I'm a mithead and therefore an idiot. And people just didn't really lift weights in like 2007, 2008.

And I I I really think that like everyone's come around probably to just the science that like nutrition and weightlifting is good. It's good for longevity. It's good for your mind. You'll live longer. You'll feel better. And honestly, weightlifting is far superior to cardio. You should do both, but weightlifting is far superior. or increases bone density. It's just you can only do one thing, you should do lift weightlifting. And um so I'm I'm really excited to see people actually embracing this. I mean, one of the things I learned about bodybuilding that I think is very much consistent with Silicon Valley is I was like, you know, I was like 120 lbs in high school. I played hockey. I remember I broke my leg and I had to do physical therapy. And um through my physical therapy, I said I'm going to like I want to learn about like my body. And there wasn't a lot of information on the internet. So, I'd like go to the Barnes & Noble and read as many books as possible. And I learned something about bodybuilding. And that was if I can change my body, I can change my life. And it was about control. And it was about this idea that you don't get in shape in one workout. It's 1% a day. And frankly, those are the lessons I learned with Silicon Valley. There's it's not about one idea. It's about grinding every single day, year after year with consistency. And overnight successes take thousands of days actually, typically. And so I think there's a lot of lessons from bodybuilding that I brought to Silicon Valley.

Yeah. No, 100%. It's a really Yeah. It's it's a really important frame of mind and we appreciate you sharing all of that with us. We know we have a hard stop, so we got to get get them out of here.

Oh, hard stop.

Yeah. But thank you so much for taking the time.

We have time for one more question.

Enjoy. I I hope to be back soon, guys. Thank you for having me here.

Awesome. Awesome. Great to see you.

We appreciate having you on the show. We'll talk to you soon, Brian. Have a good rest of your day.

Goodbye.

Wanted to ask him about uh Super Bowl ads. Oh yeah.

Specifically like how how they're thinking about I I know they've run an ad in the past that they they haven't been a annual

spender on that front, but

well they do those annual uh uh showcases sort of like the WWDC style. You wouldn't expect it from Airbnb because they're not necessarily introducing like a new iPhone, but uh Brian's been very outspoken about uh putting the whole team on an annual cadence, a rhythm of releasing all of the information, packing packaging everything up, getting everyone to, you know, rush towards one deadline. Uh it's an interesting way to run the business.

Uh we got cut off on the Sun King of Hollywood about David Ellison and Vulture. Uh I encourage you to go read it.

Yeah, you should go read it. It's a very long piece. It goes on and on and on. Uh, so many great stories.

You should go check it out in Vulture.

I'm glad that we know that the the father and son are dog fighting together.

I mean, some great some great writing from Reeves over at Vulture. Matthew, go check it out. First,