Subscribe
Share
Share
Embed
Ahmad Al-Dahle has been at the center of some of the biggest technology and societal shifts of the past two decades. He helped develop the early iPhone at Apple, spearheaded the open-source Llama models at Meta to help decentralize AI power and influence, and now serves as CTO of Airbnb, where he's using AI to help people spend more time in the real world.
In this episode of The Deep View Conversations, we sat down with Ahmad to discuss Airbnb's vision for AI, why he believes AGI is coming, how AI agents are revolutionizing executive workflows, and why technology should ultimately help people spend less time with technology and more time having human experiences.
Topics covered include:
+ Why Ahmad left Apple after over 15 years
+ How Airbnb is using AI to reimagine travel and drive more meaningful connections
+ The future of open-source AI and the impact of Llama
+ Why Ahmad believes AGI is inevitable and what comes next
+ How AI agents are transforming leadership and productivity
+ Lessons from working with Steve Jobs, Mark Zuckerberg, and Brian Chesky
If you're a leader, AI builder, entrepreneur, investor, or anyone trying to understand where AI is headed beyond the hype, you don't want to miss this episode. Ahmad's insights on AGI, open models, AI agents, and the future of human connection make this one of the most thoughtful discussions we've had on the show.
Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology.
And don't forget to sign up for The Deep View daily newsletter. We don’t just cover AI, we decode it. In a world flooded with hype, we deliver sharp, no-nonsense insights to keep you ahead of the curve and help you put AI to work every day: subscribe.thedeepview.com
Creators and Guests
What is The Deep View: Conversations?
From frontier labs and enterprise platforms to emerging startups reshaping entire industries, The Deep View: Conversations podcast interviews the brightest minds and the most influential leaders in AI.
Jason Hiner: In this episode, I talk to Ahmad Al-Dahle, CTO of Airbnb. Ahmad has been at the center of some of the biggest technology and societal shifts of the past two decades. He helped develop the first iPhone at Apple, spearheaded the open-source Llama models at Meta to help decentralize power and influence in AI. And now at Airbnb, he's using AI to help people spend less time on their devices and more time having human experiences in the real world. Ahmad and I talked about his views on what humans will do after AGI arrives, why he left Apple after over 15 years with the company, and what it's been like learning from leaders like Steve Jobs and Brian Chesky. This is one of the most thoughtful conversations we've had on the show. So here it is, our conversation with Ahmad Al-Dahle of Airbnb. Alright, so you've had this amazing journey with Apple and Meta and working on things like machine learning and autonomous systems and open models, things that are foundational to this AI moment that we're in. So why Airbnb and what are you working on now?
Ahmad Al-Dahle: I've been very, very lucky to have worked on a lot of technologies that have become part of how we live our daily life. And when I think about, I got to a moment where I was really thinking a lot about what's next. I just had twin daughters, they're three years old.
Jason Hiner: Congrats.
Ahmad Al-Dahle: Thank you. And I think about their future and what has become of human connection, what has become of, you know, what helped form me as a person. And I think most of who I became has come from traveling, has come from me experiencing the real world, seeing cultures, seeing how people live. That richness, that diversity has really informed how I think about almost everything I do. And when I think about companies in the valley that are really trying to work on that problem, that are trying to build technologies to get people out of their phones and into the real world, you know, if I take a picture of a family eating at dinner in the 1980s and the 1990s, it would look very different than 2020. Now I don't think it's realistic to go back to those days, but I do remember spending a lot of time in the neighborhood, hanging out with people and finding things to do, meeting people, building my social skills, and usually that came in the form of either travel or experiencing things in the real world. So I thirst for a company that's on that mission, and I think it's more important and relevant than ever as I look at the next frontier and what it will do to society.
Jason Hiner: So for somebody that spent all their time working on sort of building these systems, you know, very deeply technical systems, that led you to wanting to work at a company where you're working more on like letting people build connections and getting away from technology a little bit and spending more time in the real world.
Ahmad Al-Dahle: Yeah, I think the world needs it. I think using state-of-the-art frontier technologies to make it happen is the right thing for someone like me to do. But I think, you know, aside from that, Airbnb is also quite an interesting company. And Brian specifically is also a really interesting person for someone like me to work with. You know, I think the loss art, you know, I'm again very lucky to have worked with, at Apple in the early 2000s, pre-iPhone, and having seen the kind of culture and experimentation that gave us the iPhone, it was a very product-centric experiential company. We were always thinking about, you know, how would people use it, how would people feel about using it. Design and human interaction was at the core of what we were trying to do, and the company was really focused on its sole, on its values. I see a lot of that in Airbnb. So it's not just about the mission, I think it's also about the culture that will give bear to a product that I think could actually, you know, bring people out of their phones and into the real world.
Jason Hiner: You mentioned that there were some experiences, you had travel experiences that were really fundamental to that. I'd love to double-click on that for a second. Like, what are the ones that you had that are most memorable to you that were most sort of formative, you know, now that you're at a travel company?
Ahmad Al-Dahle: Oh boy, I had a lot. Let's see, I think probably one of the trips that changed me the most was a, right after college, a friend, a group of us, almost like 30 of us, went, yeah, it was a big group of us. I went to University of Waterloo, which was like a very intense engineering school, so a bunch of us, when we graduated, decided we needed a break. In Canada, it's quite cold.
Jason Hiner: You go somewhere warmer?
Ahmad Al-Dahle: We did, we went across Europe, so I went to, I don't remember the exact number of countries, but I think we went to 15 countries.
Jason Hiner: 30 of you going to 15 countries?
Ahmad Al-Dahle: Yes. And I'm still friends with everybody who went on that. I didn't know everybody very well, I mean they were friends and friends, and we became friends on the trip, but I still keep in touch with a bunch of them, and it was just an incredible experience.
Jason Hiner: Wow, how long was it?
Ahmad Al-Dahle: 20 plus days. I don't remember the exact number, somewhere between 20 and 30.
Jason Hiner: That's impressive. Okay, any other ones? That was an early one, you know, any ones more recently that you've had that have been big?
Ahmad Al-Dahle: Yeah, my wife and I, after twins, three years, it was quite an intense experience having two children at once. It was very efficient. It's a very efficient way to build a family, but it is also quite taxing. My wife and I took several weeks traveling through Switzerland and Italy and France, the south of France. It's like Europe. We love Europe. Yes, we love Europe. I have been to all parts of China, and I have been to India, I have been to Russia, I have been to all kinds of countries, so I've done a lot of traveling. That's why it's a really hard to pick one. I've been to Japan a bunch. I love Japan. Japan's probably my favorite country to visit.
Jason Hiner: Wow. Now I see why you wanted to go to work for Airbnb. This is not a travel podcast, but I'm sure a lot of people do.
Ahmad Al-Dahle: I got married at Airbnb.
Jason Hiner: You got married in an Airbnb?
Ahmad Al-Dahle: I did, yes.
Jason Hiner: Wow, you are the perfect CTO for Airbnb. Tell me about that. How did you get married in an Airbnb?
Ahmad Al-Dahle: It was COVID.
Jason Hiner: Okay.
Ahmad Al-Dahle: And during COVID, both my father and her father's been a tragic thing. Both had cancer. And so COVID, they're also quite vulnerable.
Jason Hiner: Of course.
Ahmad Al-Dahle: We brought our families together, very, very small event. We couldn't find events, but Airbnb was there for us.
Jason Hiner: Wow.
Ahmad Al-Dahle: But what I don't know, and maybe I shouldn't tell you and I shouldn't admit to, is whether I violated someone's house rules, but they were very nice about it. And so we got married in an Airbnb, and it was an incredible, and the host was an incredible human. They came in and helped us set up for the wedding.
Jason Hiner: Wow.
Ahmad Al-Dahle: We did it, not very loud, but it was a beautiful home.
Jason Hiner: Okay.
Ahmad Al-Dahle: We got photographers. It was magical.
Jason Hiner: That's pretty incredible.
Ahmad Al-Dahle: But I've been fond of the product, and we stay at Airbnb's every year. Not just my wife and I, but my entire extended family comes together every Christmas. Okay. And we stay in Airbnb. So I have been an Airbnb user for over a decade.
Jason Hiner: Wow. You are the perfect person for this job. Can I ask where the one was that you had your wedding?
Ahmad Al-Dahle: It was in San Jose.
Jason Hiner: In San Jose. Okay. Okay. Yeah. Makes sense. So what are you working on now at Airbnb? You have this long history, which we'll get into in technology. Coming here, obviously it is a technology company. It always has been. What are the kinds of things that you're spending your time on?
Ahmad Al-Dahle: I mean, AI almost permeates the entire platform. It was really interesting because coming into Airbnb, I was, I was really interested in the idea. I was mostly focused, and it's obvious that the most complicated technical area is going to be what you see. It's going to be the user interface and sort of what is above the surface. But if you look below the surface, there's quite a deep technical stack here.
Jason Hiner: Okay.
Ahmad Al-Dahle: So I'm fortunate to work with some incredible engineers across almost every layer of the softer stacks. So we have an infrastructure team that is quite deep. The infrastructure at Airbnb is actually as sophisticated as some of the, some of the large, much larger companies that I've worked at. There's a platform that's quite complicated because it's a marketplace, meaning you have supply and demand, meaning to balance the equities of both. You also need to build product on both. So while we thank you and I, I don't know if you're a host, but you're probably a guest, you look at Airbnb as like the booking engine. You go in, you search.
Jason Hiner: Yeah.
Ahmad Al-Dahle: And of course we're working on that. Of course there's AI features there that can make your search experience better, your discovery experience better, help you discover what to do. And it's always important to remember Airbnb is not just providing homes, but you know, we just launched and talked about our experiences and services, which I'm very bullish and excited about. But that's the guest experience. On the other flip side is you have hosts. We have 5.5 million hosts that are depending on us for their earnings, for their income.
Jason Hiner: Sure.
Ahmad Al-Dahle: And it's quite difficult to be a host. You know, if you're operating this property, you're communicating with guests that are both staying there, that might come there, you know, it's a heavy engagement platform. So how do we build tools to make them more successful in our platform? How do they price? How do they position themselves in the marketplace so they can get the conversion and profit that they're looking for, but also provide a good experience? And actually it's not obvious, but you know, a large portion of our hosts only have one to two properties, and a lot of them are regular people who operate in the services industry. So these people who generally are doing this because they care. And so we have to build them tools to be successful. So, you know, as you look beyond just what the guest experience is, this is a whole other side that is depending on us to create a platform that they can thrive and be successful in. And then there is obviously in between all of that, all the glue logic that operates the entire thing, which is trust, which means, you know, for you to be successful as a company that matches strangers to stay, you know, you're trusting me to stay at my place. I'm trusting you to stay at my place. That too. That too. And then, you know, there's a certain level of trust that we need to facilitate, which is quite hard.
Jason Hiner: So the problems are quite deep, they're quite complicated. And almost every surface at Airbnb could be impacted by AI. So from, you know, what I'd call like the bottom of the funnel, which is things like customer support, is really hard use cases. People call us for real things that they need result. You know, 40% of those cases have been resolved by AI. And then you got the mid-funnel kind of work, which is trust and, you know, like how do how do users understand whether a property is relevant to them? Then you have like the top of the funnel, which would be search, you know, user comes in and expresses intent to find something. And AI is useful up and down that stack. Now it's time for a word from this week's sponsor, Deel. Have you ever loved your HR, IT or payroll software? Yeah, didn't think so. Me neither. Most platforms are stitched together. So when they add AI, there's not much it can do. Deel rebuilt the entire global workforce stack from the foundation up. One AI native platform for HR, IT and payroll. Built in-house, not white labeled. And because it's one system, AI inside Deel can actually do real work. It manages onboarding, payroll, compliance and approvals, following your policies, workflows and risk thresholds. You can hire, manage and pay anyone, anywhere without juggling tools, logins or manual handoffs. AI agents take care of the repetitive work while over 2000 local experts in 130 countries keep everything compliant from five people to 50,000 Deel scales as you grow. It just works and you'll actually like using it. We use it here at the deep view so we can confirm how good it is. See how global work runs when AI is built into the system. Book a demo at Deel.com slash deep view. That's Deel.com slash deep view. We thank Deel for their support of the deep view. And now back to the show. Let's talk a little bit about that because in the big announcement today with all of the new things that are coming to the platform and really showing that Airbnb is moving beyond this rental app to more like we're going to take the friction out of the whole travel experience. I thought that was really interesting. We can talk a little bit about some of those things. Here's what I thought about the AI things that I wanted to ask you about, which is when I go to a lot of these AI events, there's a lot of grandiose rhetoric around AI. AI is going to solve all human disease. It's going to solve for human abundance. And I actually found it refreshing that it was solving some what Airbnb talked about was solving some quality of life things. We are going to make search better. We're going to make it more customizable. So when you come and look in the app, we're going to make the summary of what's in the bullet points more tailored to what you're looking for and not just the same for everybody. Talk to me a little bit about that because what I thought is this very sort of applied AI approach. We're going to use AI to solve some little problems every day and make our experience better. That's like so different than so much of the rhetoric around AI that I see day in and day out. Yeah.
Ahmad Al-Dahle: And that's what I love about this place is exactly that question. I am just should explain this, but I do believe AGI is coming. There's a reason why I think Airbnb is particularly interesting in the world of AGI, but we can get back in one second. But I think to your question, Airbnb is fundamentally like a humanist company. So we actually start and Brian's approach starts with the user. It starts with the archetype that we're solving for. There are people out there that are explorers that come to our platform to explore the world. What does it feel like to be an explorer and how should our product be a good fit to help the explorer explore the world? And so our approach to AI hasn't been to throw away that archetype and force them into a new behavior or a new way of thinking. It is to work backwards from their experience that we want them to have and then figure out how we can leverage AI to solve the problem. So again, if you look at our launch, almost every feature is grounded in intuition on what we think is hard about travel today and how can we solve it? One billion reviews. That's how many reviews we have on different things. That's a lot. Exactly. I wouldn't expect you to read a billion reviews. Maybe you can, but I can tell you what, an AI can read a billion reviews for you. And if it knows enough about you. You mean 30 is hard, right? Exactly. One property, right? Exactly. Yeah, exactly. And by the way, the 30 are completely uncorrelated. Someone's complaining about something that you don't care about or someone's talking about something great about the property that you don't care about. So how do I even give you, even if it was 31 and you're only able to read 30, how do I give you the 30 most relevant responses? And so that's the facilitation of trust. What we haven't done is take away the human signal. So I don't want you to feel like you can't read what the billion comments are. What I want you to feel is that you have an easy way of discovering the ones that are most relevant to you. Because at the end of the day, trust is facilitated between people. And so we're using AI to help strengthen that fabric and that marketplace, not to erode it. Very good.
Jason Hiner: So how about, let's talk a little bit about a couple of the other AI features. You already mentioned the one already, which is the customer service part. And humans talking to bots and AI voices, there's some friction there. People don't love it. They're like, just get me to a human. Most people are. Some people are like that.
Ahmad Al-Dahle: Actually, you'd be surprised. OK. If you engineer it well and it's a great product experience, some people prefer getting to an action card where they can resolve the issue with themselves with the ticket. For simplistic things, like I want to change my dates. I just want a card that lets me input. I don't want to call. I don't want to wait. I don't want to talk to a human about it. It's not that complicated.
Jason Hiner: OK.
Ahmad Al-Dahle: And as long as the system is well, as long as your intent is well expressed, which normally people can and would like to change the date on my reservation or some kind of expression that AI can understand. Actually, my experience, most of it is just as complicated as a human. So we're almost at the same level of comprehension. And then actually people prefer action cards. But there is a contingent of people that just prefer speaking to a human. In those cases, we route you. OK. And the reason they care about getting to a human is they don't have trust that the AI will be able to solve the problem.
Jason Hiner: But what you're saying is, or what I read from what you announced is that 40% of people were finding that it's getting solved. So in most things in life, 40% is not great. But in customer service, it's pretty good, right? Because our expectation is it's not going to be able to solve it. So I want to talk to a human. So I'd love to. Yeah, you unpacked it a little bit there. But maybe you unpack that a little bit more for where you mentioned cards. You're just trying to create an experience where as if you just have something simple to solve, we're going to try to just make it really fast for you to knock it out, change the date. Exactly.
Ahmad Al-Dahle: And what that also allows us to do is have a higher touch experience for the 60% that require agents. So the sort of edge area use cases. So overall, the end-to-end experience is higher quality. It does feel better. And so yes, we want to be able to solve as many issues as we can with AI, but make it human and make it feel like you are able to be in control of your interaction. And when you want to human, we give you human. So I think it is, yeah, I mean, customer support issues can be quite gnarly and quite complicated.
Jason Hiner: Yeah, of course. Of course. So you're tackling AI with that. You're tackling it with the experience of finding something. So there's the way that if you're trying to figure out in the app, compare properties, it knows what you want. And so I understand this product, this part isn't ready yet. It's coming later this year, but it will show you side by side. If you're looking for a beachfront property to have for your honeymoon, you put that in there. And I'm looking at these four properties. It'll pull up these four and it'll say like, how close they are to the beach, what the price is, how well they're rated. Is it a super host, all of those kinds of things. It'll give you that in a much, much sort of simpler way. And my thing was like, one, it's going to help you find better things that are going to tailor to what you're looking for. And then two, just get to it a lot faster.
Ahmad Al-Dahle: Exactly. I mean, when was the last time you took a trip? So I was just actually looking up one for Los Angeles last week, in fact. And how many places do you look at on average? I looked at like 15, 20. Yeah, it's a lot. And you know what most people do? They put it in a spreadsheet. Wow. Yeah, no, I've seen people do this. My wife does this. Yes. And we just don't want you doing that. That is so painful. It takes forever. And if you look at the average conversion pipeline of travel, it takes days before people kind of narrow in and feel confident enough that they found the right one. And so we just want to make that easier. That's probably the highest friction area. It's not really the chat experience of where to go. It's actually figuring out which one specifically to get. And so tackling that problem is actually tackling what I think is one of the highest friction experiences for someone who wants to go out and explore.
Jason Hiner: OK. Very good. Now, one of the other things that you added, and this isn't as AI, but it's really interesting from a standpoint of this as a platform that's trying to solve this bigger travel problem. Because travel can get really hectic and stressful. And there are a lot of parts, especially if then if you introduce international travel, which I know most people in the US, it's interesting. I wasn't aware of this, but the percentage of people that travel internationally is lower, far less than 50%. But if you do do that, then it gets really hectic. Right. Because then you're especially trying to figure out how do I get a ride? How do I find the right one currency? Or how do I store my bags if I'm in an Airbnb? So I was amazed and really curious about the fact that Airbnb talked about today about expanding to solve some of more of those things. Like we'll help you find a hotel in some cases. We'll help you book a ride. And they'll pick you up at the airport so you don't have to try to navigate all of the systems at the airport in a language you don't understand. So maybe talk a little bit about that. Because that was to me like really kind of broadening the lens of how Airbnb sees itself as not just a place that's going to help you find a place to rent. But we're going to help you make the whole travel experience a lot smoother because it can get pretty rough, especially if you're not unfamiliar with a place.
Ahmad Al-Dahle: Exactly. Yeah. I mean fundamentally to me Airbnb has always stood for the ability to belong anywhere.
Jason Hiner: Okay.
Ahmad Al-Dahle: Now that because you know you're booking homes and stays and that's kind of the genesis and the origin of Airbnb which means I go to Italy, I don't stay in a hotel per se. In Italy I stay in someone's home and I'm living like a local. I feel like I belong in Italy. I'm actually getting the real Italian experience so I'm an explorer in that context. And that's the archetype that we tend to think about. The explorer.
Jason Hiner: Yeah. Okay.
Ahmad Al-Dahle: So if you think about that everybody's on a different journey in terms of what it means to be that. Some people prefer hotels but are willing to try experiences and are willing to try services. And so and as you stay and try more things locally you might actually feel more comfortable staying in more local accommodations as well. So we want to be able to catch people in the different sort of phases of where they're comfortable being explorers and simplify the entire experience. Everything from how you go from the airport to your accommodation, how you experience the location. And all of these are really about expanding people's horizons and making them much more of an experiential.
Jason Hiner: Very cool. I love talking with you about all the travel stuff but I should ask some AI questions or else the audience is going to vote me off the island. So I've heard that you had this kind of amazing journey at Apple straight out of Waterloo University, University of Waterloo. To Apple working on the early iPhone. Yeah. So how did that happen? How did your journey in Texas?
Ahmad Al-Dahle: I would say it even started before that. I grew up in a place called Winnipeg.
Jason Hiner: Okay.
Ahmad Al-Dahle: Which is in the middle of Canada. It's known for being the truck stop capital of Canada. But I know I made it sound too small. It has like lots of people who live there but it really is very, very cold.
Jason Hiner: Okay.
Ahmad Al-Dahle: Extremely cold. In fact, it's nickname, endearing nickname is Winterpeg. So that's where I grew up.
Jason Hiner: Perfect.
Ahmad Al-Dahle: Well, turns out there's not a lot of stuff to do except hockey and computers and I'm a terrible hockey player. So I became a computer person.
Jason Hiner: Okay.
Ahmad Al-Dahle: And I loved video games. I was hoping to become a video game programmer when I would grow up and I would build games. I like to like disassemble games. I just loved games. And so unsurprisingly, that obsession led me to Waterloo where I did my education engineering. And then after I finished, I wanted to create a company. Actually, I was entrepreneurial.
Jason Hiner: Okay.
Ahmad Al-Dahle: Young and foolish maybe. I was like, okay, I'm going to move to California and take a job. And I took a job at Apple. And I got to work on some very early prototyping of what would eventually become the iPhone. And it was probably the most magical and stressful time of my entire life. But I got to work on very low level software. I got to work with hardware teams. I got exposed to what it takes to build consumer hardware. I probably one of the best consumer hardware companies in the world. Under probably what is the most intense leader at the time in maybe the Silicon Valley. So it was a great way to build some grit, some character, some focus. And surely enough, we built the iPhone 2007. I decided not to do a company because it was way too fun.
Jason Hiner: You were having too much fun working on prototypes.
Ahmad Al-Dahle: It was pretty incredible. We created all kinds of stuff. But ultimately, all of it got canceled and the main focus was on the phone. I think it was in some of the biographies that...
Jason Hiner: Started building the tablet first. Exactly.
Ahmad Al-Dahle: So I worked on the tablet and it got canceled and moved to the phone.
Jason Hiner: Gotcha.
Ahmad Al-Dahle: And it was like the formative years of my life. It was an incredible experience. Around 2000... And then I had done a bunch of different technology groups at Apple. I did imaging, displays, multi-touch, all kinds of systems. And then I moved... I got quite enamored. If you were doing the kind of work that I was doing, it was core technology work. In 2012, AlexNet came out. Changed the world. We got a big jump on this competition called ImageNet, which was a classification challenge. And I was like, wow, the world's changed. Everything is solved. Computer vision is solved. What's the next... Okay, robotics. It's going to happen. Let's go. So I considered moving and leaving Apple to go to Tesla and other companies. And then I got pulled into a secret project at Apple to work on the autonomous systems group. I didn't lead the whole thing. I led the AI machine learning portion of it.
Jason Hiner: Okay.
Ahmad Al-Dahle: And it was an incredible experience. I thought that. But I quickly learned that we were way ahead of our skis. The technology.
Jason Hiner: Gotcha.
Ahmad Al-Dahle: I was not ready for mass scaling. I mean, this was 2013.
Jason Hiner: Sure.
Ahmad Al-Dahle: From 2013 to about 20, I don't remember, 17 or so. I had been working on building these systems. And we got quite far and did some great work. And then realizing, again, while I am a technologist, I'm also a product builder. I like to build things and put them out in the world and see how people use them and learn.
Jason Hiner: Sure.
Ahmad Al-Dahle: I decided to move on. I worked on some other special projects that haven't shipped. And then I left Apple to join Meta in 2020.
Jason Hiner: What led to that? Why did you leave to go work on AI, essentially, and machine learning at Meta? Yeah. I felt there was a different approach. I felt I wanted to work at, having now spent 20 years. That was when I just, or 16 years, sorry. At Apple, I felt two things. One, I needed to challenge myself. Sure. And I needed to challenge myself in a totally different environment that forced me out of my distribution. I kind of felt like I had a local minima. Sure. My friends were all from Apple. They still are. I mean, they're still my very close friends. But literally, my entire network was Apple. And I had not interacted with the outside world. So I was very focused on Apple to find who I am. Sure. So I wanted to test my mettle. That was kind of the primary motivation. And then second, I think, wanted to see what it was like to work with more sophisticated infrastructure. It's cloud-based company, big data. Yeah. So obviously, the fuel for AI. But in 2020, that was less obvious. People were able to train these neural nets on workstations. And it wasn't as skilled as it is today. So it turned out to be a very smart, serendipitous decision. But yeah, those were the two motivations. I wanted to work with a company with that sophisticated infrastructure and data practices and also quite aggressive on the technology frontier. How about you went and worked on open models? I did. Lama, the Llama models and that. Tell me about that. And even would love your perspective too on the place of open models versus proprietary models. This is still something we talk about right about a lot. And there's just a really maturing perspective on the whole place of open models in the ecosystem.
Ahmad Al-Dahle: Yeah. I mean, when we put out the very first Llama model, yeah, the conversation around open source has really moved a lot. I can tell you when we first were looking at doing it, people were generating concerns that just were completely unfounded. But I think there is a balance of responsibility and empowerment that is part of the tension of what to do. But I was a very strong believer that at the current state of the technology, centralization of AI was actually a net negative for society than a true democratization of the technology. So I was a very strong proponent to create the open source movement and I was happy I did it. And it did create a way that kind of helped facilitate the future for where we are today, which I think has a really rich, vibrant open source ecosystem, not just the model layer, but orchestration layer is almost every layer of the stack now. Open source is kind of the predominant way that you do things.
Jason Hiner: Yeah. So you mentioned already AGI versus applied AI. Sounds like you have a perspective on this, of like AGI. AGI is this goal that we're going to create an AI that will be generally better at almost everything than humans are. I'm trying to define it because when you define it, you almost have to define it. Some people are already here. Exactly. It's a marketing term. So I'm going to define it as the AI that is smarter and better at intelligence than humans at most things. Yeah, every task. Okay, at every task. You feel like there's a path to that. Yeah, okay, we're going to get there.
Ahmad Al-Dahle: The more interesting question in my mind is what is the consequence? How does the world evolve? What is the consequence of society? To having that multiple cultures. If I told you there's going to be a computer in your pocket in 2003, that would have dominated the conversation. I would try to imagine what it is, what it means, what it can do, what it can't do. But almost no one would have had the conversation that every human being on Earth would be stuck to their phone for eight hours a day. And so that is, having done this before, having seen what technology can do, I do think that's the more interesting question. What happens when you have an AI that is as capable as any human, any expert in their field? What happens?
Jason Hiner: We could do a podcast just on that. Yeah. Just on that. We'd love to do that at some point. Okay, but knowing that we're short on time, I would love to talk more about that. Let's shift a little bit to a couple things. Because you're a leader, you're here at an organization that is known for having a very influential leader. You talked a little bit about Brian, known for being very detail-oriented, very visionary, and very sort of human-centric. Have you felt all of those things since you got here?
Ahmad Al-Dahle: Yeah. Yeah. I think Brian works backwards from what people want and what people need. I think he is very meticulous in the details. He does really care about the products he builds. There's a sense of pride in that work. It's hard to understate how valuable it is to work with people who really want to be proud of their craft. That permeates the company. I've worked at a company that has 16 years of my life. It's a company known for its very strict guidelines on how they ship and the quality that they ship. To me, this is very natural. There's amazing things I've learned in all the different companies I've worked at. From that, I learned how to move very quickly. From Apple, I learned how to obsess over the craft. There are elements here both. Everyone is a totally different company, but Brian drives with founder energy, with give-a-dam culture that I personally love.
Jason Hiner: Yeah. Very good. All right. How about as a leader, in this AI boom, one of the things that is almost counter-intuitive is everybody always thought that when we got to AI, it would automate some things for us. There would be maybe less work to do because the AI would take care of some tasks. But in fact, it's made us move faster in many ways. Everybody that I talk to is even more precious with their time because they feel like I've got to be... It could be like commanding my agents right now. Exactly. As a leader, what's your best tip for others, for leaders on how to spend your time? How to get the maximum leverage for your time?
Ahmad Al-Dahle: Yeah. This is a really good question. There's two things that are true. I think your first point to that people feel more pressured for time is true. I actually find more...it's more intense. My time is more peaky. So when I'm at my computer and I'm using AI to get things done, I am firing five or six or seven or even I don't even know how many parallel tasks. And so it's quite exhausting because the AI comes back. And this is kind of why I want to work with Brian on the future of interfaces and what it means to work with AI. Yeah. So at the end of the day, it's exhausting because I'm firing off a command to use a bunch of tools. It goes, it does this work for a couple of hours. It comes back, it spits out a bunch of text. And so to spool up on what I just did, 10, 15 minutes, it's exhausting because you're compressing hours of work into minutes. And I'm doing across eight or 10 tasks. So you need to protect your time, but you also need to realize it's almost like you're doing hit training, high intensity interval training. Versus like a steady run on a treadmill. And just know that going in. But actually what I like to do is spike break. Spike break. So that I can re gather my thoughts. So I'm not just constantly firing 10 or 15 agents all at once. I take a minute to breathe think because your only as good, your AI is only as good as your prompt for now.
Jason Hiner: Sure. It's a great tip.
Ahmad Al-Dahle: Yeah. And then I think the other one is I start with an AI first perspective actually start from Claude.
Jason Hiner: Okay.
Ahmad Al-Dahle: Or codex depending on what people prefer. And then I start using the other application. So I'm actually hardly ever using Gmail, hardly ever using Slack, hardly ever using anything. I start with my tool, your preferred AI tool. We even have our own harness that we call Air Chat or BNB that we engineered and it's multi models. So like generates with one and the other is the critics. So we're like leveraging all models to help improve the quality of our outputs. But but yeah, I basically start from my tool and then use the bespoke thing. So I'll create my presentations there, my documents, my emails, my responses. I've got a context graph with memory in it that like I do recall from. So I basically I have a counterpart CTO agent. I've also modeled all my peers. I review all my communication to make sure that I'm thinking about their perspectives when I'm writing things. So I have sub agents that are basically every other ET member. Yeah. So it's pretty.
Jason Hiner: That's fascinating.
Ahmad Al-Dahle: Yeah. And I treat AI kind of like a coworker.
Jason Hiner: Okay.
Ahmad Al-Dahle: I mean, you know, I treat it like a coworker.
Jason Hiner: Yeah. Not nice to it. Because you never know. I'd be shocked if you were not nice to it. And it's terrible stereotype. It's true. It's true. It's too cold for us to be mean. We'd all die if we were mean.
Ahmad Al-Dahle: But I think, yeah, I treat it like a coworker and I kind of think about agents and sub agents as coworkers and how they work together. And how do I get things done with them?
Jason Hiner: So you've kind of answered this, but what's the AI? I always ask everybody, what's the AI tool that you're using that's really making a big impact for you right now? And so it could be, you mentioned the way you're doing it, which I love. And I'd love to hear about how you start more, you know, I'd love to hear more about how you start with those tools first, Codex or Claude. But maybe is there another tool sort of besides those that you've sort of discovered recently or started using that's, you know, also have an impact?
Ahmad Al-Dahle: I mean, I'm a big fan. I test all the models. So I use all the models. I really love the image generation models. They're getting quite good.
Jason Hiner: Okay.
Ahmad Al-Dahle: Especially if you're trying to look up the new UIs or new concepts, it's getting really, really impressive. And starting from a single-shotted chat image that you've been able to generate to put into a coding agent is like saving me a bunch of descriptive language that I need to create for what I'm trying to prototype.
Jason Hiner: Okay.
Ahmad Al-Dahle: So I find that to be, it's gotten crazy good. Like the level of what I call prompted here and how close is your prompt to the image that's generated is getting really nuanced. Like you can really, you can write a paragraph description of exactly what you want. You can even condition on an image. You can take a screenshot of Airbnb and say like, okay, change this like one detail and it's quite good. So I've been pretty impressed with how far the image generation systems have become and how quick that becomes to prototype and build the concepts that are just in your head.
Jason Hiner: So you use that as a way to like, if you're telling your team, okay, let's fix this thing here. I just show them. You just, yeah, and then you can create it. Yeah. Amazing. Yeah. And even Brian, by the way, we work together this way now. Okay. So he actually sends me screenshots that he's been able to edit. And he's quite proficient at AI tools. Nice. Very cool.
Ahmad Al-Dahle: And then how do you use it? It's kind of interesting because I get asked this question like all the time. Family members ask me this. I went from fixing people's iTunes accounts when I returned Apple to how do I use AI? And honestly, the simplest answer is talk to it. You can literally ask it, how do I build agents and sub agents? I work here. I am trying to do this. I spend most of my time doing X, Y, and Z. The most important part of your tool is the connectors. So make sure that you can actually connect the AI to context. And it needs to connect to your email. It needs to connect to Slack. It needs to connect to all of the things.
Jason Hiner: Yeah.
Ahmad Al-Dahle: I still think we're figuring out how to deal with the security provisions, making sure the thing doesn't accidentally send someone something it shouldn't.
Jason Hiner: Yeah.
Ahmad Al-Dahle: But I think it will get there. It's just a, it's an engineering problem.
Jason Hiner: Okay.
Ahmad Al-Dahle: And I think the future is pretty bright from everyone's ability to buy time.
Jason Hiner: Speaking of time, thank you for your, this is a lot of fun. Yeah. And good luck in the new gig.
Ahmad Al-Dahle: Thank you.
Jason Hiner: Yeah. It's great to hang out. Very good.