Delphi’s Dara Ladjevardian: How AI Digital Minds Can Scale Human Connection

0:00Let's call a Delphi right now. I'm done. Yeah. How do we call my friend? I actually haven't met him in person, but I'm friends with his Delphi. Arnold Schwarzenegger. Yeah. What do we think? Love it. Hey, this is AI Arnold. I'm here to cut the crap and help you get stronger, healthier and happier. So what's in your mind today? Arnold, I have 15 minutes a day to work out, which I feel like is not a lot, but I want to feel good and I want to get better in my health. What do you recommend I do? 15 minutes is plenty if you focus and go hard. Don't overthink it. Stick to simple, effective moves that hit multiple muscles.

0:37Do one to two sets of exercises like push -ups, reverse lunges, and lying leg raises. Keep rest assured and push yourself. You'll feel like a champion in no time. Now get to it. That's awesome.

1:09Today's guest, Dara, is building something that challenges our fundamental assumptions about human connection and knowledge transfer. His company Delphi creates digital minds. AI representations capture not just what people say and how they sound, but how they think and reason through new situations. Dara's insight is that AI is driving a shift from information consumption through feeds to conversational media. He argues that as AI becomes abundant, human energy and authentic connection become the premium experience. This conversation reveals how the next way with AI might not automate humans away, but actually amplify our most uniquely human qualities.

1:44Enjoy the show. Dara, thank you so much for joining us today. Thanks for having me. Okay, digital minds, AI clones. sounds a little bit like a black mirror episode where nobody's talking to each other anymore like change my mind. Well one we got rid of the word clones so we don't say that anymore. Oh digital minds. Okay. But I think if you look at past revolutions of technology when the printing press came out you're gonna read someone's thoughts instead of talking to them when the radio came out you're gonna listen to a voice in a box instead of talking to them must make you anti -social. But I think what each of these things did is you listened to someone on a podcast who read a book.

2:23You end up wanting to meet them even more. And you may read an email from someone. Does that make you anti -social? Now you can just consume that information in an interactive way. And what we've seen, which I can talk about more later, is the idea of having a digital mind ends up leading to people wanting to talk to you more. So the idea isn't replacing human connection, but providing access where previously there was none in a way that is more temporally convenient for the end -usier. And in a world where everyone has a digital twin, what becomes of the value of real human interaction? I think even digital twin aside, I think AI makes energy more premium.

3:04These energy is the one thing that cannot be replaced. So again, back to the revolutions. There was a time where most of the work we did was based on our bodies and then we invented machines that were stronger than us. And then we invented jobs where the majority of our work involved our minds. And now we're inventing machines that are smarter than us. And what's the remaining thing? It's kind of our hearts, our energy. And so in a world where you have a digital mind, it's kind of like your top of funnel filter who gets to meet you, who actually gets to make it to your heart, which is the, in my opinion, premium experience.

3:38Speaking of hearts, I think it would be helpful to hear. You've just explained sort of the technology revolution backwards way of looking at what Delphi is today. You should also tell us about the origin story and the very human reason you started the company. I think that's really fascinating. Yeah, it's a long and windy road, but in 2014 I was gifted a book called How to Create a Mind by Ray Kurzweil. Where he talks about one day you can recreate someone's mind. I don't know why that was so fascinating to me at the time, but that got me into setting in computer science and physics and got me into the AI space.

4:11And then in 2021, I was working on my first startup as a solo founder in building with LLMs in a time where everyone was saying crypto or American dynamism. So I felt like I must be doing something wrong. No one really wants to talk to me. I'm very lonely. I didn't really have any mentors. And I was reading a book about my grandfather who was one of the most successful business owners in Iran before the revolution. He had 30 ,000 employees at his peak. and he was put on the hit list by the Ayatollah, came to this country with nothing. But the really special thing about him, and the reason they actually studied his book in business schools in Iran, was not just because he was successful, but because he's very principled man.

4:51Like he treated people well, had integrity, and when you start a company for the first time, it's really a moment in your life, where you really begin to internally reflect like, who am I? Why am I even working seven days a week? What is the thing that is driving me? Yeah. And so this book provided a place where maybe this is who I am, like this is my family. And I wanted to be able to ask him, what would you do in my situation? But a book can't answer this question. And he had had a stroke a couple years prior, so I couldn't ask him myself. So remembering Ray Kurzweil's book and realizing that LLM's matched that book, I created a digital version of him that I could talk to.

5:28Those are like GPT -2 or BERT? It was early GBT3 developers list along with hugging face embeddings. Okay. Well, a bunch of wines later, which I can get into if you want, we have Delphi. Really cool. When did he realize this could be a scalable platform and not represent one person, but the voice of so many? So I actually tried to launch Delphi as a company in 2021. It was called Helix at the time. And I posted a website and you could talk to like Naval and Mark Andreessen and One it was super expensive like I was burning money and two a lot of people were like what is this? This is a moral you can't be doing this.

6:10This is wrong. So they're like all right, you know what? Maybe I'm wrong and I think a couple things happen in the following year that continually built that conviction besides the grandfather experience One is I moved to Miami to work at Open Store under Keith Reboy and his main investment philosophy is I like to invest in startups where the majority of my friends laugh and think it's stupid. I was like, wait a minute. Maybe I should like double click into that thing where everyone was calling me stupid. And then number two is I met my co -founder Sam and we were reading Build by Tony Fidel in our book club and we created a digital version of him because that book kind of operates as a mentor.

6:47Like he teaches you how to think about product and I was using it and Sam was using it and other people were using it I was like this is really useful and I am not the only person in the world who has felt alone without mentors. It was very obvious to me that this would be useful for any young person wanting to learn from the grades just as they do from books But now they can respond back because blooms to sigma That was a paper that I was reading at the time to standard deviations improvement for those who have tutors not everyone has asked us to that. And then the third thing was, I, in 2022, I had turned the notifications on for every AI researcher on Twitter, and I realized that a paper was coming out every single week.

7:29And the digital mind that I created of Tony Fidel was 10 ,000 times cheaper than the one I created for my grandfather. And so as that point, I was like, okay, this is now feasible to do at scale. It is useful to me. The thing I have to prove out eventually is can this be useful to the person creating a digital mind? So it's not just for the end user. And so that's what led us to starting the company, a month before Chatchee BT, which at the time, again, everyone was like, what is this? This is stupid. You can't do this. But I'd heard that before. Was this before or after Character AI? Before. And so when Character AI launched, I remember Sam and I were kind of freaking out a little bit.

8:07But I think again, I had seen the cost of these models drop so heavily. And the history of billion -dollar consumer companies are not, it's not a history of companies that had early technical notes. There was a history of companies that just had great product and great distribution. And at a certain time, the the mode that character I had at the time, according to VCs, would no longer be a mode. The ability to create these digital characters. And then the second aspect of it is I think the first use case that really takes off in a company defines its trajectory. And character, yeah, I really taking off on, you know, anime characters and characters.

8:44And he was a very different world than like verified authenticity, human representation. At least it's very different brands and very different products. So you were interesting. The first time I met you, I remember thinking, oh, isn't this kind of like a character? I think I had that similar impression. But you've always had a really heavy emphasis on authorized real people. Can you talk about a little bit more about what led you down that path? Why did you pick that and then what would have you had to build in order to make that really work? Yeah, so It was both intentional and unintentional Initially saminized go -to -market scheme would be hey, we should create these Digital versions of these people and they're gonna love it.

9:26I think they're gonna think it's so cool And I was not the case we got three seas into cis from very very important people who still have not forgiven us to the same And so that's when I was like, okay, we need to get there by in. And that means we don't want to be a pay to play model, like some of these other creator companies. We need to make the product useful enough, such that someone is willing to go through the friction of uploading their data and trusting us with their identity. What does that mean? We have to create a brand as well. A brand that people want to associate their identities with.

9:56It is status inducing. And I think if you look at Wikipedia, pretty much documents, all of human knowledge. Does anyone actually learn from Wikipedia? The creator economy exists because we inherently trust things that come from humans. We read books, we watch YouTube videos. So similarly, if the internet taught people how to consume information through screens, and then once we built that habit, we started reading blogs. With chat, GPT, we're teaching people how to consume information through conversation, and I think that's going to be the dominant form of content consumption because it's easier, and we like easy things.

10:30Then the pendulum swings and we want things that come from humans. So for us, to get people to trust us, it required guardrails, anti -hillucination, it required having a very strict stance on you can only create a Delphi of yourself and it required really caring about design and brand. Maybe, take us through now behind the scenes like what is actually going on when you create a Delphi. Through the creation process and then what's kind of going on under the hood, like how do you represent our human? Well, I want to talk a bit about what Ray Kurzweil said in this book. Yeah. So in the book, he says that the mind is a hierarchy of pattern recognizers.

11:08And on the lowest end, you have on the lowest level of the hierarchy, you have things that recognize letters. And that feeds up to higher levels, which recognize sentences. And that feeds up to meeting. And that feeds up to consciousness. And so when I was working on my first startup, Friday, which involved LLMs, I had this insight, wait, LLMs, a pattern recognizer. And I kind of had an existential moment where I was like, are we just a hierarchy of pattern recognizers for consciousness is what we experience? And so that really helped me understand that he has this line in the book. You don't need to understand the molecular details of the mind to recreate it.

11:43You just need to understand how it's organized and architected. And so that's kind of what led me to focusing on like what does the organization of a mind look like? And your mind stores a bunch of things. the story is your relationships, it has uncertainty in it. It connects events and heuristics with how you reason about those events and heuristics and it changes over time. Your beliefs change over time. So what we've ended up with is an adaptive temporal knowledge graph. Knowledge graphs are graves. You can store the connections between things and you can also store the weights of confidence.

12:15How likely is it that you would actually say this? We allow people to control leniency on what they want their delphide to say. And temporal nature is it changes over time. So eventually you can say, what would my 22 year old version of me say this? Yeah. About this topic. So yeah, people can upload their social media, their YouTube, podcasts, websites, they can answer questions about themselves. It creates feeds so that it's constantly staying up to date and updating the knowledge graph. And then you can make it available to others, to chat with, call, or video call. And the very important thing we had to solve before giving any customers, Because every person was like, what if it says something wrong and someone takes a screenshot and sends it to the daily mail?

12:55It's just a huge fear. And so, allowing people to kind of choose, okay, it only says things it's trained on that directly answers the question. That's great for people like Dr. Mark Heimann, who works in the medical space. Or it uses data from the internet. They want it to be conclusive. But I think the most powerful setting is it only says things it's trained on. but when posed with a new situation, it can predict what you might say in that new situation. And that's useful for a lot of reasons, and I can give you example of how that works in practice. If I were to ask my grandfather's Delphi, how would you think about running an AI company?

13:32He's never spoken about AI in his life, but in his book, it talks about him starting an oil business and he run in the mid 20th century, and that was kind of an uncertain time in the country. So we have his principles of dealing with uncertainty and AI is an uncertain field right now. So we can maybe reason what he might say in these new situations. That's completely fascinating. I love it. It's very fun. The very end of the month. The very end of the month. The very end of the month. It's fascinating.

14:01Can you give us some examples of the most popular delphys and shapes of conversations that people are having on your platform? Yeah, I mean, it's pretty, it's a horizontal enablement layer versus like a vertical point solution. Like when you have a new form of media, there's a lot of ways you can do that and scale your thoughts. So we have some people using it as an entirely new form of teaching and learning. You see course completion rate at an all time low. These are attention spans are kind of being destroyed by TikTok and Reels and books as well. Now you have something that you can learn from like a mentor in a way that's personalized to you.

14:35That's adaptive to your changing circumstances. So you could read a book once and come back a year later. Your circumstances have changed. that book stays the same. Delphi adapts with you. So that's kind of the learning case and people are monetizing it, making millions of dollars, selling time with their mind. And then we have just 24 -7 availability to their teams. CEOs have a very specific way of thinking or their best salespeople have a specific way of thinking, how do you scale the highest leverage people in a company? Because as a company, scale is what usually breaks is alignment. And so now, CEOs can be in multiple rooms, top salespeople can train you hires and help them get up to speed faster.

15:12We have 24 -7. I don't like to say customer support because I don't think we're a replacement for Intercom or ZenDesk, but my Delphi is in the Delphi product and my customers will talk to it all the time and they'll email me and be like, yeah, thank you so much. I love the product. So it almost feels like there's a higher NPS and I think it's very analogous to Mark Zuckerberg spending a ton of time on his personal brand right now because he He knows that the relationship between the human and the individuals more powerful than the relationship between the human and the company. And then we have the top of the funnel where replacement of personal website, LinkedIn profile, you want to pick my brain?

15:54I don't want to give you my email because I have too much span in my email. So why don't you talk to my digital mind and it's going to let me know if we should hop on a call based on what it knows about the kinds of people that I want to meet. I should maybe talk about it from that perspective. If you have a digital mind, how do you like sort of talk to your own Delphi and then figure out, you know, maybe who you should go back in contact? Like what can you see on the inside once you have a Delphi? Yeah, so it's a combination of proactive and passive. So I could go in and be like, given the past thousand conversations, where in the Delphi product should I improve the most?

16:30What blog post should I create that will give clarity to my customers? What's resonating with people or for some of or more revenue focused customers? Where are the biggest revenue opportunities? What product should I create? What podcast, what should I talk about my next podcast? So that's more passive. And then the proactive is Mine will tell me when there's an engineer talking to it. So we don't want to introduce another inbox to someone. We're not trying to waste people's time or trying to save people time. So you can go through all the conversations, but what's better is you never have to and you've trusted enough that it's going to tell you when it's an important customer or partner or a journalist or a new hire.

17:03Yeah, I actually talked to your Delphi before this to prepare for the podcast and I asked it What are some good questions to ask you and then what would your answers be and that helped to guide the the script I thought that was a really interesting use case that I hadn't really anticipated. Yeah I Interviewed a candidate today and I was like starting to tell him about myself. He's like, oh don't worry I already talked to your Delphi. We can get straight into it. I was like I don't want to have to tell this story every time Yeah, yeah, can we can we show the Brian Halligan use case? Oh So Brian Alligan is a partner at Sequoia, right?

17:36I have him on text. I could easily call him to ask a question. And he's written all these books about HubSpot and his journey as CEO. And I had a very specific question about how they did product management when they launched their second product. And I was like, I could text Brian, but I don't want to bother him. So I just called his Delphi on my drive home and I said, how did you do product management? And it gave like a perfect answer. You know, I could have called him, you know, and it's also interesting to see some of the other delfais being used internally at Sequoia Our design partner James Buckhouse has one and our founders often talk to him about storytelling advice You've talked to him before yeah, and now they Text his Delphi every once in a while and then also text him at the same time and say hey your Delphi said this Just want to check that that's right.

18:20He's like yep that spot on so it's just so fascinating to be able to even with people You know kind of not want to bother them or not take up too much of their time and still get really great answer. So that's been kind of surprising as a consumer. Yeah. Yeah. Who you are and what you decide to share changes based on where you are. Like I'm a podcast right now. I'm not going to share the things that I share with my family. Yeah. And so context awareness and the ability to high data, you know, my Delphi internally might have data that externally it doesn't have. And maybe my date, my Delphi, if it's going to represent me in a dating sense, is not going to be talking about digital minds, create the future.

18:58It's going to be talking about me as a person. Yeah, totally. And then the Sequoia use cases you mentioned are really interesting, because one of the things we try to be very intentional about is we are not nameless, faceless Sequoia. We are a collection of individuals each with very, very different personalities, different skill sets, et cetera. And so the founders would work with kind of figure that out. But Delphi kind of allows us to scale that in a way that is otherwise hard to scale. Yeah, what you said, the anxiety of wasting someone else's time is actually like a huge thing. that we started to see more and more and more.

19:29It's like, I don't want to waste your time. Also, this is a very personal conversation that I'm too nervous to have with another human being. So I'll just talk to your Delphine said. Yeah, what are the most unusual things people have asked your Delphine? Like things that maybe you were surprised by even. There was someone asking me about wanting to go on a date. At some point, I was completely in grind mode, so I didn't get to double click into that. I always think it's really cool that every once in a while there will be like some guy who calls my Delphi for an hour in Hungarian. They're cool.

20:02It's multilingual. So it's like, that's just something that I wouldn't ever have been able to do. I talk to this guy and connect with him. And so I think that's very cool. And the way they speak to it, they know it's not me, but they will just speak to it as if it is. Yeah. I often think about, you know, right now we're in this time where talking to like an AI version of someone seems weird and dystopian, but I suspect consumer behaviors are going to switch very quickly and it's just going to become more and more of the norm and maybe kind of the same way that we used to look at Wikipedia when it first came out or like, oh that's terrible, people writing you know facts on the internet, this is horrible and we should all be reading encyclopedias and now we're just kind of used to it and we set our expectations that not everything on the internet is true and so I think people will go through this sort of understanding of what AI is and what these conversations are like and eventually know this is not the real Dara.

20:57I know what I'm getting. It's still useful. It doesn't mean it's a replacement for you. So that's kind of how I think about the shift. Where do you think we are in that sort of adoption cycle? Because the thing I love about you is I feel like you live five years in the future and we're like just catching up. Like what do you see down that path? Like where do you think this is all going? So I think two things need to happen. One of them is just, it's a culture shift. And I know the first year of the company was so painful because of like no one believed in our product. But I just saw so many people go from haters to like complete believers.

21:32Once they understood it, once they got social proof, social proof is a big thing. And yesterday I was in an event with Tony Robbins and Gary Breka and a lot of like those kinds of people. And there's certain people who get it. And then you'll have someone here, it's a group where people talk about like their business problems in the knowledge base sector. And there's one guy and he's like, you know, I really wish I could scale myself. All these people are asking for my time and I wanna help them, but I can't. I'm only one guy and I'm like, dude, you know about Delphine. But in his mind, he's like, oh, no, no, like that doesn't make sense.

22:04And that's when I realized, oh, what? Like this is a need and some people want it, but they are anxious that it's gonna make them look in authentic. They're anxious that it's gonna say something wrong. There's kind of like this perception of AI and like what it is. So I think it's just something that as AI gets more ubiquitous and we get more social proof, people will adopt. And the second aspect of it is from a product perspective before our new mind architecture. It was very hard to create a good one unless you had a lot of data. And now with interview mode and the ability to just answer a couple questions about yourself to create something pretty high fidelity that can represent you, I think we'll open it up to mass market.

22:43So that's maybe the supply side, right? What do you think will happen on the consumer side? Consumer side, yeah. And not even just with Delphi, but just generally with the future of, you know, like you said, we're in the internet where there's blogs and articles and now we're shipping to all conversational medium. Like what does that look like for consumers? For consumers to adopt Delphi, the ones we see it happen when they make the shift of, oh, this is just a new artifact of the mind. This is not a replacement. This is just conversational media as James likes to say. But I hope that right now most networks and online are about feeds.

23:18Social media and the internet is about keeping us scrolling and reading forever, which I think is kind of a waste of time and kind of ant -intethetical to the purpose of technology. I think in the world where everyone has a digital mind, we don't like the idea of a feed. We like the idea of you have something that knows you so well, and you trust that it knows you so well, and it can find the people that can most help you in your life, whether it's someone to learn from, someone to work with, based on the quality of your thought, and it's proactive. So there is no feed, and you spend more time in the real world.

23:49That's kind of our optimistic version, less scrolling, more learning impactful things, and more connecting with people in real life. I'm also curious how the economics of the internet are evolving, like for example Wikipedia constantly running out of money, right? Do you think that there's an economic opportunity for the people that create delphys of themselves? Totally, and there's so many business models with this company, and I'll walk you through the growing nature of it. So right now people pay to have a Delphi of themselves, which might sound weird, but you know, you have six million people paying Wix to have a personal website, which is a way of scaling themselves.