Latest / Tilly Talks Tech / We Tried Cloning Our MINDS with AI | Dara from Delphi.ai on Tilly Talks Tech

Transcript

We Tried Cloning Our MINDS with AI | Dara from Delphi.ai on Tilly Talks Tech · Tilly Talks Tech

• 0:00
  SPEAKER_00: What if you could talk to yourself?
• 0:02
  SPEAKER_00: I mean, that's something I do quite regularly.
• 0:04
  SPEAKER_00: But it's not actually you.
• 0:06
  SPEAKER_00: It's your AI clone.
• 0:08
  SPEAKER_00: Imagine this.
• 0:09
  SPEAKER_00: Someone can ask you a question and get an answer instantly without you being there.
• 0:14
  SPEAKER_00: Imagine if there was literally a version of you thinking like you, speaking like you, available on demand.
• 0:21
  SPEAKER_00: This platform lets you create an AI version of yourself.
• 0:24
  SPEAKER_00: In this episode, we're not just scratching the surface.
• 0:27
  SPEAKER_00: We're going straight into the questions that feel a little uncomfortable.
• 0:29
  SPEAKER_00: Like, can you make your AI version better than you?
• 0:32
  SPEAKER_00: Can it say things you never have?
• 0:34
  SPEAKER_00: Could it misrepresent you or even lie?
• 0:37
  SPEAKER_00: What are the limitations on this thing and where do we draw the line between you and your AI?
• 0:42
  SPEAKER_00: But also, what can we unlock with this?
• 0:44
  SPEAKER_00: Guys, there are AI clones.
• 0:46
  SPEAKER_00: I've just been talking to an AI clone of myself.
• 0:50
  SPEAKER_00: That is crazy.
• 0:51
  SPEAKER_00: Hello everybody, welcome back to Tilly Talks Tech.
• 0:53
  SPEAKER_00: My name is Tilly, I'm your host.
• 0:55
  SPEAKER_00: What we're talking about today is like the epitome of crazy stuff that I like to talk about on this podcast.
• 1:00
  SPEAKER_00: We're gonna be talking about AI clones.
• 1:03
  SPEAKER_00: Yes, you heard that right.
• 1:05
  SPEAKER_00: We are there.
• 1:06
  SPEAKER_00: Joining me is Dara, the CEO of Delphi, a brand new platform that lets you create an AI version of your own mind.
• 1:15
  SPEAKER_00: Its whole purpose is to be like a representative of your mind.
• 1:18
  SPEAKER_00: Like what?
• 1:18
  SPEAKER_00: I'm not even exaggerating when I say, like, when I tried this, when I set this up earlier, I literally felt like I was having a conversation with myself.
• 1:25
  SPEAKER_00: Without any further ado, we gotta learn all about this, where it came from, how it came to be.
• 1:30
  SPEAKER_00: So I want to invite our guest to the podcast, Dara from Delphi.
• 1:35
  SPEAKER_00: Welcome, Dara, welcome to Tilly Talks Tech.
• 1:37
  SPEAKER_00: Thank you for taking the time.
• 1:38
  SPEAKER_00: How are you feeling today?
• 1:40
  SPEAKER_01: Telly, it's great to be here.
• 1:41
  SPEAKER_01: I'm Greg.
• 1:41
  SPEAKER_01: I'm here in San Francisco.
• 1:43
  SPEAKER_01: It's hot for once.
• 1:44
  SPEAKER_00: That is good news.
• 1:45
  SPEAKER_00: That is good news.
• 1:46
  SPEAKER_00: And Dara, you've made something crazy.
• 1:48
  SPEAKER_00: Like I've just hyped you up to the nines, which you haven't even heard yet.
• 1:52
  SPEAKER_00: How do you feel about AI in general?
• 1:55
  SPEAKER_00: What's your what's your view on that?
• 1:56
  SPEAKER_01: I mean, I got into AI in 2014 when the coolest thing that was happening was uh someone created a model that could tell whether Donald Trump was angry or happy based on his tweets.
• 2:09
  SPEAKER_01: And that was really cool at the time.
• 2:10
  SPEAKER_01: So, like to just like think about how much has happened in 12 years, it almost feels like a fever dream.
• 2:16
  SPEAKER_01: Like it doesn't really feel real, it feels like we've entered a new reality.
• 2:20
  SPEAKER_01: And I think there's a lot of exciting things.
• 2:23
  SPEAKER_01: I think there's a lot of concerning things.
• 2:24
  SPEAKER_01: So I'm an optimist, but I think there's also like things that are valid to be afraid of.
• 2:31
  SPEAKER_01: Um, but I think you just gotta not be distracted by the noise.
• 2:33
  SPEAKER_01: So I like try to mute as as much AI content from my Twitter feed as possible so that I can't really yeah.
• 2:39
  SPEAKER_00: Oh, that's interesting considering you like are all part of like making it.
• 2:43
  SPEAKER_01: Yeah, but I there's just like a new toy every other week.
• 2:46
  SPEAKER_01: And if you're constantly like getting distracted by these toys, you're never actually gonna build something useful.
• 2:51
  SPEAKER_01: So I just try to think about you know, five years, where is this all gonna be?
• 2:56
  SPEAKER_01: And then build for that world, not for the Claude Bot, Claude Opus, you know, new model, not for that world.
• 3:03
  SPEAKER_00: It's funny you talk about like entering this whole new like dimension with everything that's out all the time now.
• 3:10
  SPEAKER_00: It is crazy.
• 3:11
  SPEAKER_00: Like I couldn't agree more.
• 3:12
  SPEAKER_00: It feels fully like a fever dream.
• 3:14
  SPEAKER_00: And I can't lie, I felt the same way when I tried out your stuff for the first time.
• 3:19
  SPEAKER_00: Literally was tweaking out like 20 minutes ago.
• 3:22
  SPEAKER_00: So before I get into any more of that, why don't you tell us what is Delphi?
• 3:27
  SPEAKER_00: Like what is the origin story, how that came about, and what it's used for?
• 3:31
  SPEAKER_01: Yeah, so in 2021, I had gotten early access to the GBT3 developer list, and I started my first company kind of tinkering with AI.
• 3:40
  SPEAKER_01: And at the time I was a solo founder, didn't really have any mentors, but I was reading a book about my grandfather, who was this larger-than-life entrepreneur in Iran before the revolution.
• 3:49
  SPEAKER_01: He had 30,000 employees, but then he was kicked out of the country, came to the US with nothing, and built a life for us here in America.
• 3:57
  SPEAKER_01: And I really wanted to get his advice and ask him, like, hey, what would you do in my situation?
• 4:02
  SPEAKER_01: Um, but that's not a question that a book can really answer.
• 4:04
  SPEAKER_01: And he had had a stroke, so I couldn't ask him directly myself.
• 4:07
  SPEAKER_01: But I, you know, was playing around with LMs and I ended up fine-tuning a model on his book so that I could talk to him and ask his question, him questions and get perspective in his voice and his style and his reasoning.
• 4:20
  SPEAKER_01: And that was obviously like very weird and like profound, and not necessarily because the information was so like deep, but rather, you know, getting information from someone that you like trust kind of makes you adopt it a lot faster.
• 4:35
  SPEAKER_01: And so that first startup got acquired, but that experience with my grandfather got me thinking about how we access other people.
• 4:41
  SPEAKER_01: You know, we can read books, read blog posts, watch a YouTube video, listen to a podcast, look at someone's LinkedIn profile, but you can't ask a question and really understand who they're about, like what they care about, what their experiences are, um, unless you have their in-person time.
• 4:56
  SPEAKER_01: And so we started Delphi three years ago.
• 4:58
  SPEAKER_01: We let people create a digital version of their mind, thinking about it like a new kind of artifact, the same way a book is an artifact, where you can experience someone through conversation.
• 5:08
  SPEAKER_01: It can answer any questions about you, your knowledge, and pretty much make you available 24-7.
• 5:13
  SPEAKER_00: So interesting.
• 5:14
  SPEAKER_00: And it's so personal for you as well.
• 5:16
  SPEAKER_00: Like with that idea where you wanted to like talk to your relative, you wanted advice.
• 5:20
  SPEAKER_00: Do you know what I mean?
• 5:20
  SPEAKER_00: But obviously, there was that that huge barrier there, which with technology you've been kind of able to conquer, you know.
• 5:27
  SPEAKER_00: Like, like you said, it's not just answering these questions with like the information, like it really captures the energy and the vibe and the tone of that person.
• 5:37
  SPEAKER_00: So I imagine you got your own Delphi.
• 5:39
  SPEAKER_00: I think it's one thing getting this advice and hearing stuff from your grandparent, but like when it was your own voice talking back to you, how was that?
• 5:48
  SPEAKER_00: Because I just had like such a fun time doing that.
• 5:52
  SPEAKER_01: Yeah, I mean, there I never listen to my own podcasts or any talks I do because it's just weird.
• 5:59
  SPEAKER_01: You hear your own voice, you're like, uh, what is that?
• 6:01
  SPEAKER_01: That's how I stuck.
• 6:03
  SPEAKER_01: Um, but it is a little weird, it's therapeutic.
• 6:05
  SPEAKER_01: You know, some people will will use it to talk to themselves.
• 6:09
  SPEAKER_01: It is really more meant as like an external essence of like, hey, if you want to learn about Dara, you can talk to my Delphi and you know, customers, journalists, everyone talks to it.
• 6:20
  SPEAKER_01: But then I think there's talking to a version of yourself that is almost like omnipresent.
• 6:25
  SPEAKER_01: It remembers everything about your life, and it can kind of know you more than you know yourself.
• 6:29
  SPEAKER_01: That's definitely an interesting experience.
• 6:32
  SPEAKER_00: So cool.
• 6:32
  SPEAKER_00: Like, I set this up.
• 6:34
  SPEAKER_00: I mean, I'll have you talk through the process of how to set it up, but you can add stuff like in articles that have been written about you or podcasts that you've appeared on.
• 6:43
  SPEAKER_00: So it takes all that information, and then when you ask it a question, it hits you with all that information you've learned.
• 6:50
  SPEAKER_00: And I don't know about you, but sometimes when I'm when I'm answering questions, there's some questions which you get asked so much, like you kind of do just have a script for it.
• 6:58
  SPEAKER_00: I was asking it these questions, and I'm thinking in my head, how would I answer this question?
• 7:02
  SPEAKER_00: genuinely, just to put it to the test.
• 7:04
  SPEAKER_00: And then and then Delphi comes out with exactly what I would have said.
• 7:08
  SPEAKER_00: And I just find that so, so baffling.
• 7:11
  SPEAKER_00: So if anybody was to set up their own Delphi, talk us through the process.
• 7:15
  SPEAKER_00: How do they go about training this AI to speak and think and respond exactly like they would?
• 7:21
  SPEAKER_01: Yeah, I mean, we've tried to make it as fast as possible.
• 7:24
  SPEAKER_01: Basically, you put in your name, we try to find whatever we can about you online, and then if you there is nothing, you can upload your personal notes, your journals, your YouTube podcast, blog.
• 7:35
  SPEAKER_01: And if you have none of that, we have this interviewer feature where it will ask you questions about yourself uh that will make him itself better.
• 7:43
  SPEAKER_01: So the more investment you put into it, the better it's gonna know you.
• 7:46
  SPEAKER_01: But then if you ask it a question where it doesn't know how to answer, it'll say I don't know.
• 7:50
  SPEAKER_01: So it'll know the the boundaries of its knowledge.
• 7:53
  SPEAKER_00: Okay, so it's not just gonna make some stuff up like ChatGPT sometimes does.
• 7:57
  SPEAKER_01: No, no, otherwise we can be working in the area of identity representation if people then trust it.
• 8:02
  SPEAKER_00: Mm-hmm.
• 8:03
  SPEAKER_00: That's very, very true.
• 8:11
  SPEAKER_00: As this is Tilly Talks Tech, I have a lot of people who I feel like would be like really interested in how this is made.
• 8:18
  SPEAKER_00: So what like coding goes into building like a clone?
• 8:22
  SPEAKER_01: Yeah, I mean, our main focus is is the mind.
• 8:25
  SPEAKER_01: Like, how do you capture someone's mind?
• 8:26
  SPEAKER_01: And the mind is a lot of things, it's temporal, the mind changes over time.
• 8:31
  SPEAKER_01: You know, your favorite color five years ago might be different than your favorite color today, and we have to be able to track that.
• 8:35
  SPEAKER_01: You change who you are depending on who you're talking to.
• 8:38
  SPEAKER_01: So it's contextual, you know, the way I am with an investor is different than the way I am with a customer, which is different than the way I am with my family.
• 8:45
  SPEAKER_01: Um so our back end is really like a knowledge graph.
• 8:49
  SPEAKER_01: Uh, can I share?
• 8:50
  SPEAKER_01: Can I share my screen or no?
• 8:52
  SPEAKER_00: Yeah, sure.
• 8:53
  SPEAKER_00: Go for it.
• 8:54
  SPEAKER_01: I can show you what it looks like.
• 8:55
  SPEAKER_01: This is loading my digital mind.
• 8:58
  SPEAKER_01: I'm about to show you my mind, which Whoa.
• 9:02
  SPEAKER_01: So this is this is like I upload all this data, we create this temporal knowledge graph that captures, you know, when something has been uploaded, uh, what the context of the data is, and you can see this big kind of lump in the middle is Delphi, because that's probably a big part of my life.
• 9:18
  SPEAKER_01: Uh so yeah, this is kind of how the backend is powered, and we have like a multi-model orchestration of agents that are on top of it on how to use it in conversation.
• 9:30
  SPEAKER_00: That's so cool.
• 9:31
  SPEAKER_00: So when we're training the AI, this is what's sort of being made while we're doing that.
• 9:36
  SPEAKER_01: This is what's being created, yeah.
• 9:38
  SPEAKER_00: That is so cool.
• 9:40
  SPEAKER_00: Can you see that on the dashboard?
• 9:42
  SPEAKER_00: I haven't seen that part.
• 9:43
  SPEAKER_01: It's it's not available uh publicly, but I can it will be soon.
• 9:46
  SPEAKER_00: So can you see like everyone who does Delphi, can you see their little like star little world that they made?
• 9:52
  SPEAKER_00: I can probably see your mind.
• 9:54
  SPEAKER_00: Wait, hang on.
• 9:55
  SPEAKER_00: What do you mean though?
• 9:56
  SPEAKER_00: What could you do with that?
• 9:57
  SPEAKER_00: What could you do with that information?
• 9:58
  SPEAKER_00: Or do you just leave it?
• 10:00
  SPEAKER_00: You just you're just chilling with like a library of minds.
• 10:03
  SPEAKER_01: Library of minds, the library of Alexandria.
• 10:06
  SPEAKER_00: What is the reaction that you get when you tell people, yeah, okay, I can make an AI clone.
• 10:10
  SPEAKER_00: I've got digital minds.
• 10:11
  SPEAKER_00: What do people say to that?
• 10:13
  SPEAKER_01: I mean, when we started this company a month before ChatGPT, like no one understood what we were doing.
• 10:18
  SPEAKER_01: And even after ChatGPT, people are like, this is dystopian, this is Black Mirror, what so it's almost been like three years of objection handling, and like people's biggest initial concerns are, you know, what happens to my data?
• 10:33
  SPEAKER_01: I don't want to give my data away to some tech company.
• 10:35
  SPEAKER_01: We don't train models on your data, you own your data.
• 10:38
  SPEAKER_01: You know, we're not like training AGI.
• 10:40
  SPEAKER_01: We believe data is best in the hands of the human.
• 10:42
  SPEAKER_01: Um, the second is like, what if it says something wrong?
• 10:45
  SPEAKER_01: Like, what if it hallucinates and makes me look bad and someone embarrasses me on Twitter?
• 10:49
  SPEAKER_01: Spend a ton of time on guardrails.
• 10:51
  SPEAKER_01: It will always say, I don't know, when someone asks it a question that's completely outside of its training data.
• 10:56
  SPEAKER_01: And the last one is, what if this makes me seem inauthentic?
• 11:00
  SPEAKER_01: And the way that we think about it is kind of what I said is it's an artifact of your mind.
• 11:04
  SPEAKER_01: The same way a song or a painting or anything that you create that someone else can experience is an artifact.
• 11:11
  SPEAKER_01: This is just a new kind of artifact that is interactive.
• 11:15
  SPEAKER_01: Instead of reading your thoughts, I am talking to your thoughts, and it's a way of uh consuming, you know, the the thinking that you've done in your life.
• 11:24
  SPEAKER_00: Yeah, that's really interesting.
• 11:25
  SPEAKER_00: Like, yeah, I can imagine like all the all the immediate questions because it just does just feel like such a like crazy future cyberpunk world.
• 11:35
  SPEAKER_00: Do you know what I mean?
• 11:37
  SPEAKER_00: But that's good.
• 11:37
  SPEAKER_00: I think that's great reassurance, and I feel like you know, we did an episode on the podcast that was all about data and things like that, and I think like that is a big concern.
• 11:44
  SPEAKER_00: So that is that is good that you collect you have the minds, but you know, there's no nobody's data's being tarnished or that's good news because I've I've already signed mine away if that was the case.
• 11:56
  SPEAKER_01: Yeah, I mean I think like the people are worried about IP and like the we're interested in a future where humans are still the main character, so like humans can own their data and choose what to do with it.
• 12:07
  SPEAKER_00: Yeah, okay.
• 12:15
  SPEAKER_00: This is my quiz time for my millions of questions that came up in the last like couple hours, okay?
• 12:21
  SPEAKER_00: What are the limits that are set on Delphi?
• 12:23
  SPEAKER_00: Like, could this be used in an unprofessional sense?
• 12:25
  SPEAKER_00: Is my question?
• 12:26
  SPEAKER_00: Like, because I find myself just talking to myself like on the daily, and I know that this is used for like a professional sort of like in a professional way, but could you possibly have your AI clone where like you can literally just converse like casually, or is that not?
• 12:42
  SPEAKER_01: Yeah, it's not meant to be a companion.
• 12:44
  SPEAKER_01: Like, I think we have a couple things where like, okay, this is we're not trying to be this.
• 12:47
  SPEAKER_01: We're not trying to be fake characters, like you can only create one of yourself, and there's like verification and all these things.
• 12:53
  SPEAKER_01: We're not trying to be like AI friends or companions.
• 12:56
  SPEAKER_01: Um, we think we want to promote actual human relationships, and this is almost like a first entry, like, almost like, do I do should I like this person?
• 13:05
  SPEAKER_01: Should I talk to this person?
• 13:06
  SPEAKER_01: Is this person good to work with?
• 13:08
  SPEAKER_01: Uh, and then we don't work with politicians or OnlyFans.
• 13:12
  SPEAKER_01: Um, yeah, those are kind of our rules.
• 13:16
  SPEAKER_00: Okay.
• 13:16
  SPEAKER_00: I was wondering if I could train my Delvi, because you kind of have the power in that respect.
• 13:22
  SPEAKER_00: You can train it.
• 13:23
  SPEAKER_00: If I could possibly train it to be like way more charismatic or like way funny and charming than I am in real life, is that something that you can do?
• 13:31
  SPEAKER_01: You could.
• 13:31
  SPEAKER_01: It's based on the data.
• 13:32
  SPEAKER_01: So if you did upload data, it's almost the best version of you because there is a version of you that is like that all the time.
• 13:38
  SPEAKER_01: And like it's it doesn't capture the version of you that is like lazy or tired.
• 13:44
  SPEAKER_01: I mean, the way we think about it is it's the version of you that your best friend gets.
• 13:49
  SPEAKER_01: You know, you're in a good mood, you're hydrated, you're like in your patient, and I think that's like you don't always get the best version of someone else, um, especially if you don't know them.
• 13:59
  SPEAKER_01: Like some people are like take time to open up, and Delphi lets you experience the best version of someone.
• 14:05
  SPEAKER_00: Yeah, interesting.
• 14:07
  SPEAKER_00: So I could like really just be like risen up all these employers, potentially from the comfort of my bed while I'm sleeping.
• 14:15
  SPEAKER_00: That's so funny.
• 14:16
  SPEAKER_00: So, what if we take it one step further?
• 14:18
  SPEAKER_00: What if I like lied about something?
• 14:20
  SPEAKER_00: Is there any sort of like fact checking that Delphi does, or could I tell Delphi, yeah, I did this, I climbed Mount Everest, and it would be like, okay, you climb Mount Everest.
• 14:27
  SPEAKER_00: So it does not work like that.
• 14:28
  SPEAKER_01: This is this is a great uh problem that we need to solve.
• 14:31
  SPEAKER_01: We haven't solved it yet.
• 14:33
  SPEAKER_01: It is uh, so there's two ways you solve it.
• 14:36
  SPEAKER_01: One is like fact-checking where someone can just call you out and you just actually look worse for lying.
• 14:42
  SPEAKER_01: Um but like YouTube, when YouTube started, they had to build a similar system.
• 14:48
  SPEAKER_01: People were uploading YouTube videos with content that wasn't theirs, and they ended up building a system that checked every YouTube video uploaded against a date a database of copyright material.
• 14:59
  SPEAKER_01: So it could flag it immediately.
• 15:01
  SPEAKER_01: So, like I think one of the things we'll have to build is that if you do upload data, one, we can check like to what degree of confidence are we sure that this is true?
• 15:12
  SPEAKER_01: And if we're not confident, we can put you through interview mode where we can like quiz you in real time and like ask you, okay, is this actually true?
• 15:18
  SPEAKER_01: Uh but we just haven't built that yet.
• 15:20
  SPEAKER_00: Cool.
• 15:20
  SPEAKER_00: And I saw like there's so many people on it now and stuff like that.
• 15:24
  SPEAKER_00: There's so much to learn on there.
• 15:26
  SPEAKER_00: It's actually so cool.
• 15:27
  SPEAKER_00: And I guess you can kind of think of it as like, I mean, everybody used to get all the information from books and things, like you say, and then we had access to Google, and that was like a big sort of advancement, which nobody knew what to do with back when books were everywhere and everything.
• 15:43
  SPEAKER_00: Um, but then you know, you adapt to Google, and this access, I feel like, is something that is amazing in this day and age.
• 15:50
  SPEAKER_00: On the topic of like the first question I asked you, where it was like, you said that you definitely don't want to be like a companion or anything like that.
• 15:57
  SPEAKER_00: I think you probably saw some of the crazy Chat GPT scenarios where like, you know, people people get a bit too attached to their AI bots, right?
• 16:06
  SPEAKER_00: And I was wondering if it was to kind of evolve into an unprofessional sense, why is it that you want to stick with the professional?
• 16:16
  SPEAKER_00: Like, do you think that could be dangerous somehow?
• 16:19
  SPEAKER_01: It's more just like my own values.
• 16:21
  SPEAKER_01: I think when you work on a company seven days a week, you want to create a future that you're excited about.
• 16:25
  SPEAKER_01: And I'm not really excited about a future where humans are talking to AIs versus humans talking to humans.
• 16:29
  SPEAKER_01: So I think like the best version of Delphi is like you have a Delphi and it can talk to every other Delphi in the world, and it can be like these are the five people that you should probably like actually meet and learn from.
• 16:40
  SPEAKER_01: And like you don't need to worry about whether they're the right fit.
• 16:43
  SPEAKER_01: I already did that work for you.
• 16:46
  SPEAKER_00: Yeah, that's cool.
• 16:47
  SPEAKER_00: Honestly, I think it's gonna be really good.
• 16:49
  SPEAKER_00: Like, Loki, I'm gonna be on there, like find a new podcast guest.
• 16:52
  SPEAKER_00: I feel like that would be one thing where it would be like that's a good one.
• 16:57
  SPEAKER_01: That's such a good that's such a good use case.
• 16:58
  SPEAKER_00: Delphi, it's it's out in the real world now.
• 17:02
  SPEAKER_00: How are people using it?
• 17:03
  SPEAKER_00: Are they using it how you wanted them to be using it?
• 17:06
  SPEAKER_01: I mean, we have two groups of users.
• 17:07
  SPEAKER_01: We have like the really famous people, you know, like Jay Shetty and Gary Brecca and um Arnold Schwarzenegger and all these people who are kind of using it as a new way of learning for their audience.
• 17:18
  SPEAKER_01: These are people that people look up to and ask for advice.
• 17:21
  SPEAKER_01: And then we have kind of the free-tier use case, which is hey, talk to my Delphi before our meeting.
• 17:27
  SPEAKER_01: So it handles the repetitive questions, and we our conversation can be more in-depth, or you know, you want to pick my brain, talk to my Delphi.
• 17:33
  SPEAKER_01: You want to learn about me, talk to my Delphi.
• 17:35
  SPEAKER_01: Uh, that's kind of the use case that we're excited about because like it's something that everyone can relate to.
• 17:40
  SPEAKER_01: Resumes are boring, LinkedIn profiles are boring.
• 17:42
  SPEAKER_01: Delphi is a way that humans represent themselves in the AI age.
• 17:46
  SPEAKER_00: Yeah, definitely.
• 17:48
  SPEAKER_00: Is there been any like stories of people using it where you're like, damn, I would have never thought to use it in that way, or oh, that's not quite what we're going for, or anything like that.
• 17:56
  SPEAKER_01: I mean, in the beginning for sure.
• 17:57
  SPEAKER_01: I mean, we we were pretty early, and so people were using it for like customer support and sales training and outbound sales and therapy, and you know, this is just a product that can be used in so many things.
• 18:11
  SPEAKER_01: And I think in AI, going back to what I said in the beginning of this call, there's a lot of shiny object syndrome, and the I think the biggest edge you can have is just like your ability to say no and focus.
• 18:22
  SPEAKER_00: That makes a lot of sense, 100%.
• 18:23
  SPEAKER_00: When I was exploring the website, one of your users said it has completely changed the game in terms of people having on-demand access to my mind, which is such a cool, crazy sort of statement, right?
• 18:36
  SPEAKER_00: First of all, was that Matthew Hussey, who's a relationship expert, or was that his Delphi?
• 18:41
  SPEAKER_01: That was Matthew.
• 18:42
  SPEAKER_01: That was Matthew himself.
• 18:46
  SPEAKER_01: You wouldn't know.
• 18:46
  SPEAKER_00: But that's cool.
• 18:47
  SPEAKER_00: So people are genuinely using this in their day-to-day, and you know, it's it's pretty new, but there's so many minds on there.
• 18:52
  SPEAKER_00: Like I said, there's so much to explore, which I think is really cool.
• 18:56
  SPEAKER_00: I wanted to learn like how does this work in the kind of economy?
• 19:00
  SPEAKER_00: Because Matthew, let's take him as the example, like, he's a relationship expert in real life.
• 19:05
  SPEAKER_00: Like, I'm assuming people will pay for his expertise in the same way you would for like consultancy, for example.
• 19:11
  SPEAKER_00: Like, for me personally, I'm like a public speaker.
• 19:14
  SPEAKER_00: So when I was speaking to my Delphi, it was clued up on all the things that I would be speaking about, right?
• 19:21
  SPEAKER_01: Yeah.
• 19:21
  SPEAKER_00: When my Delphi knows this, is it like a possibility where people would just go to the Delphi to get the information?
• 19:28
  SPEAKER_00: Basically, is it gonna bypass the actual person or are they able to monetize through it?
• 19:33
  SPEAKER_00: And basically, is this gonna be a whole new stream of income for people?
• 19:36
  SPEAKER_01: We view it as a top of funnel.
• 19:38
  SPEAKER_01: Like, we we don't think having a Delphi replaces the in-person experience.
• 19:42
  SPEAKER_01: We actually think it makes it more valuable, and we have data to show that it proves.
• 19:45
  SPEAKER_01: It's almost like when you uh when you watch someone's movie, you become a bigger fan of them.
• 19:50
  SPEAKER_01: When you read someone's book, you actually end up following them on Twitter and you're like, oh, this person actually has good things to say.
• 19:55
  SPEAKER_01: So Delphi provides access in a way that kind of opens up the funnel a bit.
• 20:00
  SPEAKER_01: And we have had people who have monetized their Delphi, and I think that could work in some ways, but a lot more people just use it as like a first entry.
• 20:08
  SPEAKER_01: Welcome into my universe, and it can guide people to the right resources.
• 20:12
  SPEAKER_00: Yeah, interesting, interesting.
• 20:14
  SPEAKER_01: You can see what people are asking.
• 20:16
  SPEAKER_01: So the insights, that's kind of one of the big features we're working on now.
• 20:19
  SPEAKER_01: Is if you know a lot of people are asking questions, you want to know who are like the top three people I should talk to, and then of all the things people are asking, what are the things that I actually should like double click into?
• 20:29
  SPEAKER_00: Right.
• 20:29
  SPEAKER_00: Yeah, it's go around in my brain, like the the possibilities for like, yeah, finding finding people who are talking about cool things to get on the podcast because I would use it that way.
• 20:37
  SPEAKER_00: Like I would get on there, ask the question, see who I need to speak to, and then get that person on, right?
• 20:43
  SPEAKER_00: And then like mingle in real life.
• 20:45
  SPEAKER_00: On the other hand, obviously, if I was getting guests on, I would go on Delphi, I would find somebody, I'd get them as a guest, I'd bring them on to the podcast.
• 20:54
  SPEAKER_00: But could I train my Delphi to kind of interview myself?
• 20:58
  SPEAKER_00: And could I train it the way I interview people and potentially interview myself?
• 21:05
  SPEAKER_01: Yeah, I mean, we've had people interview their Delphi's.
• 21:07
  SPEAKER_01: We've never had someone where they had their Delphi interview them.
• 21:11
  SPEAKER_01: Um so you could interview your Delphi.
• 21:14
  SPEAKER_00: Cool.
• 21:14
  SPEAKER_00: How would I go about that?
• 21:16
  SPEAKER_01: Call it and start recording.
• 21:18
  SPEAKER_00: Can I put that on a podcast episode?
• 21:20
  SPEAKER_01: Yeah, yeah.
• 21:21
  SPEAKER_00: It's so cool.
• 21:22
  SPEAKER_00: But it it it yaps about my life so well though.
• 21:24
  SPEAKER_00: Like it literally like actually blew me away.
• 21:26
  SPEAKER_00: Coming up with some things that even I forgot about, just because it's out there on the internet.
• 21:30
  SPEAKER_00: I have some quick fire questions for you.
• 21:33
  SPEAKER_01: Let's do it.
• 21:34
  SPEAKER_00: First one is what is the biggest misconception people have about AI in your opinion?
• 21:39
  SPEAKER_01: I think the biggest misconception is that the future is determined.
• 21:47
  SPEAKER_01: Um, and I say that because AI is changing so fast, and like the easiest thing to think is like, oh, every job is gonna be automated, and like this is where the future is going.
• 21:58
  SPEAKER_01: But I think like we Can create, we can define what do we end up in the cyberpunk future?
• 22:03
  SPEAKER_01: Do we end up in the Terminator future?
• 22:05
  SPEAKER_01: Do we end up in like the utopian?
• 22:07
  SPEAKER_01: We can actually decide that.
• 22:08
  SPEAKER_01: Like we it doesn't have to just be one way.
• 22:11
  SPEAKER_01: Um, and so yeah, that's why I think it's really useful as you think about AI is to like okay, what is the best version of this technology in five years?
• 22:19
  SPEAKER_01: It is free, it is fast, it is smarter than everyone in the world.
• 22:22
  SPEAKER_01: What futures are possible with that?
• 22:24
  SPEAKER_01: And then how can I be a part of making the future that I want?
• 22:27
  SPEAKER_01: I think that is something that I think is important if we want to end up in the non-terminator future.
• 22:33
  SPEAKER_00: Yeah.
• 22:34
  SPEAKER_00: I couldn't agree more.
• 22:35
  SPEAKER_00: And like pretty much exactly what you've just said is like a massive part and kind of how I end my presentations.
• 22:40
  SPEAKER_00: And I always say, like, no, we're not heading for inevitable dystopia, like, we're heading towards inevitable change, but within that, you have a choice, and that's how the presentation ends.
• 22:48
  SPEAKER_00: It's like, what do you want?
• 22:49
  SPEAKER_00: What do you want, what do you want the change to be in the world?
• 22:52
  SPEAKER_00: What do you see that future as?
• 22:54
  SPEAKER_01: You know, yeah, and it's harder.
• 22:55
  SPEAKER_01: It's it's harder to like think about okay, how do we end up in a future where it's not just like universal basic income, but universal basic meaning, where people like there are new things to do, and it just takes a little bit more thought, but we can we can get there with effort.
• 23:10
  SPEAKER_00: How do you think AI will integrate in our lives in the future?
• 23:14
  SPEAKER_01: I mean, again, there's the there's the black mirror wally future where we're all just like overweight because AI knows us so well, and we're just like in a never-ending uh feed of content.
• 23:29
  SPEAKER_01: And I say no to that.
• 23:33
  SPEAKER_01: I I hope that TikTok and reels are viewed as cigarettes on an airplane, where you know, there was a point in time where we did that, and then it was like, oh, that's actually not what we want to do, and we changed it.
• 23:45
  SPEAKER_01: Um, so I I'm hoping a future of minimal devices, presence in the real world.
• 23:52
  SPEAKER_01: We're not scrolling because we're not so afraid of optionality.
• 23:56
  SPEAKER_01: I think the reason we're on the internet is because the internet made the world bigger, more options, more people.
• 24:02
  SPEAKER_01: And a lot of us kind of panic under optionality because we don't we want to make sure that we choose the best thing.
• 24:09
  SPEAKER_01: And the great thing about AI is it actually makes the world smaller.
• 24:12
  SPEAKER_01: Instead of six billion options, AI will tell you these are the options that you should really focus your time on.
• 24:18
  SPEAKER_01: And obviously, there's a trust-building exercise where you have to trust the AI, but then once you do trust it, then you no longer have this optionality crisis and you no longer have the need to scroll, and you're more okay just like doing what you're doing.
• 24:32
  SPEAKER_01: So that's what I hope.
• 24:34
  SPEAKER_00: Definitely.
• 24:34
  SPEAKER_00: You seem you strike me as the type of person who's like always like thirsty for new knowledge and things like that.
• 24:40
  SPEAKER_00: And I find that really cool because now you've opened the doors for everyone else to have all this knowledge, like so much more easy, you know.
• 24:47
  SPEAKER_01: It was a platform that I created out of FOMO, literally wanting to know what other people thought and how I could get their advice.
• 24:52
  SPEAKER_00: Yeah, because it's like how cool is that?
• 24:55
  SPEAKER_00: Like, we're all are so trapped in our own little brain sometimes, like, and everybody is, like, everybody in the world is.
• 25:02
  SPEAKER_00: So it's so cool to be able to like tap into that, and obviously, like everyone, everyone who signs up is allowing you access to their brain anyway.
• 25:09
  SPEAKER_00: So it's like it's just grateful to those people who want to share that wisdom, I think.
• 25:14
  SPEAKER_00: So change is scary, but like you, I am a technology in general optimist.
• 25:20
  SPEAKER_00: What would you say to the people who are struggling with the concept of AI integrated into our day-to-day lives?
• 25:26
  SPEAKER_01: I would say be patient with yourself.
• 25:29
  SPEAKER_01: It's two real two pillars.
• 25:31
  SPEAKER_01: Start with the end in mind, don't get uh faded by you know the latest clawed bot or latest thread, all that is a distraction.
• 25:40
  SPEAKER_01: Um, start with the end in mind and then follow your curiosity.
• 25:43
  SPEAKER_01: The second thing is I remember I met the first uh prompt engineer and smart guy, but people on my team who just started prompting the last year knew more techniques than he did.
• 25:57
  SPEAKER_01: And I say that because there really are no experts in this space.
• 26:02
  SPEAKER_01: You could become the expert.
• 26:03
  SPEAKER_01: Like you, you know, even if you as a beginner right now, you could end up discovering a use case of AI that no one else has thought of.
• 26:10
  SPEAKER_01: So I think the most important thing is to have the beginner's mindset and you know to really realize that like no one knows what is going to happen.
• 26:19
  SPEAKER_01: No one.
• 26:19
  SPEAKER_01: I don't know, you know, Sam Allman.
• 26:22
  SPEAKER_01: No one is an expert in this space.
• 26:24
  SPEAKER_01: It is so unpredictable, and I think that can be empowering.
• 26:27
  SPEAKER_00: Definitely.
• 26:27
  SPEAKER_00: I think it's so cool.
• 26:28
  SPEAKER_00: Like, we're definitely like going through so much change in that regard.
• 26:32
  SPEAKER_00: Like it used to be so traditional, whereas now there's so much access, and like you said, like you can figure out anything.
• 26:38
  SPEAKER_00: You know, this whole journey started because you wanted to speak to your relative, you know.
• 26:43
  SPEAKER_00: Um, and bang, now like people can have access to minds all over the world.
• 26:47
  SPEAKER_00: I think that is just crazy.
• 26:49
  SPEAKER_00: Like, you gotta zone in on what's happening in your life, what is the change you want to see, and then help build it because it's a lot easier now than it used to be to be able to go out, set your mind to something and achieve it, I think.
• 27:00
  SPEAKER_01: Yeah, yeah.
• 27:00
  SPEAKER_01: So you just follow your curiosity, have that childlike curiosity that we once all had.
• 27:05
  SPEAKER_00: Awesome.
• 27:05
  SPEAKER_00: Thank you, Dara.
• 27:06
  SPEAKER_00: I'm gonna actually take that with me into the next, like hopefully forever, you know.
• 27:11
  SPEAKER_00: We should never we should never lose our curiosity whatsoever.
• 27:14
  SPEAKER_00: Thank you so much for talking to me and sharing with me Delphi, this incredible new what do you call it, like a platform?
• 27:22
  SPEAKER_01: Yeah, a platform.
• 27:23
  SPEAKER_01: It's it thanks for it.
• 27:24
  SPEAKER_00: It's like a messaging platform where you message yourself.
• 27:27
  SPEAKER_01: Exactly.
• 27:27
  SPEAKER_01: Or I'm gonna I'm gonna talk to your Delphi after this to see if I miss anything.
• 27:30
  SPEAKER_00: I'm gonna ask your Delphi everything I've just asked you to see if you say the same things.
• 27:35
  SPEAKER_01: Yes, that'll be a good eval test.
• 27:38
  SPEAKER_00: That will be so interesting.
• 27:39
  SPEAKER_00: Yeah, anybody who wants to find you or Delphi, where should we go to find out more about this?
• 27:45
  SPEAKER_01: Uh Delphi.ai slash Dara.
• 27:47
  SPEAKER_00: Perfect.
• 27:48
  SPEAKER_00: Thank you, Dara.
• 27:49
  SPEAKER_00: I'll see you next time.
• 27:50
  SPEAKER_01: See you next time.
• 27:52
  SPEAKER_01: Bye.
• 27:53
  SPEAKER_00: That is crazy.
• 27:55
  SPEAKER_00: I am definitely gonna go ask his Delphi all the same questions and let me know in the comments if you want me to drop the comparative doc because I am so interested.
• 28:04
  SPEAKER_00: Also, you can find me on Delphi.
• 28:06
  SPEAKER_00: What?
• 28:07
  SPEAKER_00: You can literally talk to me, and I I am Tilly Lucky.
• 28:10
  SPEAKER_00: I can back up the fact that my Delphi is saying what I would say, which is kind of mad.
• 28:15
  SPEAKER_00: So if you wanna if you wanna chat to me or my my my Delphi, you can you can find her.
• 28:21
  SPEAKER_00: Anyways, thank you so much for listening to this episode of Tilly Talks Tech.
• 28:24
  SPEAKER_00: I really hope you enjoyed it.
• 28:25
  SPEAKER_00: I had so much fun on this one.
• 28:27
  SPEAKER_00: I love diving into these like very specific techie tools.
• 28:31
  SPEAKER_00: I feel like we haven't done too much of that, but let me know if we should do some more.
• 28:34
  SPEAKER_00: It's definitely something I want to do.
• 28:36
  SPEAKER_00: If you did enjoy this episode, please make sure to like, comment, and subscribe.
• 28:39
  SPEAKER_00: Share it with a friend.
• 28:40
  SPEAKER_00: Let me know who else we gotta get on, and I will see you next time.
• 28:44
  unknown: Bye.