Building Real AI Humans: The Future is NOW

We built machines to be these unambiguous, logical communicators that can remember everything. And humans are ambiguous, we're lossy, we speak as much through our gestures as we do with our words. The future isn't that machines are entirely human, but they're actually some combination of the two.

It's so hard to wrap your mind around what it could be or what is even possible with technology. This is really so futuristic and so cutting edge that part of our job, and I think something that a lot of AI companies are really having to deal with, is actually painting the picture of what the future can look like and being descriptive with what technology can do for us. And that's part of the fun of building it today. It's so different than 5 years ago.

A significant part of applications of the future will actually be these human-like applications, these pal-like applications.

AI is exploding right now. In your experience, what are other AI companies getting wrong about what they're doing?

There's a lot of people building that don't actually have a love for technology. They have a love for money. At Tavis, we have a love for humans and a love for technology. That's what makes what we do very unique and special.

I've got one more question on hiring. I'm coming for a job at Tavis. You got one question to ask me to see if I'm a good fit. What would you make that question? Quyen, Hassaan, thank you for having me. Welcome to the podcast.

No, thanks for having us.

We're excited to be here. It's been a while since the last one we did. I'm glad you could be at the office this time instead of remote. This is so much better than Zoom.

I think we were, what, February, March last time we recorded?

Yeah, it must have been. It's close to a year. It's been— a lot has changed. It's been a while.

Yeah. And seeing your office today, I'm going to have to show some B-roll of what this is and what's happening. Can you explain the inspiration behind this office, please?

Yeah, I mean, it's a few things, right? Well, I think like an office is like a physical manifestation of the culture of a company. Um, and for us, like, I think we're really into being very hands-on and being very creative. And we have a huge respect for just vintage computing and putting ourselves in the shoes of these amazing designers and developers that had no priors. And so we love tinkering. We really love being able to use old technology. We also love old cars. So it's just like a physical manifestation where we have this amazing museum of old computers. That we can use and be inspired by. We can go fix them, we can create new things. And that's just like, that's our culture here.

And I think it's more than that too. I think part of it, part of the root of the office is we are building something brand new, right? We are building something that hasn't existed before and we need a space that the team can be incredibly creative, innovative, think about, inspired by. Exactly. And that's why, you know, we've got the brick walls and the wood, right? It feels like a workshop in some ways. You've got the natural light because it feels, you know, uplifting and comfortable. And everything was really done with such intent and attention to detail where the team can feel inspired and actually want to come here and put in the work and the time to build this future.

100%. If you're building the future of computing, then you should be able to go and see what it was like. So our team will often go to one of these old machines we have and turn them on and put themselves in the place of someone who had never used the internet before, or the person who was designing it, or they'll go read one of these early user guides and manuals and all these things. And it's really inspirational and allows them to be creative in a completely different way.

You're really blending the past and the future together. Talk to me about what you think the future looks like and how Tavis fits amongst that vision.

Yeah, yeah. I think a lot of what we think about is what does the future of using a computer and software look like? And if we think about the history of computing and why we have some of these old machines, then you can sort of see how we went from these machines that were super complex to use, and then we got to command line interfaces, and that opened up a larger population of people, these more developers, these tinkerers who can use these machines. And then we got GUIs, and now you've opened it up to everyday people being able to learn and use these machines. And so if we track that evolution, what has happened is that machines have become to some degree a little bit more human, but they haven't quite gotten there yet. And we've been like watching and reading the sci-fi for all these years around these machines that could really understand us and they felt like us. And so what we see as the future of computing is human computing where you don't have to learn like how a computer works. Like, you don't have to learn how software works. It knows you, it understands you, and it meets you where you are. And that's what we're building here at Tavis.

It's interesting because I've spent the last couple of days at Dreamforce, and one of the companies there is Figure. So they've obviously got these humanoid robots, but they're still so far from looking like a human and interacting like a human. So there's still this cognitive dissonance there where it's like, I know that that's not human, it's definitely a robot. And I remember doing research on you guys and something that really stuck out to me was you put it as the OS for human and AI interaction, right? Unpack that for me, please.

Yeah. So if you, like, Tabis in a nutshell, we're a research lab focused on human computing, which is teaching machines the art of how to be human. And humans are like really complex communicators. Like this conversation we're having right now, it's a dance, it's a waltz we're doing together. And machines are in the corner doing the robot. And so how do we get them to join us in the waltz? So a lot of our research focuses on teaching them to see the way that we see, to be able to hear the way that we hear, and to teach them the relationship between how we talk and the reactions that we have, and then be able to react back appropriately as well. right? And so, and if we can teach them to do that, then you can essentially teach these machines to like really understand the nuance of what we're saying and even be able to look a lot like us and be able to have those microexpressions as reactions. And once you do that, then suddenly like computers aren't this thing that are like, they have to learn and you know, that some people are good at using, some people aren't. Then like the future is something that you can just use a computer the way that you talk to a friend or you work with a coworker. And that's a future we're really excited about.

And it comes back to, I mean, when we were growing up, when we were starting to use computers, we had to adapt to the computer. We had to learn how to talk to it. We had to type instead of talk. We had to use emojis instead of gesture. All of these different adaptations where we had to do it. I think the base premise here is how do we have the computer or the machine meet us where we are?

Change the dynamic. Yeah, flip the dynamic entirely where instead of us learning the language of the machine, the machine learns the language of humans. And that changes who can use a machine well and who can't. Our effective ability to take our idea and turn it into reality is very different than say my mom's, who's incredibly intelligent, but maybe doesn't use a machine as well as I do. And the question is, why should she have to compromise on the way that she communicates? A machine should meet us where we are.

Do we lose anything in not doing that hard work of understanding how machines work and how code works and speaking that different language, theoretically, that's going to make our brain change and have to adapt and learn. If everything's just a conversation, are we kind of losing something there?

I think that in a lot of ways, we built machines that are opposite. We built machines to be these unambiguous logical communicators that can remember everything. And humans are ambiguous, we're lossy, we speak as much through our gestures as we do with our words. And so I think the future isn't that machines are entirely human, but they're actually some combination of the two that is special. It's not fully human, but it's not fully machine either. So it's human-like in the sense that you can build a relationship with it, it evolves, it understands your emotions, it has emotional intelligence, but it still is able to help you 24/7. It still has unlimited recall. It still is, can be very logical. And I think bringing the two together, making them meet, is really the real special place. And that's where you want to go.

Yeah. It's more man, it's more machine. You get the best of both worlds.

Something in the middle. Not quite man, not quite machine, something beautiful in between.

I want to really break down what you described as the waltz. I'm an alien, I come to Earth, and I'm trying to understand what is human communication. What's important? Is it voice, tone, body language? Like, break down those components for me. What goes into communication?

Yeah, I mean, the reality of it is that it's so much of those things coming together in perfect harmony. And that's why it's a waltz, right? Is that, let's just take like turn-taking, for example. So me saying what I need to say and you responding back. You know, if you purely look on like a mathematical or like a measurement level, you're like, okay, yeah, yeah, the humans, humans that really connect, they respond each other as low as 250 milliseconds. And so theoretically you could tell a computer, all right, after I finished, just respond back in as low as 200 milliseconds. And what you'll get is a conversation that's like shit. And the reason isn't that it's too slow or it's too fast, it's that it's unintelligent or it doesn't have the right context. Because the reality is humans are very dynamic. If someone says something thoughtful, maybe we'll just sit there and let them, let the silence. Speak. If we're like in like a really great cadence, we'll respond back really, really fast. And so it's very contextual. And so we do things by taking all these signals together. So how you said something, like, you know, how I interpreted it, like all those things come together. And so a lot of it is understanding the relationship between these different signals, right? So what is the face that you made? What is it that you said? How did I react to it? And breaking down the relationship between those things is really, really important. So all of it is really important, but like, you know, in vacuum, like each and every one of them, you know, it's like, it feels like, ah, you just get one right, but you actually have to get the combination of them right. And that's the most important thing.

And it's the fluidity of it as well.

Yep, absolutely.

Yeah.

Right. It's ever evolving, ever ongoing, live.

Yeah, totally.

Before we get into the technical components, hypothetically you want to model this after the best human communicators that you've seen, what are good examples of that?

Well, I think that maybe it's not the best human communicators, but because humans are so diverse, some of us communicate in such different ways. So really, if you think about what we're doing is we're building really robust human simulation models or human behavioral models. That can model really dynamic differences between conversation to conversation. And so it's things like in the first few back and forths, you can detect to say, oh, this person clearly speaks a little slower and they're more thoughtful, and you adapt the conversation versus like, hey, we're really hitting it off and we're speaking really fast, we're so excited. And it's like, okay, I'm gonna do that instead. And so I think the diversity of conversations is actually more important than having the best communicators. And in that way, you can be very dynamic to it.

Break down for me what you're building. From what I understand, you'll get different models that are specific to a tone of voice, to body language, to understanding facial expressions. How are you building them and how are they different to what's currently out there with other startups?

Yeah, that's a great question. So I'd say there's a few different areas, right? There's perception, there's understanding, there's reaction, and there's actions, or we say agentic. On the perception side, this is one of the most important pieces of research that we do, which is how do you actually capture the right signals in human conversation? So being able to look at your face and extract expressions from that, hearing the audio, your voice, and saying, okay, what's the tonality? What's the prosody? What emotions are there? And being able to collect a ton of signals there, right? So the perception is one of the most important things that we do. And Tavis is one of the only startups that's able to take as much signal as we can. But with it, like you have all this signal, you have to actually do something with it. So then the understanding piece is even more important. So a lot of our research work and our model work also is focused on being able to take all these really amazing input signals and actually create understanding out of it. And this is where we talk about the relationship, which is like, okay, so we have this really amazing data on what you just said and how you said it. And we also have data on how I reacted to it. Okay, what do we need to do with that next? What's the relationship we can extract from that? So we spend a lot of time on the understanding piece. Then of course, we want to be able to react and actually say something back. And so being able to take those signals and say, okay, based off of what we've seen so far, I should react this way. I should be able to— this is the emotion I should have whenever I'm reacting. And we spend a lot of time on being able to create the most realistic real-time replica models. So being able to render a face that looks really realistic. And then Once you have this amazing AI human, then you want it to be actually useful. So being able to actually take action the way that humans do. And those are all the pieces that we focus on. And each one of those is incredibly difficult, but then doing it all together and doing it in real time, that's where we really excel, which is like, okay, you have this robust AI human that can see you, that can hear you, can understand you, and it can talk back and it looks like you're in a video call, but in the background it's also doing stuff for you. So it's the combination of these things coming together that's really special. And so we build state-of-the-art models to be able to do that in these really amazing agentive systems.

So the models and the research is all very interesting, but I kind of want to understand it from my perspective. If I want to work with what you guys are doing, how do I go about that? Like, what's my experience?

Yeah, so I mean, we have really two entry points to use Tabas. One is, of course, our APIs and our interface, CVI, for you to be able to use all these models and build these really amazing applications on top of it. but we also have this new release called the Tavis PALs, which is these fully embodied experience, right? And we talked a little bit earlier about what does the future of computing look like? And how like today traditional applications are like these static GUIs. You have to learn how to use, like you have to learn how to use them. And they're sort of just like these things that like, they're just there, right? Like you have to go and decide to use them. The future of computing and the applications are something that's more human-like. These are gonna be applications that actually evolve with you. They understand you, they see you, hear you, they respond to you. You can video call them, you can phone call them, you can text them, and importantly, they can do the same to you, right? So they're proactive, they're always on. And so the Tavis PALs are this, they're like this future type of application where you can have the perfect AI assistant that reaches out to you and can meet you where you are, and you can talk to it just like you would a real assistant. Or you can have this amazing companion that has your emotional understanding and can help you and call you whenever you're feeling down or want to talk about something. We're really excited about this release because it allows consumers to experience the future of what a computer and AI looks like.

Where did the entire concept of this come from? Did you have people coming to you being like, I want this to happen, or was this more just a product of all the research you're doing, like people need this technology?

I think it came from us really being obsessed with how humans and computers interact today. I think a lot of our research focuses on this area of effective and human computing. And some of our research team has been working on this problem for the last 30 years. Trying to teach machines how to be more like humans. And I think we've all been obsessed with like, okay, how do we bring Cortana from Halo and Data from Star Trek and Jarvis and all of these to like, how can we like realize that science fiction and make it real? Because like that's really what we've all been about. And so have people been asking for it? Yeah, I think parts of it, but I think that people didn't realize that it's possible and our job is to actually make it possible.

People are definitely asking for it in the sense of like when I talk to people like, oh yeah, this is terrible. I'd love a solution to blank. But it's so hard to wrap your mind around what it could be or what is even possible with technology because this is really so futuristic and so cutting edge that part of our job, and I think something that a lot of AI companies are really having to deal with, is actually painting the picture of what the future can look like and being prescriptive with what technology can do for us. And that's part of the fun of building in today's today's age. It's so different than 5 years ago.

Totally. And it also is just like, we're really obsessed with this concept of friction and how much friction there is in using technology today and how these machines haven't really evolved as much as we'd want them to. Okay, we have these GUIs and now we have these chat interfaces, but is that really how we want to be using our computers? And so it really revolved around how can we make it How can we make using computers second nature? You don't even think about it.

I think you're uniquely positioned. I've just been at Dreamforce. A big thing they talk about is agent force, loads of companies there building agents. It's like, okay, great. What does that agent look like? Yeah, you're in Slack and you're messaging back and forth with the chatbot. And I think you're very uniquely positioned with how frictionless that you're making this. You're one of the only companies that I've spoke to that's put such concentrated effort into making this more human, into making it frictionless, right? I mean, I imagine the future of work, I want something that's conversational, right? I don't want to go in and type lines of code and put in bullet points. I want to log onto my laptop, be like, I've got this great idea, I'd love to go make this happen. How do we conceive of this and make that happen, right? And you're one to— you're the enablers of that. My question for you is, like, if you had any predictions into who's going to get the most use out of this, who is this best for? Like, because when my mind thinks about this, out of the interviews I've done, there's a million ways you could take this. You could take it into marketing, you could take it into ops. Some people that I speak to a lot are in, like, therapy, for example, and I can picture that there. Like, talk to me about what the future of that looks like.

We actually think that a significant part of applications of the future will actually be these human-like applications, these PAL-like applications. The future of therapy or AI therapy is something that's emotionally intelligent and has all these signals and can meet you where you are. You mentioned these AI agents for businesses. They shouldn't be these agents you have to learn how to use. Like, they should just feel like any other employee that you can text, you can call, that can do things for you, can reach out to you. So I think that in every domain where some level of human connection is needed, this is, this is the perfect fit for. So whether it's customer success, customer support, just having a great marketing buddy or having a great companion or assistant, like all of these will need a human interface to be able to interact with them. And that's— so we see it as a very horizontal application.

I mean, there's been some awesome use cases, like ones that we never would have guessed that have just popped up. People start building it and stuff. So my favorite part of this, you met some team members earlier when you came in. Every single team member, we ask them when they come on board is like, what is your vision for this technology? Or what excites you? Or what is the use case that you'd be excited about? And every answer is completely different. For me, it's education. I get incredibly excited to think about how education can look different in K through 12. If someone's able to ask all the questions they want to ask, if they're able to have the personalized attention they need. For other people, it is entertainment. For some people, it's healthcare. For people, it's across the board. And I think that's the beauty of it. Every single person on the team. Every area for that already.

Yeah, exactly.

And every day it's something new. So the challenge for us then is how do we build this broad interface that can work for everyone, but also work well enough that they can actually use it in all of these different cases.

Yeah. I mean, what we say to people whenever they think about, okay, like what are the Tabas Pals? And what I'd say is that they are the agents and the LLM for the rest of us. Like they're not focused on being something that developers have to learn how to use or that are for a more technical audience. It's that the everyday person can get the best of what AI has to offer. And it really it truly understands them. And so it's beyond what any LLM chatGPT can offer because it can reach out to you and it can really understand you. And it's not just chatting, you can video call it. Those are all really, really special pieces that make for an application that's just a lot closer to your friend or your coworker than a machine is.

Yeah.

When you're building in such complexity, And you're almost building in this emotional understanding and resonance. Are you at all concerned about where that ends up? Yeah.

I mean, I think that like the alternative is way, like I think that like with any new technology, there's always some degree of risk and downside. But I think that the positives and where it can go far outweigh any negatives. I think that the alternative is that these machines never truly understand us and they will get way more intelligent, but they don't get emotionally intelligent. And that means they're less human-aligned versus our vision is like, okay, how can we get machines to be so human-aligned to have emotional intelligence? Because we think that's the most important thing. And there are certainly some risks and some downsides, but we're already seeing with what people build on Tavis today that has such a massive impact where people will often talk about, well, you know, is, is like an AI therapist, uh, a good idea? If you'd asked me a couple years ago, I've been like, that's crazy. But now, if you, if you really think about it, like, that, um, that opinion of ours, like, oh, AI therapy is not a good idea, is sort of like a privileged take because we have the ability and the means to be able to afford therapy. Most of the world does not. and the alternative for them isn't a human therapist, it's no therapist. So, you know, we love working with these AI therapy companies that are clinical grade. They've been tested and they're offering therapy to people who wouldn't have had access to it at all. So like, I think we're really excited about the impact it can have.

And you see it left and right. I remember vividly there, so we had a concept called AI Santa. Which we launched last year. You'll actually, for anyone watching this, you'll see it pop up again in a week or two, which is super exciting. But I remember we launched Santa around Christmas time last year and at the end of Christmas around January 1st or so, as scheduled, we took it down and we got an email from this individual, amazing human, absolutely amazing. And he sent this note saying, hey, look, Santa was the first person who actually understood me. I I have severe cerebral palsy, I have a speech disability, I have these things. And Santa's the first person who actually took the time to understand me, to give me space and get to know me. And so of course we stood up an individual link that he still uses today and talks to Santa with. And it was this amazing story. But I think the other part of this question that's really important is it is our responsibility to bring this to life in the way that is human-aligned and human-centric. As the builders of this, we're not scared of it because we're the ones that are bringing it to life in the right way, in the ethical way. And that's our responsibility.

How do you wrangle that responsibility? What conversations do you have? How do you come to terms with it?

Yeah, I think it's an evolving thing. I think that like we, we try to have a lot of discussions, um, with our, with like our users and our customers and people who have never used AI in this form before. And so we try to do a lot of research. Part of our research is just focused on the ethical aspects of this. And I think part of it is policy. What do we not allow? What are things that are harmful? And trying to have a view on that. And I think that's an evolving thing. I think that we don't have all the answers today, but what we do try to do is really make sure that the right guardrails are in place, that we're doing a lot of really great work against hallucinations of any time or any sort of dangerous talk on the platform. So there's a lot of very thoughtful things that we do, but it's an evolving thing and it's something that we take very, very seriously, even since day zero at Tavis.

Yeah, I mean, that's where the— there's the practical and the tactical prongs that come about from that, right? When it comes to data privacy and ownership, when it comes to our model development and how that goes, when it comes to use cases and guidelines that we allow. It also just comes back to the culture that we've built here, which is people first. And this is one where it's not just for the users, it's not just for the customers, but it's for the team. Everything we do is human-centric from the benefits we offer and how we hire to the culture we've built to actually how we build our models and build our product. And it oversimplifies it to an extreme degree, but everything is rooted in people first. And if we abide by that, it comes together.

I think what it comes down to is a lot of AI companies, they're focused on machines first, humans second, like intelligence at all costs. I think Tavis is like solidly human first, machine second. Like our immense respect for the waltz, for how humans operate, drives everything that we do. And so even like disclosure, for example, like absolutely Tavis will pass the Turing test in the next year. Where you won't be able to tell, the Tavis Turing Test. Tavis Turing Test. You won't be able to tell unless you're told, but you will always be told. So AI, these AI humans will always disclose that they are AI, but you'll know and you won't care. And so things like this where we have such immense respect for humans and human connection that drive us in everything that we do.

Conceivably, the world's at your feet. Where do you see this in 10, 15, 20 years' time?

I think in 10 years' time, the people won't even think about using a computer. That won't even be a concept. You won't even, you're going to wake up and not think about it. It's not going to be something that you have to wake up and be like, I have to get on this thing and wrangle it or I have to learn how to use it. Truly, it'll be what we call the invisible interface. The invisible interface will be commonplace where like using a computer will be second nature. You'll be talking to your perfect AI pal assistant that will be just like waking you up and saying, hey, it's time to get ready for your day. It like texts you all the things you need to do for the day. You're on the run, you're running late. It's like, I got you. Let me just move some stuff around. After a great meeting, you video call it and like just tell it all your thing. You're brainstorming with it. You have your amazing like AI friends. You have your amazing AI coaches and all these things. And these are just like commonplace. I think that's the future and that we'll see in 10 years. And these things will be so aligned. They will understand you so well. They'll understand every emotion that you have when you're feeling down, even the things that you don't say, it'll understand.

And they're involved in every part of your life.

Yeah.

Right.

And whether it's your specific pal that you built a relationship or a pal that's working with or for someone else, say in healthcare or education, it will change how we fundamentally experience technology and how we're able to communicate with technology.

You'll have AI coworkers that you think of no differently than your real coworkers. Yeah. Right? Like you'll just talk to them the same way that you do and they'll be in the meetings. Like it'll just be commonplace. Like you'll have your, you know, AI healthcare intake assistant that you talk to before you get to the doctor that like makes sure that the doctor knows everything about you and has way more context so they can help you in a much better way.

Yeah, we're not quite 10 years out yet, right? But we try to build that internally a little bit. Like we have an AI HR consultant internally that any of our team members can go and talk to.

We have an AI SDR on the website that Honestly, all of these are, okay, I say AI healthcare intake assistant, but we already have multiple healthcare companies using those things.

And I think more so though, it's the in 10 years, it won't just be in this office, it won't just be in San Francisco, it will be in— be the new normal for how we work with machines. Yeah.

It sounds pretty utopian, right? I want to live in that world. But the doctor intakes, the personal assistant, all of these different jobs, what happens to those people that are displaced? Like, it's a hard question to answer because no one knows, but if you were to make predictions, what do you think?

Well, I think that like our job isn't to replace humans, it's to replace bad machines. Because the reality is, like, do you have a personal assistant?

No, but there's a lot of humans that do jobs that a bad machine could do.

Yeah, I mean, that's totally true. I think that like What we're doing is we're equalizing access to things that people otherwise didn't have. And I think that real human experience will be even more cherished. And there's a lot of jobs that will continue to exist, but what this means is that you can have a personal assistant that you wouldn't have had, but you weren't going to hire a real personal assistant. You were just going to have nothing. And so it equalizes access. And certainly we know there'll be some displacement. We don't shy away from it and be like, ah, it definitely won't be. But I think it will improve the everyday lives of everyone across the board. And I think that's what we're really focused on is there'll be some, but I think our job is to replace the bad machines that are already out there that are already like, they're just like removing the human from the experience. It's just an overall regressed experience and we're solving that problem.

Yeah.

And it's almost, so there's two takes on this, right? One of them is you take a grocery store and a self-checkout lane, right? In theory, most grocery stores could have been close to completely automated. Yet there are still humans in there because the human touch matters so much, right? And it doesn't make sense actually to completely automate it, or it is a regressed experience in many ways. And the other take on it is this is where, you know, it really matters that we're able to evolve the systems with us as well and bring in a 4-day workweek or a 3-day workweek. These humans weren't inherently meant to work under fluorescent lights from a 9 to 5, right? We're meant to communicate. We're meant to have the dance. We're meant to evolve and be able to help society evolve with us.

Utilitarian net benefit.

Utilitarian net benefit.

If you had this 2-day work week, this 3-day work week, how do you think you would spend that time? I imagine you two as founders would probably spend more time working on this, just in a different way.

For us, it might go from 7 days to 6 days.

Okay. Yeah.

I mean, I think that, I think that what it unlocks, like, I think we're very lucky here at Tavis that every person here is doing what they consider their life's work. This is what we want to do. We want to do this on the weekends. We want to do this on the nights. Like, you know, we have, all of us have alternative hobbies, like certainly these computers and the cars and everything. But this is our primary hobby and the thing that we love the most. It's a work of love, really. For a lot of people, they don't necessarily always have the opportunity to do that. And a reduced workweek gives them the opportunity to work on or find a passion that they, whether it's like just reading or it's drawing or it's painting or it's anything, like they have the opportunity to spend time doing that. Because one of the most impactful things you can do in life is look back and say, hey, I spent time doing things that I loved so much. And that was, that was my legacy. I think that will allow more people to do that.

I'm just coming to terms with that. You two have been building this company for how long now?

Just about 5 years.

Yeah.

It's been a while.

Before AI was the cool thing to do.

Way before AI. When we were going through YC, we were one of 2 or 3 AI companies and everyone was telling us, you should do blockchain instead.

You should do Web3, NFTs.

Let's come a long way.

I bought a couple of NFTs. I think I made NFTs at some point as well.

Did you have the The chimp ones.

I don't have any of them because I'm definitely not that wealthy, but I've got a couple of like Gary Vee defunct NFTs still sitting in some wallet in a MetaMask somewhere. So I got fooled into that as well.

One day it'll be super valuable.

Oh yeah, that's the promise I'm living by.

You're going long on them. Yeah.

I'm so 5 years in the industry. AI is exploding right now, right? Like the past 2 years, the amount of VC investment that directly just went into just AI. In your experience being sort of like the old heads of this industry almost, what are other AI companies getting wrong about what they're doing?

There's a few things.

Yeah. I think it's a few things.

Can name drop as well.

It's fine. They know who they are.

You know, Gio Hotz, who like, you know, famous programmer, hacker, had this recent article that basically said that there's a lot of people building that don't actually have a love for technology. They have a love for money. And some VCs love this because VCs, some VCs just love money. And I think that fundamentally, at Tavis, we have a love for humans and a love for technology. And I think that is something that is really, really important to be able to actually create impactful, longstanding impactful work rather than the get-rich-quick schemes that we're seeing in a lot of things, right? So I'd say like that's my version of it is really just like really understanding and believing in what you're building and saying, this is my life's work. This isn't just like applying AI on something that I know will make money, which is great. That's like a totally valid thing to do. But I think that's what makes what we do very unique and special.

And it goes hand in hand with that exactly. But people don't build with taste. Taste and craft and passion for what you do is so important. It shows up in the small details from how it feels when you use a product for the first time to the experience you have throughout the journey, all the way to your first impression of them when you visit a website. And that tastefulness is so, so important, especially in today's world.

What has building AI taught you about being human?

That it's really hard to emulate humans. Like, humans are immensely beautiful, complex creatures. Like, just the nuance of it. We do this exercise whenever we're at our summits or retreats that I call the art of conversation. And we pair people up in pairs of two and we say, okay, go. It's the best that they haven't met each other before. And we say, okay, go for 5 minutes, set your phone down, record y'all having conversation. Everyone comes back and we say, okay, everyone watch the conversation and start journaling how you felt, when you felt like there was a connection built. Okay, what happened? And you find all the small nuances like, oh, that laugh at the beginning broke the tension. And the way they said this, like, made me feel so special. And you— and like, it like opens everyone's eyes up to like, holy crap, there's so much that went on in this that like the smallest nuances of like how long it took to answer, the like the intonation changing, like the nonverbal cues, the gestures on your face, like, oh, you raised your eyebrow and that told me something. And so just the nuance of humans is what we've learned.

It's the smallest of small things. Truly, every time I've done that, I mean, you and I have done that interaction together multiple times and we've known each other for forever.

And you learn so much about each other.

You're like, oh wait, wait, wait.

He totally does that every time he's doing this.

It's like, oh, that small thing made the total difference. And now that conversation felt enjoyable versus— Totally. Sure. Right?

And there's like tells. You realize there's tells in people that you like, if you just analyze them, you're like, ah, I see. You're not very confident about this. Like, and those beautiful tells and the nuances and gestures are just like incredible to see and like the scale at which they operate at.

You guys might be incredible at poker. I'm just saying, don't rule it out in terms of bluffing people and being able to read them.

We have some people on the team who are great at poker. I don't think we fall into that group.

I don't think we do. Yeah, I think what has happened is instead we are too emotive.

Stuff.

Like we're just like, ah, we're ever, like we put all of our expressions on our face where other people are really good at detecting, but maybe not portraying them.

Yeah, exactly. Our perception team is great. Yeah, exactly.

Exactly.

I'm curious, last time we spoke, 25 people, now over 50. Is that right?

Just about. Yeah. It's been, it's been a crazy few months here. We've been really, I mean, things have really just taken off the past few months.

We've managed chaos.

Managed chaos. Exactly. We've been super fortunate. We were able to raise a Series B and bring on some, some great partners and the The team has just continued to flourish with just awesome people. The research and engineering team specifically has grown, and with that, the go-to-market team has followed, but it's been managed chaos.

There's two things I want to dive into there. Firstly, when you're hiring, do you look at hiring internationally? Because when I'm thinking about communication, I'm thinking about the different countries and cultures and how they communicate and how they use their hands and how that plays into the conversation.

I mean, our default, we hire the best person no matter where we are. And of course we have a preference for San Francisco, but basically if someone's exceptional, we'll hire them. And so diversity in our team is insane. I think this team speaks 30 languages. So honestly testing—

We have a great—

because we can test any language.

We are 13 countries where the team's based now, over 30 languages. And up until a few months ago, actually every single person on the team except me spoke multiple languages. And that was not something we screened for. That just happened.

We just, we tried like, so yeah, like hiring internationally is absolutely a part of it, but just hiring people with different perspectives and people who are incredibly opinionated is really important. That naturally then means there is diversity. And it is so important because how I gesture is so related to the culture that I was brought up in and like the societal like norms and all of these things. And so both from a team, like so that we have, we don't have like biases, but also from the data that we collect, all of those things, we have to really think about it. And it helps a lot to have very diverse team.

And we say culture as if it's like, you know, country to country, which is definitely a part of it. But even going from, you know, the Midwest to Texas to San Francisco. It's generational, right? Like Gen Z to boomer, like, you know, to the, to the, you know, education background here we have. We have high school dropouts that work here, right? All the way to people who have done PhDs and multiple degrees under that. And so it's culture in the broadest of senses, right? And striking a balance of diversity on all different ends of that spectrum, those axes, is what makes it so special. There's people who think about things from perspectives that we have never even thought about or considered or realized existed.

Beyond that diversity and beyond technical proficiency, what are you looking for in a new hire?

Passion, craft, and a want to do their life's work. Truly, someone who sees their work as a labor of love with the outputs and the products and the models and everything we're making as a testament to their craft and their passion. Like that's like the most important thing. And that has, that's been honestly like one of the biggest differentiators for Tavis is that everyone here is so passionate and in love with the problem that we're solving around human computing and teaching machines the art of how to be human. And that's the number one thing we screen for.

And I think it's two things on top of it. It's curiosity. It's a love for learning. The problems we're tackling here have never existed before, right? They're brand new. There's all this new frontier. We need to be able to learn it and be curious enough to understand what we don't know. And the other side is actually a sort of offshoot of being passionate, which is, and actually caring, which is having opinions.

Yeah, right.

And the way that that manifests is really important. We need people who are incredibly opinionated and who have strong thoughts and want what's best and will push for that.

And that means that they don't see, they don't only just come here to do their role. They see the large outcome as part of their work. And so they'll say, even if this is my domain, that isn't good enough and we need to fix that. And they're thinking beyond just the lines of their own role.

They don't come here for a role. They come here to build.

Yes, absolutely.

And they want to build the best thing ever and they will hold themselves and everyone around them accountable to that.

Yes. You're really looking for that independent thought. You're almost looking for them to come to the same conclusions independently of you, right?

And challenge you on assumptions or different conclusions and challenge us on that as well. But yeah, 100%.

What is something you believe that an employee then went against the grain and kind of turned you on the tight?

Hmm.

Ooh.

Try to think recently. Hmm. I feel like maybe Trying to think.

I actually think one of them was about how we hire. Oh yeah.

Yes.

That's one that we've been thinking about a lot lately and especially as we scale, right? How do we build these systems to hire? And you know, Hassaan and I have sort of thoughts around that because we have the value sort of system that we've created and we've seen work for the employees that we've brought on who have both worked and not worked out here. And I think one of the things that we've been challenged on recently is, hey, are we doing this as best as we can and screening as efficiently as we can? And I think our initial thought was, well, yeah, like we put so much time and thought and love into this and we spent an hour in a room, broke it down and we're like, oh my gosh, we need to reinvent how we hire.

Yeah, yeah, yeah. I think, I think from, from like I think the research team especially brings such different perspectives on all of these things. Yeah. And so, and so like, you know, let's just say like my understanding of, of like, you know, the, the best way to like capture human conversation. We had a researcher come in and said like, no, no, no, we're actually recording in the wrong way. Like, oh, you know, like actually like we shouldn't be doing it this way because we're actually introducing these biases. Like if you set someone up to say you're going to be recorded, it's like they're going to put up a completely different face. So like we need to like say that's a different way. And I was like, okay, this was amazing. And it like changed, it changed the output completely.

Interesting. I've got one more question on hiring, one question for you guys as individuals, and then we'll finish there. I'm coming for a job at Tabas. You got one question to ask me to see if I'm a good fit. What would you make that question?

Can't ask two? No. Do we each get one, or is it like we're individual?

You each get one.

Okay, okay. I'm gonna make you go first.

Ah, okay, okay. I'm trying to think what I ask, what are the top things I ask?

Well, I feel like I can't, I have the top things I ask, but if we only have one question, I have to reinvent that. This is going to be something that I might not even ask in interviews. This could be a completely different.

I think for me, maybe it's like, what do you want your legacy to be? Whatever comes to your work.

That's a good question.

What did you enjoy doing most as a child?

Interesting. I don't want to unpack them because I want to have the audience think about why you've posed them. You've experienced this before. Um, last time we interviewed, I posed the same question for every guest. I provide no further context, and I encourage you to take 10 seconds of just silence before you answer. You go first. Why do you do what you do?

Because I want to deliver the dream that we've all had of like, of like what it feels like to use a machine. Beautiful.

I forget what I answered on the last podcast, so I'm intrigued to go back and watch that at this point.

Yeah, let's cross-reference some bulls.

So I, I I'm sure it's the same or similar, but I grew up in Michigan. My grandparents immigrated from Armenia during the genocide. They came over and were working two factory jobs, didn't speak the language, were doing their best to bring my parents up and help them live good lives. My parents didn't go to college. They sent me to college for the first time. Funny enough, dropped out two years later. So that was quite the conversation. I grew up with this really deep responsibility for, a deep responsibility and responsibility's not even the right word, but want and desire to make the world a better place, right? And to really bring the vision that my grandparents had and my parents had to life and help sort of see that come to the world. And so I went into politics for a little while and stumbled into tech and, Throughout all of this, it was this consistent journey of where can I make the biggest impact on the world? Where can I make the world a better place? And turns out it wasn't actually politics. It seems like it's in startups. And Tavis is the place where we can have a ripple effect on the world and really make the world such a better place from, again, education to healthcare to how the elderly communicate with machines and have relationships and companions all the way to how youth are learning and able to structure their thoughts and organize. And so I do what I do because this is the best place and the highest leverage that I have to do that and achieve that.

Thank you so much for having me in the office, showing us around, taking the time out with us, and for, yeah, all of your time, attention, and what you're building as well.

Thanks for coming over. Glad we could have you over. Glad you could check out the office and hopefully everyone else is able to see a few parts of it too in the clips. But you're welcome back anytime. Appreciate it, guys.

Thank you.

Awesome.

Love it.

That was great.

That was great.

Thank you. Yeah.