Subscribe
Share
Share
Embed
Fascinating conversations with founders, leaders, and experts about product management, artificial intelligence (AI), user experience design, technology, and how we can create the best product experiences for users and our businesses.
Kyle (00:01.672)
All right, welcome back to another episode of Product By Design. I am Kyle. And this week, we have another awesome guest with us, Suman Kanuganti. Suman, welcome to the podcast.
Suman Kanuganti (00:13.963)
Nice to be here.
Kyle (00:24.776)
Looks like we may have frozen for a second. Okay, there we go. I think we're both back. All right. Welcome, welcome again.
Suman Kanuganti (00:26.914)
Yep, you froze for a second. Yep. Okay, great. Nice to be here, Kyle. Thanks for having me.
Kyle (00:37.968)
Awesome. Let me introduce you very, very briefly and then love for you to tell us more about yourself. But Suman is the co-founder and CEO of Personal AI, which we're going to talk a lot more about, I believe. But Suman, why don't you tell us more about yourself?
Suman Kanuganti (00:56.19)
Yeah, by background engineer, I have a robotics background as well. And, uh, for sure enough to get another degree in MBA too. But I'm most passionate about like creating brands and creating experiences that are trustworthy to humans. That's kind of how I would say.
My philosophy always has been like you don't solve the human, you solve the problems. So I always try to leverage the technology to kind of augment individual people's experiences in my previous company, AIRA, which is AIRA. It was about using AI plus AR with human in the loop for people who are blind and low vision for everyday activities that otherwise they would be having challenges with and experiences that otherwise they would not have.
So yeah, so fast forward if you, I'm sure everybody been to Starbucks, everybody has Disney, everybody has Amazon. Right now, pretty much all those companies kind of adapt to I run as a combination for people who are blind and low vision. And personal AI is sort of like similar in terms of philosophy, in terms of vision. But our goal is to essentially extend an individual person's mind
AI model that belongs to them. So yeah, so it's always about augmenting and extending individual people capabilities that are close to consumer experiences. That's what I'm passionate about. Yeah, but otherwise, I'm a technical person, but I have some business background as well.
Kyle (02:27.584)
Awesome. Well, I'm really excited to touch on a couple of things that you talked about just now, Ira, and your work there, and then what you're doing with personal AI, which from what you just said sounds very science fiction, but also very, you know, this is what's coming. But before we do that, why don't you tell us about anything that you like to do outside of work and outside of the office?
Suman Kanuganti (02:52.126)
Outside of work, I have a five-year-old daughter. So most of my time outside the work, I spend with her, and of course, with my family, wife. And we make a lot of different stories. And interestingly enough, she is very much into
you know, creating and doing like different sorts of experiments. So I go with the flow and, you know, kind of leave her world. So I do, I do enjoy it. I do enjoy that time with the family. But in terms of what I would do, even beyond that, I am a big volleyball player. I, oh, by the way, I do run. I do half marathons. This year, my goal was like two, but I don't know if I can meet that. I'm like currently around five or six. So those are some of the things that I do beyond work.
Kyle (03:39.912)
Okay, that is great. I love volleyball too. It's such a great sport. Yeah.
Suman Kanuganti (03:45.594)
Oh nice. You don't often see people who love volleyball. It's all about basketball and football.
Kyle (03:51.924)
Yeah, yeah. No, that's, yeah, that's totally true. Well, I'm interested in, you know, stepping back a little and learning more about your journey. You mentioned, you know, you being a very technical person, and then you're founding these two companies that we're going to talk more about IRA and personal AI. But tell us more about your journey. What led you to where you are now?
Suman Kanuganti (04:18.77)
Oh dear, I always think about it as...
you know, whatever I'm doing right now and what I will do next is built on the experience that I've built in the past, right? In other words, I grew up in India. I was like, you know, moderate family, pretty much like I had to struggle to get good education as well. And I'm fortunate enough to have parents who struggled hard for us, you know, put us in good schools and walked our way through, you know, getting good education.
And then I got an opportunity to kind of come to United States to do my master's in robotics back in Midwest, actually. I went to University of Missouri, Columbia. That's where I got, I got, I got like, you know, exposed to robotics and advanced engineering systems. So yeah, so generally, I think it's always a composition of, okay, you learned XYZR, you gain so much of different experiences and then what builds on top of that experience.
So I think my life is probably like a step function almost in an entirety of my life. And I think it's continuing to happen even like right now. Then I went to working at big tech corporations. Being a robotics person, I was actually programming robots at Caterpillar in Mosville, Peoria. That was so much fun. Oh my god, I miss it. In a way.
And slowly moved on to engineering systems, more of a software as well. And I went to Qualcomm, I went to Intuit. Intuit was a pivotal moment for me because I think by the time it was at a point where I grounded myself like, different, I guess, technologies, if you will, all the way from like hardware to the software, if you are a network person, like all the way from layer one to layer seven, if you will.
Suman Kanuganti (06:10.27)
And then I picked up this like code desire.
our passion when you create something and people touch it, there is something about it. It gives you energy. And that's what happened to me at Intuit because quote unquote, I was the person in a TurboTax team, right? Now, accounting quote unquote considered like boarding, but still, so many millions of people used TurboTax and had pleasant experience with actually fulfilling the taxes.
So I think that's kind of where my desire to building products and solving problems and touching people kind of came about. And I went to God, my MBA from Radio School of Management, and immediately started building companies.
Kyle (06:55.828)
Awesome. And one of those companies was Aira, which was the previous company that you founded. And I'm fascinated by this. And you mentioned it was for and is for blind or low vision individuals. So maybe tell us a little bit about what prompted that you're creating this company and what exactly is it?
Suman Kanuganti (07:11.669)
Mm-hmm.
Suman Kanuganti (07:20.37)
Yeah, so it's a combination of a dearest and closest friend of mine, his name is Matt Brock, but also being a technologist, I was pretty much trying every new technology that comes out. And also being a hardware person, I was much into any new hardware that hits the market. So this was back in the times when Google Glass came out. So I was close to the problem.
And then I saw a solution which is, oh geez, you know, there is camera that is located right in front of, you know, a wearable device, where this could be more useful and hang on, like, rather than thinking about, you know, giving it for people who can see, it can actually serve the purpose of, you know, the accessing the visual information for a blind person. Because I was close to the problem, so I think it was a natural...
kind of, you know, progression for me is like, hey, let's try this thing out. And then, yeah, I mean, this is back in like 2014, 2015.
And Google Glasses, there are all this press and issues and camera and everything else. But then I'm just going from one blind person to the other blind person, putting the glasses on. We connected to human agents behind the scenes who basically sit in front of a computer. And we created this dashboard. It's kind of fascinating. It's almost like mixed reality. So it's almost like they teleport into your world, and they're seeing for you. And the dashboard is just not the camera,
environments and the surroundings and everything else. So you get the feel of like, you know, looking left, looking right, straight ahead and the maps and all the information that is coming to you as like a remote person. And of course there is a little bit of like AI involved in terms of, you know, fast processing the information that is coming at you. So yeah, I mean, that literally was a, it changed like how like, you know, people conceptually this idea of like this visual information, like what does it mean to, you know, reading book for
Suman Kanuganti (09:26.93)
a six year old daughter that you cannot see for the first time, right? Whatever like going to the Disneyland and experiencing, you know, what does it mean like beyond the audio descriptions? Um, so yeah, we had like fascinating stories and, uh, I fell in love with the problem and, uh, I made a lot of friends in the blind community. Um, yeah. And today it's like even every airport has Iver for people who are blind.
So it's pretty widespread adopted by major, not just government, but also big tech and retails as well, including banks.
Kyle (10:01.36)
Yeah, no, that's incredible. And being close to somebody as well who has lost their sight is to take the problems and then apply the technology and all of the different things that we can in order to help in what tends to be a very visual-based world. You know.
Suman Kanuganti (10:27.922)
Mm-hmm.
Kyle (10:28.172)
offer the same type of information and experience to somebody who may not be able to necessarily navigate in some places. That can be a really, really challenging thing, like in airports or in other environments that are very... There are a lot of people, there's a lot of things happening, and being able to help navigate those types of things. I think that's just... It's a fascinating problem and something that I think most of us don't think about most of the time.
until you're potentially with somebody who is trying to navigate some of these things that aren't the familiar, they're very new and very difficult to go through.
Suman Kanuganti (11:07.646)
Yep, totally.
Kyle (11:11.272)
So I'm interested then because you founded and you built Aira and then moved to your next company, which is Personal AI. So tell us what made you move from what you were doing into Personal AI and what exactly is Personal AI?
Suman Kanuganti (11:32.018)
What made me move? So the genesis story for personal AI kind of ties back to Aira as well, because at the core, the inspiration for personal AI is my desire to have a conversation, both emotionally and intellectually, with a last co-founder as well as investor of mine. His name is Larry Bok of my previous company, Aira. And he created around 27 different companies.
Suman Kanuganti (12:03.376)
And we as human beings, we generally tend to learn from different places. Internet obviously is one of those places, but we also learn a lot from our professors and the people around us as well. The interesting thing about learning from individual, like humans, is this element of trust, this element of connection that you carry.
that you probably sometimes do not get, even though you're taking long periods of time to do the research that is available on the internet. How do you negotiate? How do you think about strategies? How do you think about experiments? He taught me a lot. So then the question becomes, oh geez, like, you know, he passed away and I cannot remotely remember everything that he told me. Or, you know, I still want to have a brainstorming session with him.
I wish I had Larry's AI during my time building Aira because he passed away within 18 months of building it together. So it was in the back of the mind for a long period of time until the board and myself kind of came together and decided to bring in a new CEO to grow the company, Aira. That gave me a good opportunity to step into my next.
passion that's been building for almost two years, right? And the technology was also right about in the beginning stages in terms of AI. And luckily given, I was also like technologically pretty, you know, heavy that includes like, you know, wearables, hardware, human in the loop, AI techniques. I carried pretty much all the things that I learned, you know, putting together the business and putting together.
the team for personal AI. So what is personal AI? As I was mentioning, it's as simple as thinking about an extension of an individual that you can have a conversation with, right? It's an extension of somebody else's mind.
Suman Kanuganti (14:05.83)
In general, the most common modality of people learning from each other is, of course, the computers and the other is human beings. But when it comes down to humans, it's about communication. It's about talking to each other. So we wanted a system that is so grounded in an individual, that is grounded in their facts, that is grounded in their opinions, that is grounded in their, you know, ways of thinking.
When I found Sharon Zung, who is my current CDO, she basically said, hey, we cannot use large language models. This is back in 2020, right? So even then, you know, it was cutting edge research. We cannot use it for so and so reasons, which is, you know, hallucinations and, you know, fast forward the problems we know right now. And she said we had to take an alternate approach, which is building small models. So we took a contrary approach to the large language models. We actually started building small language models.
that is grounded in one person's memory. So what is personal AI? Personal AI is a model that basically mimics an individual person that is grounded in their facts and opinions in their own voice that is backed by the memory that belongs to them.
Kyle (15:18.952)
That's super fascinating. And I'm interested in how it works because you describe personal AI as an extension of yourself that can interact with anyone in your own voice and your own style, with kind of like you're saying, trained to an individual, which again, if we think almost science fiction, this feels like the progress towards what we would have considered to be
an absolutely crazy idea, you know, a decade ago or even a few years ago, but this idea that you can extend yourself.
Suman Kanuganti (15:56.187)
Yep. I mean, even two years ago, I had hard time convincing, you know, even technical people and investors like this is going to be happening. Well, here we are. AI is everywhere right now.
Kyle (16:11.312)
It is. And so tell us more about how does that work? How would I extend myself with a personal AI? So not necessarily a large language model or something trained in just general knowledge, but something that is very specific to me.
Suman Kanuganti (16:27.982)
Yeah. So I would describe the components, and then I will describe the application that you would use to get there. So at the simplest level, right, you should be thinking about personal AI as, if you have experienced Alexa, right, I'm sure most of the people are, or even some form of smart speakers, you know, we say, hey Alexa, and ask a question, and basically it's a query to the internet, and sometimes backed by large language models as well.
Think about this one as Kyle AI, like, you know, hey, Kyle, right, and then I'm talking to you, right? So the experience is pretty much that's form of conversation, that's form of communication, except that it's not a general intelligence as much as it is an individual personal intelligence, right? So that's the basics. How it works from a...
technological perspective is that every person has a memory. What does a memory mean? Think about every minute, if you would wish, capturing a piece of information that you have said, that you have captured, right? And it's time bound. And over a period of time, as you learn as a student, as you grow in your career, as you become trailblazer like in the future, influencing other people, you have all these memories that are formed.
So that is essentially the data that basically is then used to train your unique AI language model that anybody can have a conversation with. So how can anybody use it today? There is a mobile application, as well as there is a desktop application. And people can download, and they can choose to add any of the historical memories that they would want.
And they could also choose to add memories on an ongoing basis. So that will continuously improve or increase the quality of your model, as well as the synthesis of your own mind. In terms of use, the application will also support being able to invite other people to chat with you.
Suman Kanuganti (18:35.471)
with your AI in the loop. So imagine text messaging experiences or WhatsApp kind of experiences, right? Where Suman sends a message to Kyle. That's Kyle, you know, I'm coming down to Utah. What are some of the favorite places to eat? I'm with my wife and kids, so I need a family-friendly one.
Well, I could go look up in the internet or I could like simply text message you and on your side Your AI since it already knows you if you have said that to somebody else before I added to your memory. It will automatically draft message for you Right and that you can choose still choose to review and yeah that sounds here is the place that I would go
and you would simply swipe right to send that message. So there is a applicability in terms of how simple it is to be able to train with this core mechanism of texting, which is the communication medium that we use almost every day basis integrated. And as you communicate with more people, as you communicate with yourself more, the model just gets more promising.
And the use of your own model in multiple different places also gets more prominent.
Kyle (19:47.024)
That's super fascinating. And I'm interested, what is it, what are some of the training data that you use now? You mentioned some of the communications and what do you potentially see that expanding to in the future? Because I can imagine a scenario where if I'm training a personal AI, it has potential access to literally everything. And I assume that's not the case right now, but this idea of if I want it to...
Suman Kanuganti (19:56.398)
Sure.
Kyle (20:14.78)
know me as well as possible and be able to reflect all of these things to, you know, the broader the data set and the broader, you know, whether it's just about anything, you know, the more opportunity you have for that. So what does that look like now and what will that look like in the future?
Suman Kanuganti (20:23.182)
Sure. Yep. Yeah.
Yeah, sure. So you can think about personal data.
to train your own model as pretty much anything that you have ever authored historically. So that could mean that you know there are autobiographical fees that you have written, the emails that you have sent, maybe the historical text messages that you have sent, maybe there are documents in your google drive that you have written, maybe blocks that you have written, maybe podcasts that you have spoken, right. So it can pretty much be any author
content on your behalf if the purpose of your AI model is indeed to start kind of almost reflecting and replicating you know your mind. Currently the system does support you know ingesting the data in the native format which is you know information directly fed into your model through the personal AI apps.
You can also feed in the URLs. So let's say if there is a blog or if there's a YouTube video, you can post those in as well.
Suman Kanuganti (21:35.926)
you can integrate into your existing Google Drive accounts. And if you have documents that are authored by you, you can use those as well. And over a period of time, we intend to add even more easier mechanisms to sync your historical communications. Now, the most interesting thing is with personal AI, we give the ownership of the data to the people.
We give the ownership of the model to them as well. What that means is the company will not or cannot use the data for ads business at all. Zero chances because we remove ourselves out of the equation too. So yeah, so from a data perspective.
It's anything and everything, all the personal data that you would wish. And right now there are automated mechanisms and there are some manager mechanisms and it will only get, you know, more and more easier over a period of time in the company progress.
Kyle (22:38.2)
Well, that's really, really incredible. And I'm interested too in, what do you see as the applications of this? You mentioned some of the texting and communication. What are people using this for now? And how do you see that expanding in the future?
Suman Kanuganti (22:55.994)
Yeah, yeah. So that we think about like two modes, right? One, you can think about it as an alternative communication channel, right? Which is if you care about privacy, if you care about every communication that you're having with other people to work in your favor.
I have your AI kind of help you draft these messages automatically. You can, you know, essentially choose to think about personally as an alternate communication mechanism.
Now, the most beautiful thing is it's not just a human-to-human communication. You can choose to go human-to-AI communication as well. Let's say I would want to communicate with my coach and she probably is not available, but I still want to have a conversation while they are in airplane. They can set them, instead of airplane, more autopilot, more we call them. So I'm having a direct access to their mind talking to their AI. So that's like one way to think about it.
is like an alternate communication medium, then what you may be using at work or in personal life, such as WhatsApp or Telegram or Slack. The other form that we also see a lot of more users coming in is more like Omni communication media.
In other words, if you indeed are in existing communication such as WhatsApp, such as text messages, such as websites or email, right? You can choose to have personal AI draft those messages for you in those platforms as well. Where you still have the ability to kind of control and approve the AI messages, but at the same time...
Suman Kanuganti (24:46.706)
It kind of goes into multiple different channels, including social media. For example, for business people, they would want to have their AIs automatically representing their brand in Facebook Messenger. But still, the control is within that individual person.
Kyle (25:03.896)
Awesome. It's interesting because this is the kind of thing that my wife and I were just talking about recently was, you know, hey, need to step away for a minute. Why don't you, she's like, why don't you just turn on an AI to respond to some of the messages while you step away from your desk for a minute? And I laughed and I was like, that's exactly what I need is to be able to step away.
for a minute and if anybody needs something, like they're still able to get answers, but I don't have to constantly be, not that I am constantly monitoring it, but I don't have to be, which I think is a fascinating thing.
Suman Kanuganti (25:43.354)
It's magical, Kyle. It's actually so good because right now I'm living in it and I know my team is continuously using my own AI that could vary from brainstorming with your AI or writing specific pieces of blogs or maybe replying to somebody else in the email or writing a byline to a Wall Street Journal article as a response. So like, look, these people are asking about these interview questions and the ethics and privacy. How would I respond? Oh my God.
how many things that happen in parallel and believe it or not once you are in the personal AI ecosystem it's not about talking to the bots anymore it's about still talking to people but they are getting help a little bit more from their own AI and there is that identity associated with it it's not somebody else's chatbot right
It's not somebody trained that you actually control the idea of this training. It's actually yours. It's representing you, right? And you know what data went in there, right? That's the level of control that you have with this model. And it's not dependent on the large language model. So your data is not going to any big tech as well. So, oh my God, I think it's just a matter of like people getting to really understand the, like the depth associated with.
you know, what we built. And I think like, you know, some of the existing community people, once they're in it, they love it. It's that there's that emotional attachment because it's not a service, it's an asset. It's an asset that you build for your life and there is no stopping to it. And it will live even when somebody leaves the world, right? Like you can still have these conversations with them. That's the beauty of it.
Kyle (27:39.236)
Yeah, and that I think that's another fascinating part because you mentioned it earlier as part of the genesis of it. And I see more and more people looking at that sort of thing, being able to not to have some that continued connection where it doesn't necessarily have to mean that somebody is gone if you're able to still have, like you said, a brainstorming session or
or a conversation very similar to what you would have. And to me, this is, it's kind of a new area, but an absolutely fascinating one that I think we're starting to see more and more of. Have you seen that as part of building personal AI and what has been some of the things that you've seen regarding that specifically?
Suman Kanuganti (28:29.906)
Yeah, very good question. So the uptose that we have taken is given it is so broad, we let the market decide who are the ones who has appetite for this.
So what we've done is, like late last year before charge GPT came about, we started taking some beta applicants to kind of see what their motivations are and build their models and in the learning process, we built our applications and the products as well. Because first two years of the company, it was all about building the tech, right? Building the data models and models and how do you scale the models, et cetera, et cetera.
And since April, so April is when we went like GA, in other words, like anybody can come in and like train it and use it for themselves. And we saw all kinds of use cases. We definitely saw all kinds of use cases and people's expectations are all over the board as well.
In a way, it was intentional too, because we wanted to see what kinds of needs exist out there given this platform and what can people do. And some people are like, it's personal AI and steroids right now. Like they're basically like 80% of the job is done by their personal AI in terms of communications, automations, drafting and everything else.
Some people probably never have invested enough time to actually train the models. So we invested some.
Suman Kanuganti (30:01.094)
easier mechanisms to make it more effective, even with one or two memories. So long story short, we do see a lot of people building their legacy with their AI, one of the original applications. So that is still happening. We are seeing a lot more appetite for small businesses, small business owners specifically, who probably are dealing with multiple different clients or communities or customers and people around them. Has a lot more need, if you will.
for automating some of their daily communication needs. Either it be writing email, writing tweets, generating blogs, text messages, responding to clients, all those things, and you have a mechanism to attach a AI phone number to your AI too. So it's almost like, hey Kyle, you have a phone number, your AI has a phone number too. So you can technically think of it like routing, and you can get creative there. So yeah, so those are some of the things that we are seeing,
And one thing that seems to be pretty consistent across all the people who are building personal AI is their moat and their desire is their personal data, is their personal brand. You know, their individual themselves is their currency. And they don't want to submit the data, right? They want to use the AI because that's the next generation of optimizing your work life as well as personal life.
they care about privacy and how these models are built and how much control I have, which all suits to what we are in the first place, principally, right? We want to give that control to the users, you know, own your data, get your model to work for yourself. It's a small model, it scales horizontally. No data is shared between any two models. No memory is shared between any two memory stacks, we call them.
So yeah, so I think that's kind of where, you know, my energy is spent lately as well, like just telling out and understanding, making market understand the differences between what it says, other players, and what personal AI is, and you know, which are optimal for what needs. Personally, obviously, we're not solving everything, but we have a place where people care about personal data and personal models.
Kyle (32:23.752)
Yeah, I'm interested because you mentioned the idea of being able to talk as well. And we continue to see just an absolute some incredible advancements in some of the voice and sound applications of it. How are you incorporating that where maybe it's not just written or textual, but also being the spoken part of AI?
Suman Kanuganti (32:39.887)
Yep.
Suman Kanuganti (32:52.042)
Yeah, so we have. So it's a matter of weeks, if not days, that will be available in the Personal AI app. And not just any voice, it will be your voice. I mean, it has to be personal. Come on. Yeah.
Kyle (33:05.692)
Yeah, absolutely. And I'm interested in, where do you think that goes? Cause obviously we're probably pretty early in it, but we continue to see a lot of musicians and things like that where you can replicate the voice very, very well. And some of the earlier ones are not as good, but now it's getting to the point where it's very, very good. How...
How have you seen that advance, and where do you think that's going?
Suman Kanuganti (33:38.142)
Yeah, I think the issue that we are probably stressing on is not just about voice or mind or visuals, right? Like generally the deep fakes existed for a long period of time, even in the celebrity world. And this problem existed all the way since the images become a thing like on the internet, right? So I think generally the...
companies who are building AI technologies now because it poses a greater risk to the identity and how the government kind of plays the role in protecting the identity, I think it'll become important.
In other words, internet give as a good medium to communicate with each other and exchange content and somebody produces the content and other people are consuming the content. And so far we have figured out what the transaction looks like and there is some giving and there is some taking. Now with AI, it's almost like a 2.0 version of the internet or even web 3.0 kind of like falls into that realm as well with the AI.
So I think protecting the identity and what kind of transaction need to exist needs to mature. And my message to the people that I speak to is you cannot just wait for the lot to kind of step in and figure out who took your data to train their models. As much as you have to step forward and take control of your own data, right? Build your personal models so that way it can start benefiting you and the people around you and your business. And you know, your family.
or friends. Otherwise it will be another social media. But at the same time like with every technology step up the maturation about you know some of these issues I guess will emerge and will be fixed. So I think it's natural progression to how
Suman Kanuganti (35:35.594)
the technology kind of penetrates into our lives. And we have seen that in the history and the history kind of repeats, but in a different form and in different modalities and different fidelities.
Kyle (35:47.536)
Yeah. What have been some of the things that you've seen as you may be some of the biggest hesitations that people have? I know you've talked about owning the data, which is probably top of a lot of people's minds. Are there other things that as you're exploring this and talking to people about it that you've seen either some hesitation with or some of the biggest concerns with?
Suman Kanuganti (36:11.302)
Ownership of the data and the control, I think are the two big pieces, right?
Suman Kanuganti (36:19.618)
So when, I mean now, okay, what does the ownership mean? What does control mean? You can kind of double click on what is exactly meaning the experience.
Let's just say if you, like let's just say back, like 10 years ago, what you did, like when we had Google search and Bing searches, et cetera, you would go to Google.com, you would search and you would find that information is available on Kyle's website or Kyle's podcast. So I will go to Kyle's website and then consume that information. Now, if Kyle puts it publicly, great, I will read it. If Kyle has a paywall, then I would have to pay and then read it, read that information.
But technically, we are routed back to Kyle's work. We are routed back to an organization work. We are routed back to a specific business. That, in the AI world, is attribution. So in other words, when you ask a question or when you push a prompt to a model, what is the data, underlying data, that is supporting it? And who?
is responsible for that data. And right now, large language models are not good at attribution of the data.
So I think that's one of the important things. And I think personal AI by design is associated with an individual, right? And we do have the mechanisms of attributing the data, the response back to the person. And it's very fascinating because we do have like community of people. So we have this concept of like groups or launches we call them. It's like AI launch.
Suman Kanuganti (37:55.69)
where a group of people can come in and then contribute on a specific topic. Let's say you are super passionate about volleyball, get your like 20 volleyball fans and then like basically train a model.
And then when you ask a question, basically the response will have, hey, these are the contributors potentially for this response. So that is almost like the next level of what model release that's gonna happen. But yeah, attribution I think is really important. And just ensuring it's not just one size fits all, right? Ensuring a little bit of payback, a little bit of...
value democratization that is economically, you know, kind of beneficial to the people involved. So I think that's important. It's not just about technology, so I can take your data and it's available publicly so I can give it out. It's kind of a pathetic world in a way. So yeah, so I think, you know, as responsible human beings, we get to do, you know, what we think is the right thing. And I think personally, I has a lot of those characteristics and the future may be
you know, billions of personal AIs and a few, you know, big models, they all work in conjunction. That's also fine. Yeah.
Kyle (39:14.712)
Yeah, I'm interested in diving into that just a little bit more. How do you see that playing out as the future of personal AI and the future of more personalized AI in general? We're obviously seeing the beginnings of it, but how do you see that going forward?
Suman Kanuganti (39:37.714)
Yeah, so I think there are use cases for every, different forms of models, right? So there's one general models, obviously that's large language models that exist today. And there are at least, I can count 10 to 15, if not more players that are happening. And some of them will survive, some of them will stop. And I'm sure some of the big techs will survive. And...
there are then personalized AIs which I would still define it slightly different than personal AI. personalized AIs are you have a base intelligence let's just say food
There may be a collection of food intelligence, and based on who you are and based on what your preference is, it may suggest specific things for you. That I would consider personalized AIs. It's almost like still foundationally built on an aggregated model, but because there is an input, then you have some personalized experiences.
Personal AI, our definition of personal AI is a completely separate, like unique model that is like only or solely built on a person, right? It doesn't know like the other things. It may consult with the other models to give you an opinion, but fundamentally the model is trained on individual persons, that way we don't conflate, like, you know, who is Kyle versus who is Suma?
So I believe there will be some sort of a exchange or transactions that will likely happen in terms of like negotiation, if you will, between a personal AI and potentially a personalized experience. So imagine this, right? Airbnb. Suman wants to book an Airbnb in Santa Barbara.
Suman Kanuganti (41:41.898)
Right now I have to go to Airbnb website and I have to tell the locations, the preferences, and I have to search. And then ultimately, Airbnb AI engine is providing me some recommendations and I'm making the decision.
So right now, I'm giving the data to Airbnb to kind of tell me or give me that personalized experience. Well, in the future, imagine you open Airbnb app, and Airbnb will consult with your personal AI and do the negotiation and give you what recommendation it is, right? Same thing with Uber. Like you open Uber app right now, and you are sending your location to Uber's servers. What if that is flipped 180 degrees? Uber will make a call to your personal AI.
to give you the ultra personal ride experience based on, you know, that current situation.
Suman Kanuganti (42:32.118)
So I think there is a good trajectory in here and I think it will take time.
Kyle (42:32.456)
dads.
Kyle (42:37.252)
Yeah, absolutely. I'm interested in what is it like going to market in the AI space right now? You've been working on this pre- explosion of AI and LLM specifically, but what is it like you're taking some of these things to market right now in
as there's lots of things happening, the space is continually changing and it's becoming very, very it's a hot growing space right now. Yeah. What is it like going to market both with the company and the products and some of the new features that you're coming out with?
Suman Kanuganti (43:12.462)
counted.
Suman Kanuganti (43:23.174)
Yeah, it's very fascinating for me building this company, primarily because we went from, I don't know what you're talking about, to everybody's talking about the same thing. I mean, there is a reason why we are personal.ai, right? We set out to build a category, and then the category just happened.
Just like that, it happened. So I think going to the market right now is about educating the market, right? The differences of the market. The interesting thing about technology is like, once you have a core foundational technology that is built, you have a whole number of different experiences and applications that you can build on top of it, right? And then you can unlock those experiences. It's like internet. Once we have the internet, we have.com. And when we have, once we have the protocols of, you know.
standards of HTTP and HTML like boom you have like millions of websites then you don't know do I go to Amazon to buy you know there are like million other shopping websites do I go to other places to buy so it will take time to kind of mature like what brands to sustain and what brands to trust and what is the difference between you know one versus other and for me my focus is trust and control like I want to solve for people and humans first
There is tons of money that is getting poured into the AI industry. And I perceive that as like, you know, some people who are powerful can also just like simply buying the market, right? Well, as a person that is grounded, building the startup from the ground up, I would want to win the market, right?
and do the right things because it's very easy to get tempted and take shortcuts and give people something that you will repeat in the future and I don't want to do that.
Suman Kanuganti (45:22.43)
So yeah, so going to the market right now for us is like clearly articulating what does it mean to get a personalized experience versus build something of your own with personal AI versus what are the use cases that a large language model will solve and what are the use cases that we are focusing on and let them make choice, right? Let people make the choice. And it isn't like, you know, everybody...
uh is like one winner as much as there will be like multiple players and multiple winners uh and multiple different people solving on different specific uh use cases and issues and you know for personal ai i just want personal ai to be an asset for every individual person on the planet because it belongs to them it's almost like you know you are a growing second you alongside like learning and you know growing in your career for yourself and that will be an asset for your
Kyle (46:19.012)
Yeah, I think that's absolutely fascinating to think about. And I am excited to see some of this new direction and some of the possibility, because I think it's just it is so immense. And I feel like we're just at the very, very beginning of a lot of this. Like, what can we do with a lot of these things that are coming to fruition right now? I'm interested from your side. You mentioned, you know,
coming to market and being a founder and you know have founded this your second company What have been some of the lessons that you've learned broadly? You know founding multiple companies and you know helping to grow them and scale them up
Suman Kanuganti (47:03.586)
I mean I would say stick to the problems that you are focused on solving. Honestly I'm sure like everybody has like multiple different distractions right and when you have big players making moves and you know pouring in money and doing like tons of marketing as well you do get intimidated you know sometimes
Suman Kanuganti (47:30.814)
I mean, for example, right? Two years ago, personal AI wasn't a thing on Google or anywhere else. Well, guess what? Now everybody is building a personal AI. Literally, every company has some version of personal AI. So it's a big deal.
I think the most important thing is focus, dedication, vision, and being close to your customers and focusing on the problem that you are at and not get carried too much from the noise that tends to happen. And by the way, I'm saying it, but even I have a very hard time being able to not get distracted on other things. Now, there are two ways to think about it. One, you got to keep up with what's happening in the industry.
And you got to filter out and use that information to make the appropriate decisions. But it's not about making a nature of reaction and then changing the company's direction overnight because something else happened. So yeah, it's a delicate balance and I think being grounded and being very thoughtful and curious about what's happening in the market and what does it mean to you. When do you make the most?
Appropriately, I think all goes into like, you know, the learning and making progress towards the dreams that you envision versus other people envision.
Kyle (49:04.144)
Yeah, I think that's absolutely great. What advice would you give to somebody who is thinking about either, you know, founding a company or building an AI driven company or an AI product or something like that?
Suman Kanuganti (49:20.298)
I think probably two things. I think one, be grounded. Meaning there is ups and there is downs and there is excitements and there is issues, right? And we can only do so many things at times and sometimes, you know, a lot of things can happen very fast because of like other variables involved. But I think staying grounded and staying present, I think is really important.
Suman Kanuganti (49:48.322)
And within the context of AI, I would say, try to understand what exactly is the moat that you have versus other companies before you jump into building it. For example, if you're leveraging existing technologies, that's great. But what is your moat? Is the moat your connections with a specific market?
or you have a community and you have a you know specific ways of penetrating into the community whatever that may be and maybe it's the data that you collect I don't know but I think like trying to understand what gives you differentiation before it's too late I think is important
Kyle (50:36.092)
Yeah, I think that's really, really good advice to understand, to one, be grounded, and then two, to understand the market and your company and what are some of the differentiators, like you said, super, super critical. It's easy, and I think we've probably all seen this, it's easy to have a really great idea, but it can be just that, a great idea if it doesn't have some of these other things like the focus to...
to do it and then the, some of the good reason, the value that it will deliver because ultimately if it doesn't then it probably should just stay a good idea.
Kyle (51:18.348)
Well, Suman, this has been a really, really great conversation. We've touched on a lot of different things, but is there anything that we touched on or didn't get a chance to touch on that you wanted to add?
Suman Kanuganti (51:32.138)
I think we touched on a lot. All I would say is, if you're not already, start building your own model, put it in your day-to-day use. It almost gets too magical once you invest a little bit of time in it. Yeah, I would want to see a world where everybody takes control of their own lives and own models.
Kyle (51:54.172)
Absolutely. I couldn't agree more. Well, I've got a couple of questions to wrap up, but before we do that, where can people find out more about you, about personal AI and the things that you're working on?
Suman Kanuganti (52:00.791)
Mm-hmm.
Suman Kanuganti (52:08.382)
About personal AI is pretty easy. Personal.ai. You wouldn't miss it. You just Google personal AI and it will be the first one. Specifically, if you want to communicate with me, seriously, go to s.personal.ai and you can chat with me, the human version, but you can also get a taste of my AI version as well. It knows a lot about me.
Yeah, but I'm on Twitter and LinkedIn channels. I do not tend to do messaging on any of the social channels because I know my data is not serving up for my benefit. So yeah, so all my communication happens in personal AI.
Kyle (52:49.285)
Okay, great. Well, we'll put the links in the show notes as well so you can check that out. But do have, like I said, a couple of questions to wrap up. And the first one is have you read or watched or listened to anything that you found particularly interesting? And this can be business or AI related, but it doesn't have to be.
Suman Kanuganti (53:10.562)
Uh, um.
So speaking of the whole ethics and how you would want to build a company for what reasons, sometimes you may have too many different shortcuts and stay grounded. I think it's very interesting because I'm enjoying the painkiller TV series right now.
You know, me and my wife has some time at night where we sit together and kind of, you know, enjoy some time together and we watch TV and typically like one of the shows. So I think it definitely gives like a little bit like perspective on what you should not do, but at the same time, like how do you carry the passion and vision for making it happen? So I think there's a delicate balance that you will have to do. Yeah, sure. That's like, you know.
too real because like people, but AI is no different in many forms of fashion to a degree, right? So being able to, you know, do things like a little bit more ethically and you know, for human, I think is important and it's hard.
So that's like a TV show, but I enjoy a lot of science fiction for sure. Like, so if you haven't watched Altered Carbon, it's an old series, but a lot of the concepts of personal AI such as like memory stack, you know, a memory that basically remembers you forever.
Suman Kanuganti (54:44.566)
the word memory stack actually came from Altered Carbon TV series. So I can go on and on, I have many of such things. My favorite books is like, you know, Homo Sapiens from by Yuval Harari, I really like it. Like how like the past kind of influences the future. I think there's a lot to it. So yeah, so there's some things I would say.
Kyle (55:05.2)
Okay, great. Yeah, I think those are some great, great recommendations. And finally, are there any products that you've been using that you've been enjoying recently, either digital products or physical products, anything?
Suman Kanuganti (55:20.13)
Digital products or physical products? Digital products, you know, I was never a big designer as much, but lately I've been doing some wireframes in Miro, so I think it's kind of cool, especially for like product people, because before I go to Figma and do high-frequency comms, which like takes a lot of time, when it is like simply wireframing, I think.
you know, Miro gives you like a lot more flexibility. So I'm enjoying that. Other products, I mean mid-journey and stability AI for image generation, trying to figure out how it incorporates. You know, I got better at prompt engineering generally with language models. I think I still have some ways to go getting prompt engineering right for the image generation.
But eventually that will be part of first layer as well. So I'm enjoying that. Physical product, oh dear. I don't know if I can, oh no, nevermind. I don't know if I have anything that is super fascinating lately that I can think of, yeah.
Kyle (56:37.932)
Okay. Well, yeah, those are some great product recommendations and shout outs to those. All right. Well, Simon, this has been, again, an amazing conversation. Appreciate all of your insight and am super excited to see what the future holds for a lot of personal AI and kind of some of the things that we talked about in really people taking control of their data, their AI in a way that I think is
Like we said a little bit ago would have been science fiction and totally out there, but is coming quicker and quicker. And I just, I think this is a fascinating space.
Suman Kanuganti (57:16.686)
Thank you guys, I had fun chatting with you.
Kyle (57:19.892)
Great. And thank you everyone for listening. We'll see you again next time.