The Democratization of Artificial Intelligence

Tullio Siragusa (00:11):

Good day, everyone. Welcome back to Tech Leaders Unplugged. I'm Tullio Siragusa, your host today. I am joined by Dmitry Shapiro at YouAi. Hi, Dmitry, welcome to the show. Looking forward to speaking with you today.

Dmitry Shapiro (00:25):

Thanks so much for having me.

Tullio Siragusa (00:28):

So we're talking about the democratization of ai, and particularly how expanding access to personalized AI will revolutionize everything. This is going to be a really interesting topic, especially as we talk about making it useful for the individual. We've had a lot of conversations on how it impacts businesses and, and changes and disrupts the technology industry, but I want to dig in and see how YouAi is helping to democratize access to it. But before we do that, let's talk about how you got here. Let's get unplugged today. Dmitry, tell us how you got here and tell us a little bit about your company.

Dmitry Shapiro (01:13):

In 1984, I went to the movie theater with my dad and saw a movie called War Games. And as I left the movie, I was like, I need a computer. And we were poor. We couldn't afford a computer, but my school, my high school had a couple of Apple, two computers. And so I spent the rest of my high school years from 14 onwards in that computer lab. And then went to college and got a degree in electrical engineering. I've never done a day of electrical engineering, have always done software. And then, you know, had a long career now including being the chief technology officer of MySpace Music. Spent four years at Google on the product side, actually working with three machine learning teams. From 2012 to 2016. I built two well now three venture-backed companies. The first one was called Iconix Systems, which was a cybersecurity company that I built from 2000 to 2005. Raised 34 million for that in venture capital. Then I built a competitor to YouTube called Veoh from 2005 to 2010, raising 70 million for that. And then went to MySpace, went to Google, and then left Google in 2016 to found the company that is now YouAi. And here we've raised 36 million so far. So I'm an old nerd

Tullio Siragusa (02:37):

All right. Well, you keep, you keep at it. Tell us about what YouAI is all about.

Dmitry Shapiro (02:45):

Look, I think it's clear to you know, many people that we are at an inflection point in how we're able to engage with information technology and this new powerful set of technologies that we have access to. You know, these large language models or these diffusion models for image generation, basically these generative AI technologies are extraordinarily powerful. But the interfaces that were given to be able to sort of wield their power is, you know, as a command prompt is a, is a text box and It's, it's like DOS it's, no, no, I was going to say that. You got the words out of my mouth. It's like DOS for those of us that are old enough to remember that. And, you know, most people have a really hard time engaging with technology through command lines. And of course, these days we're all on mobile, and so, you know, we've got two fingers that we're using or some of us, you know, with the swipe one finger. And so that's, that's one of the problems is, is like, how do we sort of leverage all of the power of this information technology given an interface of a, of a text box. The other thing that I think many people may not realize yet is that, let's take ChatGPT or Bard, like any of these large language models, right? They've been trained on all of the publicly available content that's in the world. And so you could sort of say they know everything about, you know, the world at large or the universe at large, let's call it, right? The output of humans. Everything we've ever written you know, they've basically digested and that's amazing. And so you can say to ChatGPT tell me about astrophysics and it'll tell you a bunch about astrophysics, but who is it that it's telling it to? Meaning are you an eighth grader? Are you a college student? Are you someone that works in astrophysics? Meaning are you an astrophysicist? It does not know who it is it's responding to. And so it gives you basically generic responses, but you can actually tell these large language models, again, via these prompts. You can say, let me tell you a bit about myself and then engage with it. And they can take that into account so you can inject context into these large language models and have them behave, you know, quite radically differently than if they have no context about what it is they're talking about or whom they're talking to. But the problem is like, how do you do that? What do you just start typing? Let me tell you a little bit about myself. And then again, you're sort of doing this. That's one problem. Another problem is they have no long-term memory, so they forget things after the sessions. You'd have to do that every session. So basically YouAI is addressing those problems by one creating interfaces that are not just textual interfaces in order to be able to engage with these generative models. Also, think like Windows as compared to DOS. And the other one is allowing sort of anyone, you know, non-technical people to be able to create new ais to basically fine-tune existing AI systems and create AI for various bespoke needs, whether those be enterprise needs or consumer needs or anything of the sort. And of course, people can check it out if they just go to YouAi.

Tullio Siragusa (06:31):

Let me see if I can unpack a little bit of what I've heard. So the current models, you cannot train the current model. You can only give it input and context so that you're informing it within what context you want the data back. And those who know how to use it well seem to do a better job at personalizing the output. But what you're suggesting is that by democratizing this, I actually can inform the model something differently. And through that human-to-machine interaction, which is missing right now, it's, it's really just a, a query, let's call it what it is. And it gives you back input based on what you've asked it. What you're suggesting is that we could actually co-create a new ai with the user. Is that, is that the thinking, and how does that come to play?

Dmitry Shapiro (07:27):

Not, not exactly.

Tullio Siragusa (07:29):

Maybe that's a feature state thing I just asked.

Dmitry Shapiro (07:33):

No, no. Look, it makes sense, but let me clarify it a little bit. So we have a, we have a, basically an integrated development environment, an IDE called Mind Studio. And Mind Studio allows people, again, even non-technical people, to be able to create new front-end interfaces for the back ends that are large language models and take those front-end interfaces and make them contextually relevant. Meaning configure a preamble as it's known in these models to be able to inject via API into ChatGPT and Bard and Cloud and all of these, you know, large language models. And to be able to create custom applications that are sort of leveraging these generative models on the backend. But the front ends are specifically meant for any kind of use case. And so today, if you go to our website, you will see, you know, AI that helps you do financial planning or help you create blog posts that will automatically create blog posts for you based on your writing style, you know, based on whatever audience you specify that, that therefore, and, and these AI, basically when you engage with 'em, they collect all the information from you human about what they need to be able to then fine tune, you know, at runtime y your experience with these backend large language models. Meaning today, again, if you just sort of show up to ChatGPTand say, write me a blog post, a 500-word blog post about how AI is going to transform medicine, it's going to spit something out for you. But the thing is, it doesn't know what writing style you want it in, so you'd have to type that in. It doesn't sort of know who your influences are. It, it doesn't have a sample of your writing style to be able to write it. It doesn't know who it is that you're writing this for. Is it for general consumers or is it for physicians? And so, for example, those, you know, blog writing AI that we have, the blog generator, blog post generator gets all of that input from humans. And then, you know, as a backend service calls out via API too, I think that one uses Cloud, which is from a company called Anthropic and, and then brings back a, a blog post that is written in your style because it's digested your writing style. Because you've given it some of that, it's written it for the audience that you specified because you were able to specify and, and, and that AI was created by a non-technical person in a matter of a few minutes. Because again, all the hard work is really being done right by Cloud in this case.

Tullio Siragusa (10:30):

So how does that work? Yeah, I mean, it sounds like you've it reminds me a little bit, I'm thinking back on your experience in MySpace, you know, where when you first joined it has certain preferences or what you like, what you dislike, et cetera. And a lot of social networks are set up that way when you first join, or most you know, kind of gamified learning who you are is, is it some little concept, some little principle where when I get in there, I get to sort of go through the initial learning process of letting them know what my preferences are, how did it, how does it learn my, my writing style, for example, do I load content or I put a URL to some of my content? How is that how is that working? Is that fairly mature?

Dmitry Shapiro (11:15):

Exactly. Yeah. You're, you're describing accurately, it gives you a number, we call 'em prompts where it prompts you for answers. So it gives you basically interfaces and say, give me answers. And so the blog post generator one of the things it asks for is for you to either upload or, or copy and paste into it a sample of your writing up to 50 pages of your writing as much as you want to give it. And then it instantly parses that and, and sort of understands your writing style as opposed to my writing style. That's one of the things it does, it asks you again who's the audience for whom you would like it to generate a blog post, how many words the blog post should be, and what the general topics are of the blog post. And, then it generates the blog post using again, the power of in this case [ INAUDIBLE] clause.

Tullio Siragusa (12:10):

Let's talk a little bit beyond the blog. Where are some of the other applications for this? Can give us…?

Dmitry Shapiro (12:17):

We, I mean there are countless applications anything you can imagine that can be built for any team or any enterprise or any consumer things. There are already, you know, sort of hundreds of them that are YouAi.ai. There are things that are for entertainment purposes. There are things that help you, again, content generators of various types. There are things that are summarizers that will, you know, these large language models are very good at being able to take a lot of content and summarize.

Tullio Siragusa (12:47):

What about in the context of learning? Is it also adaptable in terms of sharing content based on the user? How does that work?

Dmitry Shapiro (12:57):

Sorry, could, could you repeat that?

Tullio Siragusa (12:59):

When it comes to learning, for example, if I'm looking to understand something about astrophysics, we talked about that earlier. Does it put it in digestible form based on who's in, who's making the inquiry?

Dmitry Shapiro (13:15):

Yes, that's specifically where I'm going. Meaning, when you ask ChatGPT to summarize something, who, who is it summarizing it for? Is it summarizing it again for an elementary school student or a high school student or someone that is an expert in, in the subject that it's summarizing for? Unless you provide it with context on who it is to summarize for, it's going to give you generic summaries. But one of the apps that somebody created is a personalized summarizer. And so, when you first load the summarizer, you have to train it and train means to answer its questions, engage with its prompts, as it sort of queries you on, you know, who you are and what kind of summaries would be interesting to you. And then it's able to personalize its summarization specifically for the user. That becomes extraordinarily powerful.

Tullio Siragusa (14:07):

I'm thinking about some possible business applications and two come to mind. The educational system could adopt this in terms of how it teaches students of various levels and can make it so that it is optimized for each individual student. So the lesson being learned is optimized for each individual student. And the same thing could be said about employee onboarding, I guess, you know, in terms of how you train a new employee based on optimizing it based on who they are and how they digest information and learn information. So, what's the disruption that this could introduce across learning and, and leveling up, for example, employees you know because I think sometimes you have some really talented people who learn at different paces. And so, it's a one-size-fits-all now, you know, you're training someone, here's the training. Some people will need to see it differently to digest it differently. That doesn't mean they're not as smart as the other person. They just don't process the information the same way and end up failing. But this is kind of a, you know, we talked about demo democratizing. It's really about equivalence. Where do you see this really progressing in terms of enhancing the ability for people to be successful, whether they're learning or they're doing their job based on who they are is, I mean, is how far away is that from, from reality in terms of adopting that? You know,

Dmitry Shapiro (15:40):

I, I think it's here now. We're starting to see these types of, of you know, new ways, again, for us to engage with information technology. I think you're spot on. One of the biggest issues, that people have is that we don't know what we don't know. And so, it's hard for us to be able to go learn something when we don't know that we don't know it. Or what is it that we need even need to learn? And so I think the opportunity of systems like YouAi is to allow sort of AI to probe you to figure out what it is that you don't know and then to be able to fill that in for you on a personalized basis to create on the spot, you know, on the fly, completely personalized curriculum for everyone about everything on an ongoing basis. It sort of completely turns on its side the way we've been creating content. For example, up until now, up until now, we create content for some kind of a persona, typically as it's known in marketing, like who's the target audience? And we create sort of a caricature of this like group of people, right? These personas. And then we create content for these personas. And of course, there is no such thing as this persona. It's just a caricature. Everybody's different in, in massive ways and sort of like what we know, how we see the world, you know, etcetera. But now with generative AI, as long as the AI knows you know, what you need specifically versus what I need specifically, and again, it can query you to figure that out, then it can create custom learning materials for us and custom sales materials and custom entertainment. And it's going to radically disrupt the creator economy as we know it now.

Tullio Siragusa (17:33):

It's amazing. You know, I remember 30 years ago Peppers and Rogers were talking about personalization, you know, one-to-one personalization and it's 30 years later and it's beginning to actually happen. So, what I'm hearing is that there's also the possibility for you know, do-it-yourself projects where you go and look on YouTube, how something is done. I mean, that's going to be something of the past because I could actually do go to this AI, it knows me and I could say, I want to fix something in my car right? And you'd be very specific. The model of the [INAUDIBLE] can actually generate a tutorial on the fly on how to go do this. Is that what the promise is?

Dmitry Shapiro (18:20):

Yeah. Specifically for you.

Tullio Siragusa (18:22):

Right, right. Based on mechanical skills or lack thereof. I mean, this is going to, you know, basically enable every single one of us to become a do-it-yourself expert to a certain degree. Is that kind of like.

Dmitry Shapiro (18:36):

For sure whether people will want to do it themselves separately?

Tullio Siragusa (18:41):

Of course.

Dmitry Shapiro (18:42):

But yes, I am certain that most people don't yet realize the radical sort of change that we are all experiencing right now in our capabilities to be able to, again, leverage information technology. Like I'm old enough to remember that if you wanted to find some information, you had to go to the library, right? Or to a bookstore. And then we got the internet and then we could sort of look things up and we got search engines, and then we got Google, sort of the ultra-search engine. We could find anything and then find anything on our phones. But there's a huge delta between going to the library and being able to look anything up on your phone. I think the delta between looking up anything on your phone and what we have access to now in the form of these generative ais that's as large and perhaps even larger of a delta in capabilities than we saw with the simple digitization of materials and access to it via search. This again, today, what we're generally calling AI is a new set of technologies that will touch everything and, and radically transform the way we engage with each other as humans. The way we engage with the world around us, the way, we engage with technology. And again, we at YouAI are focused on making it possible for regular people to be able to create artificial intelligence, not just use artificial intelligence.

Tullio Siragusa (20:22):

You know, it's interesting when we think about how many industries, the technology or the knowledge industry has impacted or disrupted it's, it's massive, you know and all of them have one thing in common. They've all moved to a self-service model, but it turned every user into an employee, right? In the travel industry, you don't need to go to a travel agent anymore. You do the booking yourself. So, you are now you've become your own travel agent. And, even in banking, you know, you in investment, you've got tools, you can do it yourself. You don't need a broker. Some people still use the broker, but for the most part, you have the empowerment to do it yourself. And for the first time, the knowledge industry is actually disrupting itself with this. Because what, what you're talking about, and I'm, I appreciate this conversation because it's opened my eyes even further. You know, sometimes you're in this space but you can't see beyond, you know, the limited view that it is right now. But essentially I can create on-the-fly access to all kinds of information, whether it's do-it-yourself projects, learning creating new ideas, new content, or whatever I need to, to fulfill whatever purposes I want to fulfill. I could have a tool like this that's my partner that essentially puts me squarely into being a professional from day one, instantly, which removes the reliance on external factors that I don't control and so how fast do you think the adoption of this will be across the board?

Dmitry Shapiro (22:08):

Well, it depends on, I guess if we're talking about sort of enterprise adoption, consumer adoption for what purposes, meaning it's already happening. Lots and lots of people are using, you know, ChatGPT, Bard being like all, all of these sorts of AI-enabled services to do things. There are lots of people creating graphics now with MidJourney and Stable Diffusion. And, and things of that sort. So I think we're seeing extraordinarily rapid adoption of this because, you know, we in a sense, we're sort of now been trained to understand that those of us that can, you know, better leverage information technology to solve our own problems are, are going to have an advantage, competitive advantage as we, you know, do business or, or live our lives. But maybe if I could take a quick step back to this thing, you know, that you were sort of describing this, this agent, let's call it, that helps you fix your car. You could go to YouAi.ai right now in, in the next hour, less than an hour, create such an agent and publish it and get a link and share it. And people that engage with that, this agent will be an auto mechanic in the form of an AI and will prompt them with diagnostic questions in order to understand what is wrong with their automobile and then help them make the changes. Now they still have to like turn the wrench and, and go buy the parts, but creating the intelligence that does that for them, you can do that. You right now, you've never tried it before, but you can do it in under an hour and then after that, you can do it in like five to 10 minutes. Because again, all the heavy lifting of like knowing how to fix an automobile that's already in these large language models, because that's all been written, and these models have already digested it. And now you can just create an interface so that regular consumers can show up and, you know, call this Tullio's bottle shop and, and you've just created an AI that lets you do that. And then later you can create an AI that helps people plan their, you know, meal prep for the week. Like, that's the democratization of AI that we're talking about. And it's here now.

Tullio Siragusa (24:25):

Yeah. Pretty impressive. Well, it's a brave new world, so we're looking forward to seeing how it all comes together. Sounds like YouAi got to dial in. Definitely go check it out. It's Youai.com What's the URL?

Dmitry Shapiro (24:40):

It's YouAi.ai

Tullio Siragusa (24:42):

Awesome. Great. We'll have to check it out. Stay with me as we go off the air in just a second. Next week we've got Andrei Girenkov forgive me if I botched that up a little bit. Who's the CIO, CTO and digital transformation executive at Greystar. We're talking about technology, the tip of the iceberg and what digital transformation executives need to know. So if you're still undergoing your digital transformation or on your digital journey, come join that conversation. That's going to be on Tuesday at 9:30 AM July 11th. See you all again very soon. And enjoy your weekend. Take care.