Dmitry Shapiro Podcast Transcript
Dmitry Shapiro joins host Brian Thomas on The Digital Executive Podcast.
Welcome to Coruzant Technologies, Home of The Digital Executive podcast.
Brian Thomas: Welcome to The Digital Executive. Today’s guest is Dmitry Shapiro. Dmitry Shapiro is an American entrepreneur. He is the founder and CEO of MindStudio. MindStudio allows users to easily launch AI applications without any coding knowledge required.
Shapiro was previously founder and CEO of Koji, an app store of social mini apps that are added to various links in BIOS. Koji was acquired by Linktree in 2023. He also created the video sharing site Veoh in 2005. Shapiro served as a CTO at MySpace before moving on to Google, where he took up the role of Group Product Manager.
Originally from Russia, Shapiro moved to the US at 9 years old, where he found his passion in coding. Well, good afternoon, Dimitri. Welcome to the show. Thanks for having me. Really excited to see you. Absolutely. Appreciate that. Dimitri, I’m going to jump right into our first question here. Could you start by sharing maybe what inspired you to found MindStudio and how it aims to transform operations across every enterprise in every industry?
Dmitry Shapiro: Yeah, look, I’m a techie. I’ve been building products since 1984 when I was in high school, and I started writing code when I was 14. And so, I periodically get inspired by the opportunities to build things and I get inspired for, you know, kind of for different reasons. This is my 3rd venture back company as a founder that I’ve built.
Another 1 was a competitor to YouTube called via. I was a major competitor. I raised 70M dollars there, did that for 5 years. Before that, I built a venture backed cybersecurity company called Iconic Systems. Raised 34 million for that, uh, and then I was at MySpace as chief technology officer, MySpace music, and at Google for four years on the product side.
So, I’m a product slash engineering guy that loves to build things. What inspired us to build MindStudio was really the release of, of ChatGPT and kind of this new era. Of all sorts of consumer and business accessible models, AI models that have amazing capabilities. These generative AI models. And it became very clear to Sean, my co-founder, and I shortly after the release of chat, that while the models are extraordinarily powerful people’s ability to actually get those models to do anything meaningful.
Is sort of prohibited by the interface to the model being sort of this command line interface being chat, right? Being conversational. And while that’s amazing that you can talk to the model, you know, and type with your fingers in order to get these models to do a lot of things that they can do that you might want ’em to do, requires you to type too many words.
And that there really needs to be a better interface to models that. So that was sort of a realization. Another realization was that, that GPT, you know, chat, GPT wasn’t going to be the only game in town. Open AI was going to have competitors, and now they came much faster than we expected that they would, especially now with the open source.
Sort of part of this landscape, and it was going to be important to have a model agnostic abstraction layer that would abstract all of this innovation and the models and then allow. You know, businesses organizations, like, you know, people to build applications on top of all of this new intelligence on top of this intelligence layer.
And so, we often refer to mine studio as the application layer of. Sitting on top of the intelligence layer, which is sitting on top of like this infrastructure layer where you’ve got, you know, hardware, net compute, et cetera. So, it was just very clear that there was a going to be a need for this, and we had a vision for how to do it. And, and here we are.
Brian Thomas: Thank you. I appreciate that. And that helps our audience, how you broke that down and explaining the need for an interface and what layer you’re at for interfacing with these LLMs. Right? So, I appreciate the share. And Dimitri, MindStudio is positioned uniquely at the intersection of AI and user accessibility.
What challenges did you face in making advanced technology approachable for the average enterprise employee, and how did you overcome them?
Dmitry Shapiro: It’s a big question. I think the important thing for people to understand is that while you can tell, let’s say a large language model to do something, and it does it right, like, write me a poem and it writes you a poem and you say, oh, my God, that’s incredible.
Look, these things are brilliant, and they can write poems or write me an article or, like, solve this problem or, you know ask me a bunch of questions and then try to figure out, you know, what ails me basically diagnosed me. Like, there’s a lot of amazing things that sort of conversationally. You can do with these models again, the real power of them comes when you use them at the right time for the right things.
And so, again, it was clear to us that. Or there to be many sorts of use cases, and we’re focused on enterprise. So, it’s saying enterprise. And when I use the word enterprise, I don’t actually mean large companies. I mean, like, any kind of organization. It could be a one-person organization. It’s like any kind of an endeavor that is trying to get something done.
That, like, cares about productivity and cares about efficiency and cares about output and quality and things like that, right? Because there are also, like, sort of countless consumer applications, too. And by the way, many people use mind studio to build those also, but we’re kind of not focused there. So, when I say enterprise, I just mean any organization that there are sort of countless things that generative models can be used for inside of enterprise, whether their language models or image or video or code, like, not just language models, but these let’s call them generative AI models.
What really needs to happen is the ability for business users to be able to sort of articulate what it is that they want to happen. And have sort of a, I do that and the right way to do it is not to just sort of give a giant prompt of instructions. A better way to do it is to create basically a multi-step workflow.
That says, you start here, then you do this, this, this, this, this. And then you end here, and this is the process that I want you to do for me AI model. And so we prior to this built another platform called Koji, K O J I, that allowed people to basically sort of connect things together really easily. And before that another platform was called Metaverse Studio.
And so, we’ve built these kinds of platforms before that allow people to have a canvas. Sort of a digital canvas onto which they could place components and connect those components together and configure them in order to be able to do stuff. So, we kind of already had that in the bag of, like, how that could be done.
This is a completely new code base, but what kind of learning was there already? And so, we use those learnings to, again, very quickly sort of create the 1st version of mind studio. You know, it’s like last summer and launched it and people very quickly understood how to use this type of canvas of these, you know, components that you configured.
And then since then, we’ve obviously expanded the capabilities dramatically and sort of expand it all across the board of, like, what you can do with my studio.
Brian Thomas: Thank you. I appreciate sharing some of your iterations of where you started and some of the challenges you faced. Certainly, appreciate that.
And Dimitri, what mistakes are you seeing in regard to enterprise adoption of AI today?
Dmitry Shapiro: Yeah, there’s a lot of confusion, you know, rightfully so. It’s a completely new thing that’s been thrown at all of us, you know, over the last year and a half of like, like, what is AI and what do you do with it as an enterprise?
Right? Right. And. A year ago, if you would’ve asked this question of everyone, it would say, well, duh, you, you go to chat GPT, and you create an accountant, and you sort of practice the sort of prompt engineering and you maintain sort of like these long lists of prompts in A PDF. And then, you know, you collect them, you might even buy some from some people on, you know, on LinkedIn that sell 5000 prompts and a PDF, and you can just buy it from them.
And then you can just find the right prompt and copy and paste it. And while that’s how you do AI. As an enterprise, and that sort of set the stage for people and a lot of people, a lot of enterprises are still sort of seeing the world in that way. For early adopters and enterprise of that type of an approach, they very quickly realized that that does not lead to massive gains in productivity.
In fact, oftentimes it leads to loss of productivity, and the primary reason for that is now employees instead of doing the work are sort of nudging AI to try to do the work for them. And that process of nudging AI by typing things to it, typing commands to it. Well, on 1 hand, it’s amazing that you can do that.
Keep nudging it to do work. That’s certainly not very productive and not very efficient. And oftentimes you actually don’t get the best quality work out. Then you’ve sort of spent a bunch of time doing the wrong thing. And you didn’t get the output that you wanted anyway, and you sort of settled for it.
And so that’s not a good approach to really leverage the power of AI in there. Then companies started taking another approach of like, when a whole bunch of these sort of vertical SAS solutions popped up, you know, Jasper, copy, writer. And then there’s like, there’s AI SAS for sales and marketing and operations and all that.
And so, some enterprises, you know, once they sort of tried the, the give everybody access, direct access to a model and see how that works. And that didn’t work. They said, oh, okay, that makes sense. The right way to do it is just to cobble together this, like, new set of SAS solutions that we can train our people to use.
And here we go, we’ve transformed that our enterprise with modern tools, but that, too, is not the best approach because while that is a better approach than the 1st, 1, this approach has other limitations that you’ve got to cobble together all these tools. You’ve got to train your people. You’ve got to manage them.
And people are then sort of copying from 1 tool to another to another, and they’re sort of orchestrating this manual. Okay. So, mine studio takes a radically different approach to this and sort of looks at the world in this way that the right way for enterprises to do is to understand what I can do and what I can’t do.
And that’s a matter of education. And so, we provide some simple YouTube tutorials that explain to people how generally this stuff works. And once you got that, you can then sort of look at your enterprise holistically and clearly see that there are some things perhaps in your enterprise that can be completely automated.
And that you yourself can now build these applications, these automations that do that. And so, you go ahead and build those rapidly check. Then you realize there’s some things that can be partially automated. And you can do those partial automations for things that still require humans. The work that humans have to do now is different because you’ve automated a bunch of things or partially automated a bunch of things.
And so now you can look at what humans have to do and say, can I give them better applications that fit them more like a glove. To do those things better, we can use custom business apps, and you can use mind studio to create these ai powered custom business applications. For each job function for each employee, you can sort of update your business application, tech staff.
Yourself using mind studio, and then sort of finally for all things that still sort of come up like ad hoc, because employees need the ability periodically to use intelligence in a conversational way. Instead of just letting them use chat, the enterprises can give them now sort of custom-built enterprise grade specialized assistance for each job function.
That knows about the enterprise has access to enterprise data is integrated with enterprises other applications. It can be logged, can be archived, can enforce policies. And so, you sort of get a complete solution of taking an enterprise and making it an AI enabled them. And so, we now have over 50, 000 of these eyes that have been built and deployed across, you know, giant enterprises, government agencies, thousands of SMBs.
It’s overwhelmingly that’s completely self-service. When we don’t have any salespeople at all 0 enterprises show up, learn about it and sign up for an account start building. There are now over 100. You know, it integrators and various types of agencies that have showed up. And have said, hey, our enterprises, our clients are asking us to.
Use mine studio to build things for them. And so, can we partner with you? So, we just launched a partner program and agency partner program. There’s a lot of experimentation going on right now. I would call mistakes. I think, like, experimentation of. Of enterprises trying to figure out what to do, and we sit at this point where, by the time they come to us, they’ve tried a bunch of things.
Realize that those weren’t sort of the things that paid off in a big way and, and, and are reporting that this actually works, you know, dramatically better.
Brian Thomas: Thank you. I appreciate that. And you’re really helping the user adoption is the way I see it. The way you explain things while people are pushing the envelope as far as using the large language models.
You’re actually a partner for all these people, or like you said, any enterprise for enterprise adoption at any level across an organization, big or small. So I appreciate that. I really do, Dimitri. Dimitri, last question of the day. Looking towards the future, what emerging trends in AI and technology are you most excited about? And how does MindStudio plan to leverage these trends in its offerings?
Dmitry Shapiro: So MindStudio sits on top of all of this innovation in like all of these trends, right? These models get better. MindStudio allows enterprises to leverage those models. We’re completely model agnostic. You can connect MindStudio to any model or models that you use.
A bunch of larger enterprises have their own models that they’re running. They run inference in a private cloud or on premise, et cetera. You can connect to those, you know, we support sort of all the public ones, open source, et cetera. As innovation happens, MindStudio, because of, again, what it does, that it’s the subtraction layer between what the businesses need, their applications, And this intelligence layer sort of benefits from it.
What am I excited about the most? Yeah, that’s a that’s a tough one. One things are evolving so quickly that every day, even if this is all you do full time, you can’t keep up with all the innovation. I think that the general path is, is straightforward you know, multimodality. And so, you know, what we saw, for example, with, uh, GPT 4-0 yesterday, right?
And so, you’ve got. And by the way, multimodality is going to come to many, many models. I’m super excited about. Small models. Becoming much more useful in enterprises, large models aren’t the solution for all use cases. In fact, in our use case, meaning some process that needs to be automated or some business tool that needs to be created for the enterprise.
Using just large models to facilitate that is inefficient and, you know, has a bunch of latency and things like that. And it’s, it’s much better. And we see this already happening in these 50, 000 eyes where enterprises have realized that using actually smaller models and sort of chaining a bunch of them together in these multi step workflows intelligently.
You get better performance, you get lower cost, you get an easier way to customize them and train them and keep them updated in all of that. And so, while sort of watching the amazing, you know, innovation in these giant models, I think the, the real heavy lifting. In enterprises is not going to be done by the large models.
It’s mostly going to be done by smaller models that are orchestrated in the right way inside the enterprises that becomes powerful. I think the sort of societal change is the other thing that arguably becomes kind of the most important part of all of this is. Is like, okay, great. You can do a bunch of things with efficiency and enterprises, right?
And that’s where we’re focused. Right on that. But it is going to dramatically change the way we communicate with 1 another. You know, today we need somebody new in business or somebody new out in a bar. Right, then we start a conversation and that becomes the beginning of a dialogue and of us getting to know 1 another and that moves pretty slowly.
Because conversation. Is fairly slow, but what if we can sort of show up and bump our phones and instantly and I can look at our sort of personal digital wallet of information, mine and yours in instantly say, okay, let me print out for you. All the things you don’t need to talk about because you already completely agree on all of these things.
And I can see it. Oh, wow. We’ve got all these things in common and now we can instantly jump into the deep end of conversation. That’s amazing. And that’s going to transform relationships. And when you take that to teams. That becomes extraordinarily powerful building teams is the hardest part about building a company, keeping teams aligned because things are constantly changing is a never ending, you know, struggle and everybody’s failing at no matter what you and the best teams are extraordinarily poor in being able to stay aligned.
Right? They’re very fuzzy, but I can actually keep us all aligned by being able to do the same thing that I was just describing. Right. But at a team level to be able to observe the operations of teams and realize that while people say that they are aligned, they are not exactly behaving as if they’re aligned and sort of this detection of human miscommunication is an extraordinarily valuable thing.
We see that there is already a bunch in enterprise. Of enterprises that are using mind studio to build, you know, automations and assistance inside of sales processes. For example, you know, sales people tend to, you know, sales is this, like, again, multi-step process where, where, where you’ve got to build rapport and you’ve got to communicate value propositions and you’ve got to understand objections and like all of these things and, you know, great sales people are radically better than, you know, average sales people.
Why? Because, you know, there’s stuff there that matters that can be changed. Well, AI now, first of all, everybody’s transcribing their conversations. Like, you and I just turned off my fireflies assistant before we started that, but that thing shows up in all the conversations. And typically, those things just go into some bucket and maybe someday compliance is going to look at that.
That’s cool. But now what you can do is you can build AIs really easily that are constantly looking at sales people’s communications with prospects sort of through the lens of let me see what’s happening between these people communicating and let me see where they’re miscommunicating. See, the salesperson thinks the prospect is ready to buy, but the AI knows the prospect isn’t ready to buy because the prospect has not asked any of the questions that they would ask in order to understand, you know, what, that they’re ready to buy.
And so, the AI can proactively, you know, help the salesperson understand the situation, automatically collaterals, for example. That proactively teach the prospect of things. And so that becomes, you know, when human to human communication becomes dramatically more efficient. And nuance, well, wow, you know, kind of everything changes beyond that.
By the way, I also think that’s critically important for the stability of our world. We live today in a society that’s extraordinarily fractured is the term people like to use. Why is it fractured? Because we’re not able to communicate with 1 another. Why? Because we are all very busy and we have very different perspectives now, because the polarized sort of media and social media.
And so, nobody’s got time to sort of work out the wrinkles, but the has time to work out the wrinkles. And so, I can help people communicate radically more efficiently and bring them together and therefore is the thing that could possibly help save our democracy and other democracies and create more stability in currently a very, you know, seems like an unstable world.
Brian Thomas: For sure, for sure. And Dimitri, I appreciate your perspectives. Obviously, we know where we’re today with technology, but you went through some of the examples where you see it evolving. And I think that’s promising in a lot of ways. It’s exciting for technologists like you and I, but we need to proceed with caution and hopefully AI will come with some solutions, especially in this very fractured world we live in today.
Do appreciate it. And Dimitri, it’s been such a pleasure having you on today. And I look forward to speaking with you real soon.
Dmitry Shapiro: Thanks for having me.
Brian Thomas: Bye for now.
Dmitry Shapiro Podcast Transcript. Listen to the audio on the guest’s podcast page.
