When A.I. Isn’t Enough – A Conversation with Chaitanya Hiremath from Scanta

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Alexander Ferguson 0:00
Artificial intelligence machine learning, these emerging technologies are changing the way we live, work and do business in the world for the better. How is AI actually being applied in business today, though, in this episode of UpTech Report, I interview Chaitanya Hiremath also goes by Chait. He’s the founder and CEO of Scanta, who’s a revolutionary adaption of augmented reality, and machine learning will potentially change the way you commute to work, play a game, and schedule your next meeting. So tell me, Chait, how did you get to where you are today?

Chaitanya Hiremath 0:34
that dates me back. So we started Scanta in 2016. It was it was September, I remember. So that’s when we did all the paperwork. And we incorporated the company. And at that time, just just before that, I had another startup that was closing down. And we were focused on 3d animation for VR, creating hyper realistic visualizations. But you know, I used to go with the suitcase of virtual reality headset, you know, computer, and they all you know, the buyers and everything else, and try to sell that share this technology, right? With, at that time, my target market was real estate. So fix that multiple times and make, you know, this big suitcase that I got, specifically, like, you know, a sticker put on and did all of that stuff. And that’s where, you know, I focused on integrating or working on the Google Tango project and seeing what could be made. Specifically for the A for mobile AR in late 2016, or return 17 worked with Coca Cola for an AR campaign, it was, it is known to be the most successful AR campaign for brands and for specifically Coca Cola, you know, from their innovation team. Since the last five years, it’s been on the top of the charts. It was the most successful campaign in South Asia. And the CEO talks about it a lot. And it’s been it was his it’s like a great learning curve for us. And what we did with Coca Cola was, you know, from a tech standpoint, at that time was like, wow, this is this really happening, you know, when we did the campaign in India, and, you know, his people were not exposed to anything like that, you know, you’re talking about, you know, you putting your phone somewhere and robots entering your dimension and doing some crazy dance and giving you a Coca Cola. So it, it got a lot of appreciation. And you know, that God has some revenue, so that was nice. But after that, we realized that the business was not really scalable. After which we started a project called pica Maji, which is an AR avatar library. So we build over 2000, late 2017, early 2018, the biggest library of AR avatars in the world, with 110 characters or an IP that I realized that people actually want to interact with these characters, were allowing 3d characters and MLPs to be integrated with our technology so that it is possible for us to find me have the person connect with our machines or our devices or our 3d characters.

Alexander Ferguson 3:42
That’s brilliant, because machine learning and the ability, as you say, to to give a personality to the these interactions that that is a game changer. And that’s what fascinates me when I look at your technology. So you say, on your sexy, we fuse augmented reality machine learning to synthesize smart animations, thereby instrumenting a better way of communicating feelings. Interesting, the fact that being able to have a computer have feelings. So I think it’s a fascinating concept. I want to dig into more of the technology in a second. But first, coming back to the company itself, as obviously you’ve evolved over time as every business does. How would you say you found like product market fit with this pivot? Or is it still kind of in that exploration phase?

Chaitanya Hiremath 4:28
Yeah, so fundamentally, you know, that’s something I’ve learned over the last three years, you know, product market fit is something that you know, what we did earlier with the game and everything else, that’s something that we probably should have done earlier, rather than after or during, you know, and now, before we started this entire process, we went to a couple of players in the market and try to understand one of them Being very, the most famous game engine that’s out there. So we went and we had multiple discussions with the r&d team. And, you know, one of the funding fundamentals that came about was, this was about a little bit more than geared backwards. We’re talking about an era where, you know, when even when you’re playing a game, you know, you’re, you’re there on, you know, live streaming platform, and you have 10 other friends, and you’re interacting with them. But that’s not always the case, there isn’t there is a need for you to fundamentally be more engaged within the game, to have that sense of immersion. So why are you not able to speak to your character, in the same way that you speak to your friends, and when you when you have those interactions, you need to have a unique personality, so that the engagement is higher, there’s a higher level of immersion, right? So you need to identify that, you know, your God has a different name for your system, and it talks in a different way. And the other car talks in a different way. Fundamentally, they might do the same functions, but that the uniqueness that needs to come out, and that’s the process with the gaming. So even when you’re talking about gaming, right, if you’re talking about why can’t we create complex and LPS for each and every those characters, right, you’re talking about hundreds of characters, how can you possibly create that right, and all of them have different personalities. But if you have the same data set, just with the same data set, if you add a different layer of personality, on to each and every one of them, you can just tweak the personality and have the core knowledge set, which is the same. And for the user, it would just be different because they sound different. And the construct of the sentences and how they’re reacting to it is different. So you can apply scalability.

Alexander Ferguson 7:05
What Amelie comes my mind, I’m also enjoy games very much. And I play several, I think of a massive undertaking, like Skyrim, where has all these stories, all these unique characters, but they had to program and do voice actors on all of them. And it probably that’s why it’s such a massive undertaking. But if every character in a game like that could be its own interaction done through machine learning that would scale a game instantaneously. All they’d have to do is set a few settings. Am I getting your your vision? Correct?

Chaitanya Hiremath 7:40
Absolutely. And also, from a voice standpoint, one of the technologies that we have done a lot of research on is neural voice transfer, or, which is a part of, you know, some people put it as a part of voice cloning, you know, where we already have about 10 distinctive voices. Which, if you looked at the WaveNet demos of DeepMind by Google, right, I mean, they have, like, really human like voices. And you must have maybe seen it on the Google I O conference, where Sundar Pichai had this, this demo of China calling a Chinese restaurant and booking a reservation. And you know, and that’s all generated, right? That’s part of the whole NVD space. And that’s what we foresee right before that, though, different voices could be integrated with it with different characters

Alexander Ferguson 8:44
see painted a vision like that I find enthralling and fascinating. But my next question is, okay, are we there? Now? I mean, could Can you roll this out to a client right now with your platform?

Chaitanya Hiremath 8:55
So right now, at the stage that we’re in, we’re in talks with one of one of the really the biggest platforms in the world in terms of gaming, right, in order to see how this could be integrated? Because it’s fascinating, right? It’s the whole scalability component of it is really interesting. And where the technology is, right now, there are two fundamentals to it, right. One is the personality engine, and one is the voice. Now, both of them were able to achieve some level of accuracy that’s ready to go. But that’s something that we constantly want to evolve, you know, over a period of time, make it more accurate, right. And that’s something right now will right now, as we speak, we’re in talks with a few players regarding integrating this technology as it is. So yes, it’s good to go. But is it something that will, you know, consumers or users in users can interact with within three to six months? That’s unlikely, but definitely what we’re looking at is working with these players so that we can create, you know, higher accuracy and release these release these solutions in about one year, two years per 12 to 24 months timeframe.

Alexander Ferguson 10:16
So right now working with this one undisclosed a large developer platform. And then also, the idea is that other individual game developers could reach out and potentially start integrating this and then over the next year or two, it could become a reality. That is that right?

Chaitanya Hiremath 10:36
Yes, that’s right. And like, so that’s why we’re approaching platforms. Right? So we’ve got two approaches. One is platforms, and one is industry specific and industry specific, we just focused on automobiles as of right now. So so we’re trying to sort of add that layer of personality for the existing solutions. So both of these areas, right, one is sector specific. And when we talk about platforms, you know, that’s where the really the technology reaches, you know, the average developer, and they can, you know, make beautiful things with it. And that’s the game changer in Rs.

Alexander Ferguson 11:15
So then, if, if another platform or let’s say, car car, manufacturer would want it to do this, do you have like certain packages, or pricing or methodologies like, Yep, this is how we can work with you.

Chaitanya Hiremath 11:29
It really depends on like, you know, when we talk about the requirements, right, most of the automobile players have recently really woken up to this possibilities when we talk about the voice, and moving beyond just the rule based interactions. So it’s really contingent on what is that they’re looking for, you know, I were talking about the voice plus personality engine, or they’re focused on, they just want one voice in there. You know, what sort of tweaks need to happen on the core NLP, you know, and there are multiple different elements to it that define pricing. So it’s not like, I wish it was that easy that we’re like, okay, just do this. And this is how we’re gonna charge you know, but it just, it’s a very, it’s a new and upcoming space. And that requires, you know, in the first in the beginning stages to sort of settle in, and to understand the market a little bit more as to how these technologies could be integrated, what value could be added, and from there on in will have a much more pure understanding.

Alexander Ferguson 12:44
So let’s dig into the technology itself more, because that’s, that’s what you’re kind of like morphing? Shaping is really are your sweet sauce is this machine learning ability. And I think on your site, you say, it’s your own type of NLP. Deep learning.

Chaitanya Hiremath 13:04
Yeah. So we have our own NLP that we made. That’s what that’s what we researched on. But what we understand from it, then somebody wants to integrate that NLP, we’re happy to do it. But in most cases, when you’re talking about either automobile space, they have their own very specific NLP that they require. So our focus area is not that we’re selling an NLP solution. If somebody does not have it, or need assistance with the current one, we can do that for them. But more importantly, our thesis is that we’re able to integrate our personality engine on top of the NLP and reconstruct it in a way that is required. So if you so the personality engine itself is based on the ocean model, the ocean model is, which is predominantly used in order to set up all personalities within human beings. So it’s about five basic principles to it, you can look it up. It’s really interesting. And the objective was that all human beings come under the ocean mode. So similarly, what our solution does is it gives you n number of personality types for the machines. So it’s, you know, it’s like sky’s the limit as to what is the what is the, you know, which kind of personality you specifically want? It’s not five, it’s like, n number of personalities, right? The question comes in, like, which personality suits which industry? That’s research on its own. So we go through heaps of data and understand, okay, you know, this is how a voice assistants personality should be, you know, and that could be linked to the relevant personality that you have. So based on user’s personality, you know, you might want someone that has more agreeableness. such openness, somebody might want different attributes that’s completely contingent on mapping your personality, to what the personality of the system should be.

Alexander Ferguson 15:12
What’s fascinating to me is I’ve had a conversation when it comes to AI with another expert. And we were talking about this same concept in a while you’re driving, and you have some voice assistant telling you directions. And let’s say you’re actually really frustrated because either you’re running late or it’s traffic, and you’re angry. But this voice assistant is so pleasant, and life is just so wonderful. And so you get mad at this voice assistant, because it’s so happy, you’re like stopping so happy. I’m imagining now with your ability, and this this concept is vision, your painting that now can match your own, potentially your own emotions a year, right? This, this sucks turn right here and the other drivers.

Chaitanya Hiremath 15:52
Exactly. So you don’t want that, you know, you have different emotions throughout the day. And it needs to be dealt differently. And we have the capability to identify the user’s emotions anyway, with the camera inside the car, you know, we can use it for various different things. So the system needs to know when to shut up. And when to talk, you know, and based on your emotions, it needs to basically adjust itself, you know, it because if those capabilities are not there, then you know, it’s the same old thing that’s been going on for such a long time.

Alexander Ferguson 16:35
That concludes the audio version of this episode. To see the original and more visit our UpTech Report YouTube channel. If you know a tech company, we should interview you can nominate them at UpTech report.com. Or if you just prefer to listen, make sure you’re subscribed to this series on Apple podcasts, Spotify or your favorite podcasting app.