Latest / Convo AI World / Interactive Digital Avatars with Trulience's Richard Bowdler

Transcript

Interactive Digital Avatars with Trulience's Richard Bowdler · Convo AI World

• Hermes Frangoudis (00:07) Hey everyone, welcome to the Convo AI World Podcast where we bring you interviews and insights from companies building solutions in the conversational AI space. I'm Hermes Frangoudis from Agora and today our guest is Richard Bowdler, Head of Growth at Trulience. For those that don't know, Trulience builds some amazing interactive digital avatars for conversational AI. Richard, thanks so much joining us today. Richard Bowdler (00:27) Thanks so much for the intro and for having me on the show. Great to be here. Hermes Frangoudis (00:31) So let's kick it off. Can you share with us a little bit of the origin story of Trulience? Like what inspired the founding team to focus on like digital avatars in this space? Richard Bowdler (00:39) Yeah, The founders of Trulience, this is not their first rodeo. They have a previous exit. It was video streaming tech that they sold to one of the large EdTech platforms out there called Blackboard. And well after that, wanted to go on and build something that is still, again, very much a a frontier Marek started building Trulience, it's about five years ago now, kind of pre, you know, pre the explosion of AI. And it was a combination of him wanting to build something that was really on the on the kind of frontier of what was possible with technology. And also, on the personal side of things, one thing that sort of drove him to this, struck a chord with him, was a thought that he had an old elderly grandmother who was in a care home quite far away from where he lived and his thought was it's just such a shame that she sort of sits there languishing and doesn't have the level of attention or interaction that she could have. I wonder if there's a technical solution to this as well as a human solution. And it was a seed of an idea from which Trulience was born. Now, that idea has evolved and the kind of guiding vision is how can we give AI a face and a body to make it more human? Hermes Frangoudis (02:08) It's super interesting. It's like very human touch to that, to the story. Like the inspiration is how do I bring that, that face-to-face interaction back? Right. It's super cool. So in, in building that, like what were some of those initial challenges maybe in developing this sort of Avatar Technology? Richard Bowdler (02:19) Mmm. The world we live in now is very different to what things looked like when we were starting out. When we were starting out, it was heavy studio builds, heavy lifts, full 360 Cages of surround cameras and intensive filming days to make sure that we could capture people and get you know, lifelike imagery from as many different angles as possible. So there's a kind of a huge front-end lift in building even just fairly rudimentary avatars. So that was a big challenge. And then the other piece, which again, the world we live in now is very different, was trying to get the avatars to be meaningfully and reasonably conversational. I mean, now with the kind of explosion of LLMs and the voice AI layers that sits on top, we were kind of spoiled for choice really in terms of how we get that to work, which we're now in a world where you can have conversations with AI and it feels incredibly fluid. But way back then, it was tricky and it's taken a while to get to this point. Hermes Frangoudis (03:42) I couldn't even imagine the early days of training the AI to understand from all angles at the same time. I remember the capture cages because we came from the AR background. So we used a lot of that for bringing virtual avatars into the scene and animating them. And boy, were those things heavy. So it's super cool that that's kind of the roots of this. Richard Bowdler (03:54) Mm-hmm. Yeah And it's kind of evolved, right? So the tech, you know, it's evolved through these capture cages. And then when I joined, we were doing mocap stuff, capture suits with dots on, black suits with dots on, and filming that so we could then dress, sort of digital mannequins. When we started out, we were working in Unreal Engine, a gaming engine, and building with We no longer build with gaming engines, but but we do still have some CGI avatars because they fit with certain use cases, certainly, in gaming and so on, where there's less of drive or a want for hyper real. But the techniques have evolved so much even just in the last six months. So this side of Christmas, it's kind of things have very much exploded in both appetite. And I think it's probably because of where the product is at. Not just Trulience, there are others out there in the market, but certainly the techniques that we've developed. Hermes Frangoudis (05:04) Yeah, seeing that maturity, it's definitely like, created a new interest and invigoration probably in the market. And keeping on that thread, what are some of these key components that enable Trulience's like real-time avatars to seem so interactive? Richard Bowdler (05:20) Well, you have some of the guts of it. Let me just break it down in simple terms. Trulience, we deliver the visual front end. We give AI a face and a body where it needs to be. And that's both hyper real humans or characters. So, you know, like an off trademark Mickey Mouse or like a fun fluffy dog. So we are that. We sit on top of the shoulders of LLM and voice AI giants. We have built the platform so that you can kind of "bring your own brain" as we're terming it. So you can plug in any language model, any voice AI tech, and we play around with different combinations of those to see what is it that really plays nicely together. Like which language models, where are we hosting them privately, publicly, etc. And which versions of those models are best fit. And then both on the speech-to-text and text-to-speech, we play around with those pieces too. And by fusing those together, we can then have a real-time, realistic latency. A conversational interaction with an avatar. It's worth pointing out a kind of curious quirk here. Now, there is voice-to-voice tech, and we have integrated voice-to-voice tech, but the latency on that is. You can dial the latency kind of up and down. There's a threshold of about 500 milliseconds response time. So from when I speak, then there's a gap, and you speak, and natural interrupts. Anything below about 500 milliseconds, like you ask a thing a question, if it's below 500 milliseconds, to a human, it feels like it's happening too quickly. Yeah, it feels weirdly too quick. Hermes Frangoudis (07:13) instant, right? Richard Bowdler (07:16) We have got to the point where we are as fast as we need to be. Faster than we need to be. So it's kind of quite an interesting quirk. One of the technical difficulties is in lip syncing. So having, it's called the visemes, the shapes that lips make with different people speaking different sounds. And then you multiply that across into different languages. And then you have all different other shapes that people make with their mouths when you say, perhaps speak in Arabic or Hindi or what have you. Which that is its massive technical challenge. But we have a team that they focus exclusively on lip syncing and mouth shapes. Hermes Frangoudis (08:02) Feels like lip sync and mouth is like 90% of it right like in terms of the avatars responding it's moving, huge, but the second that lip sync kind of falls out, it breaks the magic and every time I've seen your avatars they seem to like really stay on cue so it makes sense you have a whole team focused on that. You brought up multilingual and like doing it across languages. Richard Bowdler (08:19) Mm-hmm. Yep. Yeah, yeah, Mm-hmm. Hermes Frangoudis (08:25) So how does your system handle like maybe multilingual conversations or like real-time language detection? Is that passed through to the LLMs and then it's more just so about like syncing of the motion? Richard Bowdler (08:36) Exactly that. We don't handle that. We work with that. You know, there are obviously lots of providers out there that do that. Hermes Frangoudis (08:38) Nice. So you don't So you don't really constrain the customer in that sense. You let them bring their.. Super cool. Richard Bowdler (08:49) Exactly. And there are some. Azure is pretty good for multilingual within the context of one conversation. So an avatar can listen in Arabic and if asked to, can respond in Spanish or whatever combination of languages from their selection. It's kind of awesome to experience. You're like, wow, that's crazy. Hermes Frangoudis (09:16) It probably, for like someone that is multilingual, probably feels actually very like at home. Cause like I speak Greek and English, right? And with my parents, it just depends on what words like the simpler one to set. So you get like this weird like mix of the two languages. I would be curious to like play around and see how it works with that. Yeah. And pushing the boundaries there. But moving along on the subject of like core technologies. Richard Bowdler (09:22) Mm-hmm. Yeah. Yeah. Hermes Frangoudis (09:42) Can you discuss the role of transformer-based models in really enhancing your avatar's conversational capabilities? Richard Bowdler (09:50) Yeah, I mean, you know we work with, like I said, any large language model and you can bring your own to the party. your point just now, when people use our platform to build avatars, they're not constrained in terms of any transformer model that they want to bring. So for one of our clients, they have built quite extensively a knowledge base inside OpenAI. And it's very specific. It's in an area of, I won't give away too much, but it's in an area of coaching and training for adults. So a learning and development coaching tool. And they have a very rich library of prompts and a knowledge base that they built out inside OpenAI, which is like, hey man, we don't tend to use OpenAI and ChatGPT for our stuff, our demos, just because there are other models that we've used that play more nicely. But it's pretty good. Yeah, there's no point. Plugs in and it works, right? Exactly that. I mean, when you log in and it's free to sign up for an account, you get a free 30 minutes a month and Hermes Frangoudis (10:54) plugs in and works, right? Richard Bowdler (11:03) You can use any of the ready-made avatars. So if anybody's watching or listening, you can just go to Trulience.com and get started without any barriers. And you'll see in that you go straight into what's called the Avatar Creator. And there's a section in there. So, you choose what your avatar looks like. And then we can also do custom avatars, so to clone people. or characters or whatever. Obviously with consent, it's worth me saying that. And then there's a section for creating "your brain", which is, what what does the voice sound like? What's the language model, the transformer model? And how do you want to output? And then how do you want the avatar then to behave? Which we kind of drive that through prompting, as well as bit of RAG, so Retrieval Augmented Generation, to kind of tweak the models. But it's possibly worth touching a bit more deeply on models. So we can plug, you know, can just plug into any public model or, you know, we have we have clients who have sensitive data and they want to host everything themselves. So I'm talking about in like financial services, pension funds. These sorts of customers. They want to have their own LLM privately hosted. Same kind of thing. And with some, they want them specifically tweaked at the language model level, not the knowledge-based Retrieval Augmented Generation piece that sits on top of the language model. They want the underlying model itself to be fine-tuned. And that's something that we can do. And that then feeds up and feeds up into how the avatar behaves and operates. Hermes Frangoudis (12:51) Super cool. Seems very plug and play. Bring your existing infrastructure, give it that face, as you say. In terms of giving it a face, one of the really important pieces of face and conversation is the emotional nuance and naturalness of the interaction. So how does Trulience really work in that sense and try to capture that piece of the magic? Richard Bowdler (13:18) Yeah, two main categories of avatar types are one is an image-based avatar and one is a video-based avatar. So the image-based avatars you can take, you could take a screenshot of you or me and we could then turn that into an avatar. Or we could take, take a video section of us speaking and turn that video into an avatar and both would give you know very realistic lifelike representations. What we then do is have a range of emotions that we have baked into our model and we will then map those to you know a surprised response, raising of eyebrows, furrowing of brow and so on. There are things that we can do with that. It is one of the areas, like with lip syncing, that is tough to get really good because there's a lot that needs to happen to kind of change a face, to represent emotion, and also to then synchronize that with whatever is the kind of appropriate emotion to display. But yeah, there's a... It's an ongoing bit of work. Hermes Frangoudis (14:31) No, it makes sense. It's like, it's one of those really challenging pieces because everyone's face moves a little bit differently and there's lot of nuance into that, but getting it close and from what I've seen on your avatars are really awesome. Richard Bowdler (14:44) I mean, also just to say, you know, get over a bit of the tease, what I'll do, and it would be cool if I'll share with you a link to a repository of avatars that people can go and play with. Hermes Frangoudis (14:56) That'd be epic. Yeah, that'd be awesome. Richard Bowdler (14:58) It's awesome. And that's a combo of.. I think it is nicely, I'll make sure that there's a bit of a mix of both video-based avatars. I'm one of them. There's a clone of me. Hermes Frangoudis (15:08) Nice. Richard Bowdler (15:10) It's me sad in my home downstairs. I've done some fun experiments like, you content experiments where I've done it like recorded me talking to AI me, which it's always a slightly creepy experience. Hermes Frangoudis (15:24) Very uncanny valley. Richard Bowdler (15:26) Yeah, they're like, ⁓ dude, what? Hermes Frangoudis (15:30) So speaking of like implementations and use cases of it, what industries or applications have you guys seen the most impact from these sort of Richard Bowdler (15:37) Mm-hmm. Yeah, I mean we are industry agnostic, it's worth saying, as a technology. People build businesses on top of our tech, so they take it, they run in their own direction, in their own vertical, and that's cool. That is exactly what we want to enable. Where we see a lot of interest in application, so I mean, broadly, as with any kind of chat functionality, it sits underneath an umbrella of customer service agent, right? But then if you kind of start slicing into that, that looks like in healthcare, patient after care. In customer support, kind of frontline customer support in e-commerce, both as a sales agent and genuine customer support. And then like in the world of product, we have this baked into our platform, which you'll see if you can kind of sign up. We have an avatar that is an onboarding agent inside the product. So you go into the avatar creator and you have an avatar saying, "Hey, you're now in the avatar creator." Blah, blah, blah. "This is how you need to create an avatar." Slightly meta, but we're seeing quite a lot of interest in application where, historically, in e-commerce and in product onboarding, where you may have a chat function in the bottom Hermes Frangoudis (16:54) Mm-hmm. Richard Bowdler (16:55) whatever bottom right corner of a website, people want and see value in there being, hey, here's a visual representation of where previously I just had some text input. And then also in where because of a lot of my background in learning and personal development is in the learning and development space. So experts, coaches, advisors wanting to use our technology to leverage themselves and kind of clone themselves, which you can do with a knowledge base and custom GPTs, this kind of thing. But then also take it that one step further and create a visual clone of themselves that their customer base, their advisor base can have access to 24/7. For me, that is like, I mean, Hermes Frangoudis (17:46) Huge. Richard Bowdler (17:47) the appetite is huge there. Huge. And the industry is just vast. Hermes Frangoudis (17:52) Yeah, because when you think about it, these are people that have figured out how to successfully build a business around themselves, right? And the biggest constraint is probably their time and their ability to meet with these people because to train someone else to do this exact same thing is sometimes a risk, sometimes just not doable. There's not the right people. So being able to have that clone of yourself, I feel like is huge for those sort of creators. Richard Bowdler (18:14) Yeah, I think it opens up kind of quite an interesting space, where you are both, it's like in the sort of information and advisor expert space. Like there seems to be for a lot of people a barrier or a fear to, you know, they feel like, I mean, we were talking about this before we kind of came online, came on air, giving away the secret sauce. It's like, man, give away the secret sauce. Give people access to the secret source, people will still want to pay you to go and implement the thing because they don't have time. I mean, we're talking about the open source community and the business models that sit on top of that. I kind of see the same thing playing out, but in the information and expert space. It's like, man, open source your stuff, give people access to you 24/7. Create, you know, AI Hermes, like go. What that does is democratize access to expertise. And, you know, we're already seeing it with all of the language models, a sort of democratization or like an infrastructure layer of access to intelligence. I think it sits on top of that and it kind of slices into the verticals. Like democratize and, you know, give open access to expertise. So yeah, it's a really exciting, it's huge, it's really exciting time. Hermes Frangoudis (19:32) Huge. In terms of the customers, right? Like what feedback have you been hearing in terms of like the effectiveness of the avatars in these sort of real world applications? Richard Bowdler (19:46) Yeah, yeah, I love talking to customers. So I was asked earlier today, like, where do I spend most of my I spend most of my time, it's across probably three areas, but most of the time talking to customers and figuring out like, how do we get the product and exactly to work for them. Then, working with partners, so like yourselves, like LLM, you know, big LLM companies, voice providers. And then getting on podcasts and talking to people like this, cause I love doing this. It's really interesting with speaking to customers, I'll give you an example. So one of our case studies is working with a government in India, a state government. And they wanted to provide access to healthcare to an illiterate population. What we did, we worked with a local healthcare authority to set up screens around a village so that the illiterate population could then walk up to a screen and speak to an avatar in Hindi and have a real interaction and gain access to healthcare knowledge and healthcare understanding and be signposted to how to go and seek the appropriate healthcare. So the kind of feedback that we get is, I mean, like it's a transformative technology, right? When you can, literally transform someone's situation from not being able to access something to then being able to access it and transform whatever is going on with them from a healthcare perspective. I mean that's that's just huge. Hermes Frangoudis (21:39) That sounds epic, like the fact that people no longer have to rely on their ability to read and write in order to access knowledge, right? Because that's been a significant barrier of entry and we know across the world there's significant pieces of population that just aren't in that position, right? They can speak, they're able to communicate in that sense, but the world just shuts down when you talk about books and writing. Richard Bowdler (22:03) Exactly. And this, you know, having natural language be the interface is a game-changer. And then having an avatar visual front-end as the kind of the Hermes Frangoudis (22:13) Humanize it, right? Richard Bowdler (22:15) Humanize that, absolutely. And to draw people to that as well as have people feel comfortable interacting. I mean, it's a massive deal. I mean, when we went into this project, I started digging into the figures about illiteracy globally. You would not think it, but have a guess of how many people globally, how many adults globally are Hermes Frangoudis (22:39) I'm gonna I'm gonna need your help on, in terms of super surprising or just higher than you'd probably expect. Richard Bowdler (22:40) You ⁓ kind of large. Slightly higher than you might think. Hermes Frangoudis (22:46) I'll go with 25%. Richard Bowdler (22:50) Not far off actually, of the world. I mean, yeah, at 800 million. So kind of directionally, you know, getting up to a billion of the adult global population. So maybe that is about a quarter of the global adult population. Astonishing, I'm not suggesting Hermes Frangoudis (23:09) a little Richard Bowdler (23:11) just having interactive avatars is a panacea to illiteracy. I think, you know. firm believer in... Hermes Frangoudis (23:17) But it can open up access to so many things that were traditionally inaccessible or people had to bring someone else to help them through this process. They can now do it themselves in the same way that they communicate with the people. Richard Bowdler (23:28) Exactly. Hermes Frangoudis (23:30) And I feel like that's just the power of conversational AI in general. And then you're putting a face to it, which makes you feel even more comfortable. Like you're talking to something, not just like a dot on the screen that's like visualizing your sound waves, right? Richard Bowdler (23:43) Yeah, Yeah, yeah. Hermes Frangoudis (23:45) It kind of moving into the competitive side of things. Like when we first started, you mentioned there's some other avatar players in the space. How does Trulience really differentiate itself and like Richard Bowdler (23:49) Right? Hermes Frangoudis (23:56) what do you guys do to stand out? Richard Bowdler (23:59) Well, I'll tell you what it is and I'll also tell you a bit of the backstory to it as well. So standout differentiator between us, there are a handful of comparable, in terms of look and feel, avatar platforms out there. They, to my knowledge, none of them render avatars how we do. So we render everything client-side. So it's all we're essentially leveraging the compute power of the end user. So we render the avatars in the browser in simple terms. Now, what that means is it's pretty cheap for us to do that compared to if we're needing to spin up a new server for every single instance. Now, there are use cases where it doesn't matter, right? So We have projects where we've had avatars in the hologram, holo boxes in shopping malls as a meet and greet direction finding person. You're only ever going to have one person interacting with that at a time or it's one instance, which is fine. Costs don't spiral. But as soon as you want, let's say, to have an avatar in an app or on a website, what have you, and you start to generate multiple concurrencies, tens, hundreds of users hitting a landing page and wanting to have their own conversation. That becomes prohibitively expensive. That simply does not scale. Now, the reason, I mean, this kind of comes down to bold, calm leadership from our CEO, from Marek. So we had a deal fall over because we advised that what they wanted, this is before we re-architected our platform, would not scale affordably. And the deal fell over. I mean, because we said it's not gonna work. And we took the decision then, and huge kudos to Marek for this, to rebuild the platform from the ground up so that we wouldn't face that problem again and that we had an inkling that it would prove as a competitive advantage. So what's interesting now is that we can see competitors and we get customers coming to us from our competitors. And it has been an initial signal that we saw when we first got our first few customers going like, yeah, but it doesn't scale, right? Talking about competitor X. And we're like, yeah, we know. That's why ours is a kind of no-brainer in a way and kind of genius to have pivoted in the architecture of the platform to do that. And then now we've tapped into that scene and it's more and more and more and more. And what we've seen competitors do is, because look, we're in the game where we obviously, I'm head of growth, I do competitive intel. Where I have previously had access to interactive avatars with some of our competitors, we've started to see that become gated further and further back into higher tier packages, like enterprise packages. It's just because they can't afford to run it. So we think we probably have maybe a couple of years of a headstart. Who knows? But that's our... Let me give you a kind of order of magnitude. Maybe we're in the orders of magnitude 1,000, 2,000 times cheaper to run at scale. Obviously, depending on the scale. But the more you scale, the cheaper the difference in working with a platform like ours running in browser. So it's kind of a big deal. Hermes Frangoudis (27:42) That's massive. The orders of magnitude are massive. So you mentioned like you have X head start, right? You guys learn the hard way about scale and that's the startup way, right? Like that's what helps you find that piece that really skyrockets you. So now that you are so far ahead, let's say, how do you keep that, that gain, right? Cause you don't want to be the hare that Richard Bowdler (27:43) Yeah. Hermes Frangoudis (28:05) gets really far ahead and gets lazy, you want to be almost like the tortoise that's just like, going, right? Richard Bowdler (28:10) Mm-hmm. Yep. Yep. I think we keep doing what we're doing, right? So my focus, like I said, where I spend my like spending my time, customers, making sure that we have super happy customers and that the product is working really well for them. Because like when you have that, that becomes your marketing. But it's kind of pretty cool. Hermes Frangoudis (28:29) A word of mouth is like the best marketing. Richard Bowdler (28:34) Yeah, and without that focus, then there's no point. You're filling in a leaky bucket. I can very much focus on our own backyard for one. I mean, look, without me going into all the ins and outs of the sort of slightly geeky technical growth stuff. There is all of that stuff, but I also like to keep it simple. Focus on what we have. Focus on our customers. And I'm a huge believer in working with really solid partners, which we are we're very lucky in that we kind of have a roster of super high caliber kind of world leading tech partners that we are going to market with and coming on shows like this and speaking with. Hermes Frangoudis (29:16) Appreciate it. Speaking of partnerships, you name drop a few? Or not right now, it's okay if you can't. It's all good. Richard Bowdler (29:26) There are some that are already out there in the public domain and then there are a few that I would love to mention. When I come on the show next time, I will. I definitely will. Hermes Frangoudis (29:36) No worries, mate. Don't mean to put you on the spot like that. So we'll quickly rotate around and go over to... Richard Bowdler (29:41) No, no. No. Hermes Frangoudis (29:49) So earlier you talked about go to the website, log in, set up your avatar. Is it all drag and drop or do you have from like the technology side, like APIs and things that people can connect with at a programmatic level. Richard Bowdler (30:02) Yeah, both of those things. You don't need any, you know, don't need to plug anything in. We have a kind of ready-made off the shelf a user, all of our accounts basically into the LLMs and what have you. Yeah, also if you want to get a bit more geeky and kind of get a bit more fine-grained control over things. Yep, we have APIs that you can plug into. Also, with any avatar that you can create, just each one has its own iframe, bit of code, so you can just embed any avatar in an app or in a website through our SDK. Yeah, you can get it as deep in the weeds with the control as you want to or like super non-techy just tap and like press some buttons and that's it. Hermes Frangoudis (30:52) So it's It's like the quick deploy or if you want to control it all, you have those different mechanisms and layers. Richard Bowdler (30:58) Yeah, we have customers that run the gammut, that sit at every point along that spectrum. And yeah, we kind of cater to all of that. There's some super deep documentation on our site and the front end, the marketing site. So you can get into that if you are so inclined and that's something that you do want to do. It's kind of a neat platform. Huge props to the dev team, of which is not me. I'm the person making noise about it. But huge props to the dev team for getting it to this point. Hermes Frangoudis (31:27) Super cool and it's huge to allow to have customers like the different levels of control because everyone's application and use case is really different in that sense and some people just don't need that much detail, right? Like I need this that does this very simple thing I'll pass it here we'll do that and kind of walk away, right? Richard Bowdler (31:28) Thank Mm-hmm. Yeah, and you know, look, when we partner with, I say partner, I mean, clients, but we view it as a partnership. When they are kind of massive in their space, such as like large e-commerce platforms, and they want something bespoke. They want deep integration. They want full control. That's great. We go there. We have the firepower, we have the technical know-how to do that. And then, right down on the other end of the spectrum, if you have, like I said, say a one-person shop financial advisor who wants to clone themselves, they can do that with pressing a few buttons. So yeah, it's super cool. It's such a fascinating space to be in. The technology and the market aside. But also, you know, thinking about the ethics and the philosophy, like, you know, I spend time with my head in that space as well. Because I think you really need to with building something like this. I think building in the AI space, it behooves you to at least think about it. And also, you should have guardrails around stuff that Hermes Frangoudis (32:51) That kind of stuff has to be part of the conversation when you're developing. Otherwise, it can go very off the rails very quick. Richard Bowdler (32:54) Yeah. I've seen this in not direct competitors, but in adjacent categories. And it's not good. It's not pretty. I mean, just thinking about the education space, I'm a huge believer in tech for good. I'm a father of two. I've also been a kind of campaigner, activist in the social movement. Across the UK, to look at and understand, try to help raise awareness and understand the impact of not AI so much, but social media, early access and kind of prolonged exposure and access of young kids to social media and smartphones. And it fits within the broader conversation of technology in society, technology in humanity, and what are the pros and what are the cons? And I'm very clear with my kids, I want them to use AI. I get them to interact with AI. Because we have lots of different types of avatars in Trulience, I often get my kids to say, "Hey, come and talk to Einstein and have a chat". And it's super cool because you see it's just like, yeah, there's an avatar. For me, it's like, but look, it's seriously cool. But I'm super clear with them that this is not a human. Like it looks really crazily human but you can know the difference. I'm just really clear on that point. And I think there will come a time where... it will be legislated that there needs to be indicators as with advertising standards agencies, right? You need to say if you're promoting something like on an Instagram profile, whether it's a promoted thing, if it's an advert, think the same will and should come through into the AI space and avatar space of like, are you interacting with a real human right now? Or are you actually interacting with a language model? I think that's super important. Hermes Frangoudis (34:54) No, that makes sense. It's same. I have two kids and they love to, especially with the conversational AI that they can like speak to it and it speaks back and they can ask all sorts of questions. I've set up a couple demos for them that are very guard railed for their age groups, but they have a lot of fun with it because they can ask stuff and these are questions that like I would normally have to go and look up things, right? And so the future is very bright for them in terms of having that access, having it Richard Bowdler (35:08) Mm-hmm. Mmm Hermes Frangoudis (35:22) basically, at their fingertips versus when we were kids, there was maybe a few thousand websites and that was kind of the internet. And it slowly expanded and just kind of hit that inflection point where you can never visit every page on the internet again. Richard Bowdler (35:28) Mmm. It's a different world from Encarta to interactive avatars plugged into LLMs. Hermes Frangoudis (35:44) Yeah, we had to go, even like the old school school, go pull off the encyclopedia from the bookshelf and like look up and hope that you could understand the way it was explained. Richard Bowdler (35:51) Mm-hmm. Yeah, yeah, yeah. Hermes Frangoudis (35:56) So thinking of the future and like where these things could go, where do you think the technologies developed by Trulience can really kind of like evolve? I won't say five years because in the AI space that's insanely far ahead. So let's look like six months, a year, two years. Where do you see that going? Richard Bowdler (36:15) Yeah, see. mean, I mean, for us, the technology is evolving and like, you know, that's what the team is doing on an ongoing basis, improving lip syncing, improving facial expressions and emotional nuance. It's like continuous refinement. At the same time as, you know, like I say, we have had significant shifts in techniques, both for the kind of affordable scalability piece and also shifting from the sort of slightly clunkier, like CGI into the visually super realistic. I don't think there's that much more to do on the visual hyperrealism side of things on a kind of foundational level. There's the nuances, like I said, lip syncing and so on. But then really it's like, let's scale. Let's distribute and drive adoption. So that is kind of where we're focused for the next 6 to 12 months, obviously longer. I mean, I'm saying this. Hermes Frangoudis (37:18) Yeah, but in the short long-term, there's going to be a certain level of growth and change. So it's really cool that you guys are looking at like, those nuances and getting it out there because the technology is really good, right? And while we have seen it here and there, it could apply everywhere. Richard Bowdler (37:24) Yeah. Yeah. From a, you know, the sort of honest appraisal is I think that we need a period, perhaps of narrowing and narrowing focus because, you know, with a sort of pure go-to-market strategy hat on of, you know, sort of classic thinking, crossing the chasm kind of thesis is always go an inch wide and a mile deep. I think we will possibly come to a point where we want to go after, pick our beachhead and nail it and dominate in one space. And then what that does is it earns you the right to then go after others and with more firepower, with more ease and so on. So yeah, that's kind of my thinking from my kind of growth and go-to-market perspective. And like I said, the tech is gonna keep evolving. Hermes Frangoudis (38:29) In that sense of the evolution of it, how do you really see these AI avatars fostering more natural HCI, Human-Computer Interactions and that evolution? Richard Bowdler (38:39) Mm-hmm. Yeah. Yeah, yeah, yeah. So I had this very conversation at a friend's house on Sunday. It was a neighbor's house, a friend of theirs, works in language technology. And I think there's a kind of, barrier to adoption and a kind of a barrier or a line, a threshold that people will start to cross. And I kind of likened it to now in the world of transacting, people pretty much across the board will make payments online. It is so normalized that you don't think twice about it. Your bank card details are probably everywhere. There's fraud risk and whatever associated with that. There will always be risks and increased risks with new technologies. But where kind of pre-PayPal, really, online payments existed in people's minds to where online payments now sit in people's minds, we are a sort of pre-PayPal era. when it comes to human-computer, human-avatar interactions right now. I think people for the general population sort of, it's either so novel that it's kind of like, this is a fun thing. Or it's so novel that it's scary to a lot of people. I mean that, for the most part, a lot of people that I deal with are, you know, they want this technology to exist. But there are loads of people out there who I'm not speaking to who have no idea that this kind of technology does exist. So where I see the future of human-avatar interaction as is people going on a journey at large to become comfortable. People will inevitably become comfortable interacting with LLMs. And I mean, they are already like with ChatGPT, but interacting with humanoids, avatars and that that is a a place that people will move to. Hermes Frangoudis (40:38) No, it makes sense. The PayPal was one of those like game changing technologies that really put the seat in people's mind that you could do this securely and have it actually work. And there's APIs for it. So similarly with your avatars, there's a great opportunity for adoption still. And I totally hear that. Like we're very early in this space for developers that are watching and looking at this technology. Like it's not too late. It's a very much like... start adopting it now because maybe in a year still won't be too late, but you'll be significantly different perspective than where you would be now. Richard Bowdler (41:13) Yeah, for sure. Hermes Frangoudis (41:13) So I appreciate your time and as we kind of come to the end of our discussion here, I do have this little out-of-the-box question I ask everyone. Well, if you weren't working in avatar technology, like what area of AI would you be passionate about exploring? Richard Bowdler (41:31) Mm-hmm. Yes, that is a good wild card. So I have a background in human memory development. So in a former life, I went down a rabbit hole of going and doing memory competitions and these sorts of things, which is very fun. Also, run a project dealing with a human-computer interaction, but with neurofeedback. And I think there's an interesting intersection there with what is now possible with AI and leveraging AI in the space of human memory development and with the intersection of a non-invasive neurofeedback kit. And seeing what comes out there. I mean, the project that I ran before, was wanting to output audio from brain signals. I mean, how it actually translated was a pre-mapping of sort of audio escapes that you could navigate through by sort of controlling and navigating through different emotional states and having that correlate to outputting different audio. Super cool, super fun to run. But yeah, I'd probably possibly be spending time doing something in that arena. Hermes Frangoudis (42:52) My mind is just blown. That kind of stuff is super cool, super interesting, and I think very much like the frontier bleeding edge of that human-computer interaction. Hermes Frangoudis (43:07) The thoughts in your head actually controlling things. Richard Bowdler (43:10) Yeah, it's my happy place right on the front. Just doing weird stuff right on the edge. Hermes Frangoudis (43:17) It's awesome. Well, Richard, I appreciate your time today. Really amazing conversation. It was a pleasure having you and hearing about Trulience's real-time interactive avatars, and we encourage everyone to go check them out. Anything you want to say before we head out? Richard Bowdler (43:32) Super appreciate talking, the opportunity to come on the show. There will be a link, I'll share with you, Hermes, so it'll be somewhere around this video, wherever it goes out. And on that link, there'll be a link to Trulience.com. You can just click and go and sign up for free and play. There'll also be a link to some of the pre-built avatars that are already up and running. So you can just click and play around on some of those, and they are. It's crazy. Yeah. Hermes Frangoudis (44:03) I'm looking forward to it we'll definitely share those once we post the final edit of the video. I do want to thank our audience for watching along with us in real time and helping us kind of grow this conversation in the conversational AI world. I do invite everyone to like, subscribe, and share. We'll be bringing you conversations hopefully every week. So see everyone soon. Richard Bowdler (44:26) Ace. See you again. Cheers, Hermes. Hermes Frangoudis (44:29) Cheers.