With over 10 years of experience in B2B sales for innovative IT solutions, John Son is Global Marketing Manager at DeepBrain AI, an AI startup with top-notch video synthesis technology.
He has a bachelor’s degree in mass communication/media studies and fluent in Korean and English.
His core competencies include new business development, sales & marketing, project management, and generative AI, which is the field of expertise of DeepBrain AI.
The integration of AI avatars into corporate branding and communication strategies marks a significant shift in how businesses engage with their audiences. These digital personas, capable of delivering personalized and consistent interactions, are not just futuristic concepts but are becoming essential tools for modern enterprises.
John is at the forefront of developing these innovative AI Human solutions.
What you will learn
- How AI avatars are redefining customer engagement and the potential they hold for enhancing corporate branding.
- The role of AI technology in creating dynamic and interactive digital personas that can embody a brand’s identity.
- The future of corporate communication and the impact of AI avatars on the way companies connect with their audiences globally.
- The ethical implications of AI use, including the need for responsible content creation and the company’s commitment to building tools that detect deep fakes.
- The importance of combining AI with human expertise, emphasizing that AI should serve as an enhancement rather than a replacement.
Transcript
Jeff Bullas
00:00:04 – 00:00:51
Hi, everyone and welcome to the Jeff Bullas Show. Today I have with me, John. With over 10 years of experience in B2B sales for innovative IT solutions, John is Global Marketing Manager at DeepBrain AI, an AI startup with top-notch video synthesis technology. He has a bachelor’s degree in mass communication/media studies and is fluent in Korean and English.
Jeff Bullas
00:00:51 – 00:01:16
His core competencies include new business development, sales & marketing, project management, and generative AI, which is the field of expertise of DeepBrain AI.
The integration of AI avatars into corporate branding and communication strategies marks a significant shift in how businesses engage with their audiences.
Jeff Bullas
00:01:17 – 00:01:58
These digital personas, capable of delivering personalized and consistent interactions, are not just futuristic concepts but are becoming essential tools for modern enterprises. John is at the forefront of developing these innovative AI Human solutions.
And welcome to the show, John, it’s actually an absolute pleasure to have you here and you’re dialing in from Silicon Valley and I’m dialing in from Sydney. So, John, what brought you into this sector? A I? Is there anything you were curious about? Did you want to work in this, you know sector of Silicon Valley? What led you here?
John Son
00:01:59 – 00:02:43
First of all, thank you for having me, Jeff. Um It’s an honor I’m dialing in from Palo Alto. Yes. Um I started off my career in LG electronics companies. Um It’s a manufacturer with a lot of hardware. And luckily I was stationed in A B two B uh sector where I was able to sell laptops and air conditioners. And then after a full experience of selling the boxed products, I was always interested in um associating with technology solutions as well. So I had a chance to dig into A R solutions a little bit. And then while I was doing that, I met deep brain A I um creating avatars and
John Son
00:02:43 – 00:03:09
they just looked real, to be honest with you. I’m more of a late bloomer but I’ve been with the brain A I, a little less than four years now but after seeing those avatars, I just felt I’m gonna work for that company. It’s so exciting. I wanna actually see how it’s done. Um Talk about how it’s done and that’s how the journey started with deep brain and myself.
Jeff Bullas
00:03:09 – 00:03:27
Cool. So Dick Brian’s been has been around um Sorry, I’m calling it the wrong name all the time. So uh Deep Brain I has been around since 2016. So what was the initial uh problem they solved? Uh back then? And how did Deep brain get going?
John Son
00:03:28 – 00:04:04
So after I was introduced, everyone was kind of introduced, interested in creating their own chatbots. Um Eric, our CEO um is a software engineer who started this company was also interested in creating its own chat bot called um Well, we started off our company as Money Brain. So um our chat bots were oriented for financial industries. He had some experience there with the networks there. He understood what the problems were. Um He wanted to solve some problems we facing the customer support
John Son
00:04:05 – 00:04:30
and having to build our own chatbots. Um We were able to solve some of the major problems that the banks were facing. And luckily we were one of the uh providers of a big project of one of the biggest banks in Korea and that’s how we kind of um established our business and was able to fall over to follow our vision on having avatars.
Jeff Bullas
00:04:31 – 00:05:09
So this is 2016. So this is basically six years before Chat G BT stuck its head up above the parapet of A I and said here I am and it was so A I went from like, you know, a geeks play thing to more a, you know, basically total humanity has basically discovered it now. And so has from that start did 2022 with Chat G BT being released. Did that change the game for the deep brain?
John Son
00:05:10 – 00:05:51
Actually, it did, I think I have to introduce a bit more about our solutions before I dive dive. Um A studios is our A I video generator platform, which the outcome is a video. So you have an avatar presenting and the outcome of the video um that’s called A I studios. And when chat to T came in, it impacted more towards our A I human side before chat PT four large language models were actually popular as it is right now. We built our own static chat bots. As I told you, we had experience and specialties with chat bots. So we’ve been doing good with specific sectors like finance
John Son
00:05:51 – 00:06:30
areas, retail areas, creating customized um chat bots so that they can use them with their customer facing environment. But after chat GP T was introduced, the large language model was introduced, it’s familiar with everyone and we didn’t have to create a complex version of a chatbot, but to just use and leverage the large language models that’s out there, which impacted a lot of what we do, how we do and the interest that was gained from our potential customers and our existing customers as well. Tom brought up many businesses that we can work on.
John Son
00:06:31 – 00:07:11
I did start with A I human, but also with A I studios itself, we need scripted text or our avatars to present an idea. So scripting something has always been the basic thing to do when you’re starting with A I studios. But with the introduction of Chat G BT, um this was also helped with A I. So we had the fundamentals of our A I avatars and how the programs should be learned. But with the introduction of Chat CPT, we had a boost of introduction on new ways to leverage A I to our products,
Jeff Bullas
00:07:12 – 00:07:44
right? So essentially with large language models, they learn from a whole just basically billions and trillions of points and information. Uh So essentially you were almost having to write the scripts before and give it information. Whereas now the large language models just learn from all the information out there, whether it’s out on the web or whether it’s within the company or business. So a little bit, let’s dive into that a little bit in terms of. So
Jeff Bullas
00:07:46 – 00:08:32
essentially you are now how do you train your A I avatar and A I studio. How does your platform train on content both within a company? And does it actually go and train itself on the larger web to actually add to that? Because what’s happening with larger English models and training is that it’s very expensive. You’ve got to do a lot of data hosting. Um A lot of training and the cost of I’ve heard the cost of actually chat G BT per user is actually far excessive, far in excess of actually what they’re making at the moment. So how do you try your avatar? Is it a company? Is it the web or is it a bit of both or is it just, I’ve been intrigued about that.
John Son
00:08:32 – 00:09:26
So our avatars are an in -house developed algorithm that we have. Um we have many types to create avatars. Um Studio Avatars, custom avatars using the web camera or uploaded videos. Um You can even create an avatar just by using a photo. That’s something that we do internally within ourselves. Whereas I think the question is more towards how we set up the LLM, how we modify it, customize it. Um I think that’d be the bigger question from you. Um Just keep it simple. Um Lots of the companies that we’re working with, the big companies that we’re trying to work with, with humans, have their own um cha G BT type of sort. So either they’re working with Cha BT or they’re working with I BMW um
John Son
00:09:27 – 00:10:20
autonomy there. I can’t think of lots of companies at the moment just now, but, you know, there’s lots of them out there who bring in a bulk of information and create custom um chats for them. So it’s customizable. The large language models are now customizable before we had that. It was a controversial um idea on whether the LL MS would support our customers better or just having the static chatbot um work better because with lots of power, um the internet usage and the computing power that’s necessary to have the LLM it did slow down the latency, which um takes down the customer experience, which with the avatar part, customer experience is like mostly all about it. So
John Son
00:10:20 – 00:10:59
um there were issues on it but as time went by and it’s been almost 23 years towards more mature chat EPT more mature large language models and the model uh in the industry. And we’re getting much faster responses and we have more flexibility on customizing the language models as well. So we don’t do it ourselves, but we work with partners, we work with customers and we try to define what their needs are and wants are and try to narrow down so that we can leverage the data as efficiently as possible.
Jeff Bullas
00:11:01 – 00:11:53
Yeah. So let’s ask some more questions. So we actually delve a little bit deeper into this. So for example, I’ve got a website that I’ve been doing with a lot of content about digital marketing, social media, marketing, digital entrepreneurship. Now, we’re learning how A I tools for business and how to use it for basically assisting us as humans, human companions and so on. If I wanted to create a training course, I wanted to provide education or to my potential customers. How would your avatar, would your avatar one be able to do that to? How would you approach that? Do you hand over all your content to the website? We’ve got about 67 million words, 223 podcast episodes and thousands and thousands of articles. A lot of them are very long.
Jeff Bullas
00:11:54 – 00:12:06
How would I approach that? If I wanted to create an avatar that actually people could ask questions from within, you know, our area of expertise. How would you approach that project?
John Son
00:12:06 – 00:13:00
So it would have to start by defining if it’s going to be an interactive um conversational um information that you’re gonna toss on to your users or it’s going to be lectures or presenting the ideas on a, in a type of videos. So basically, when we talk about education, when we talk about corporate educations, corporate trainings or even um college universities, um they already have tons of academic um researches or lecture data that their professors are lecturing daily, but it’s just taking up too much time to do that every time. And um we have introduced an efficient way to do it by using our A I studios. Having to create uh
John Son
00:13:00 – 00:13:50
uh lecture using the avatars. We also do custom avatars. So Jeff, don’t get me wrong. We can create you as an avatar and you’ll be lecturing. Um So having to have scripted lectures like a lot of classes and we can modify that into creating a presented video with an avatar with all the lectures that’s actively being used for corporate training. But recently, this is a use case that we have in Korea, we call Cyber Universities. I think you get the point of what it is. It’s an internet university that we call Cyber Universities in Korea. We have uh several of them and we recently actually had an official mou with one of the biggest ones, they wanted to have their CRE uh professors created into avatars
John Son
00:13:51 – 00:14:43
um and have the lectures created with the avatars. And this in the university’s perspective will decrease the cost for every lecture shooting because the professor has to be in a studio, full studio facility with crews that need to shoot the lecture together, which is a team of 5 to 4 members per hour. And if you think about the manpower and uh yeah, it’s expensive. To create an hour-long video is pretty expensive and it’s nothing that I actually knew before. Um They explained it to us because it’s much greater than we actually think it is. That’s one way you can leverage the brain A I. But also if you want to have a conversational A I human on your website,
John Son
00:14:44 – 00:15:31
um We also provide our A I humans in a web P kind of format for you to on board it, it on your website as well. Now, this is going to be much more complicated. You have to have your own um language model customized with us. We’ll have to get all the information from you first to learn it through our um Ior as well using third party large language models. Yep. So it’s a more um depth um steps that we need to take to make it conversational. But that’s one of the reasons to start with our partners. Our clients are starting to use our A I studios to create their contents more efficiently, easy and consistent.
Jeff Bullas
00:15:33 – 00:16:00
Ok. So really, we’ve got one which uh so let’s put it into two boxes. First box is more scripted, recorded, with a lot of structure around it for specific training on a topic because the topic has, these are the sort of things you need to know if you’re going to be, you know, a digital marketer. Here’s the things you need to know if you want to be a psychologist, here’s the things you wanna be if you wanna be a historian, right? Whatever it is, right?
Jeff Bullas
00:16:01 – 00:16:28
So it’s almost like in, within every discipline is structure. And uh then, then you’ve got basically the A I avatar which can be just free range and ask it anything out of your own data where it can basically be, ask anything out of the world wide web. So, uh so you’re saying that uh basically a more unstructured approach is much more complicated because there’s increased learning and it’s more complicated. Is that correct?
John Son
00:16:28 – 00:17:04
Yes, that is exactly correct. And because it’s unfiltered, it also has points where we need to, so it’s hard for us to control what the avatar is answering. Because if the data is right from the web, that’s one thing. Also, our Avatar’s core technology basically is video synthesis. So the LLM models create the text, feed it to our avatars and we video synthesize it and then that’s how our avatar is giving out the answers. So um
John Son
00:17:05 – 00:17:56
if we have the same answer cached, we can uh we can fasten up the process. We can also um save up the usage of data and everything because we have it synthesized that one time using the large language model, it will be constantly synthesizing new videos to provide the answers. So it will have to be in a very good internet uh environment to see it and then synthesize and answer. Um But I must tell you if it is situated in a very good environment recently, for instance, let’s say in our office, um we have lots of demos using other large language models. But one of them is cloud, one of them is um yeah, cha PT as well. There.
John Son
00:17:57 – 00:18:48
My response time is less than two seconds at the moment. Um with searching the web video synthesizing it and then presenting it to the user um is less than two seconds, which it’s pretty amazing. It’s a great improvement. So it’s getting there, everyone’s getting more interested because that was one of the bottlenecks. When first large language models were introduced, the large language model took time to give answers. We took time to video synthesize when it’s added up. The milliseconds do count when you’re asking a question verbally and waiting for an answer to come back. But two seconds um is a dramatic increase of uh customer experience because a second or two people are willing to wait and that’s what we discovered um throughout the years,
Jeff Bullas
00:18:48 – 00:19:12
right? So there’s, maybe , two parts of this equation as well. Number one is you’ve got just making an inquiry on tech such as chat G BT or a chat bot without the avatar. So that’s just really, it’s not simple, but that’s maybe the base layer, let’s put it that way. Then you’ve got the video synthesis to take that and actually turn that into the avatar, which again is another level of complexity. Is that correct?
John Son
00:19:12 – 00:20:10
Yes, that’s the best way to put it. And that’s one of the reasons we kind of position ourselves as a gateway for humans with A I technologies because A I avatars can be in front of actual people um acting as a gateway to all the other um A possible A I technologies. So if the script is fed to the avatars, we can deliver. But from the back end, us being a start up, um we’re able to work with lots of variety of solutions that’s released yearly. We’re experimenting with this and that every day. And that’s, I think, one of the most exciting parts to work with the brain A I because we have our avatars as a gateway. But on the back end, we’re trying to implement this that and that um too
John Son
00:20:11 – 00:20:24
deliver the newest and finest technology to our sectors which we consider financial fields, retail fields, education fields, entertainment fields also. And that’s what we’re trying to do here,
Jeff Bullas
00:20:25 – 00:21:01
right? So let’s go back to the example of a Cyber university. You’ve got very structured training. So let’s say that you want to move on, right? So here the professor asks him anything. OK, for example, rather than being constrained by the structure of the lecture or lectures or the course. Um But it’s good to have a structure, isn’t it? Because that’s, it’s basically like a guide to take you through a, a certain level of expertise or uh you know, a, a pillar of learning if you like. So um
Jeff Bullas
00:21:02 – 00:21:48
as you sort of intimated that this is maybe two different things and one’s more complicated than the other. So let’s say an organization got a whole bunch of data and information and got a lot of training. But on top of that, some things are changing all the time because we’re in a very fast moving A I world now, a very fast moving, you know, global web that is being updated both with shit and great stuff, right? So we, we, this is the other challenge I suppose is trying to sift the crap out of what’s really the gold. So within, do you, when you’re working with a corporation, uh let’s say a bank or something that you want to provide customer support, that would be constrained all within the actual data set of the company itself, wouldn’t it?
John Son
00:21:49 – 00:22:14
Yes, that is correct. And we do comply with the security and privacy regulations, with our company, with our client companies and their um founded countries as well because that’s what we should do. Um And yes, that’s sorry, but that’s, that’s how it’s done. That’s we take that as a
Jeff Bullas
00:22:15 – 00:23:15
cool. So let’s say I wanted to create a training course. OK. There’s a lot of online digital training. Um uh education really hasn’t changed much in the last two or three centuries. Right. University’s around for a long time and everything’s been basically built upon a certain structure and a certain way of doing things and a certain way of verifying. And so whereas today information is changing fast and so does your platform help companies who have a certain level of expertise or have a lot of content? Can it help them structure that, that free range content within that? I suppose boundaries of expertise into courses or basically that’s not done yet because that’s to say I want to create a course. Alright, I wanna train some people. Um Do you guys enable, does your technology enable people and organizations to create
Jeff Bullas
00:23:15 – 00:23:18
courses out of free range data?
John Son
00:23:18 – 00:23:30
I would say partially. Yes. So we started off as just a text video with an avatar. Now, our users increasingly um
John Son
00:23:33 – 00:24:20
After using our studios, people would want more additional features and that’s what we’re currently very oriented to providing. So recently, we did provide a feature called docs to video. It’s as it’s named, if you upload, let’s just say it’s a, you have a powerpoint document that’s already there that you had for a while. You wanted to create a video with it, but you didn’t, you didn’t script it, you didn’t do anything about it. It’s just the PPT file. If you upload it into our A I studios, we automatically use the slides that you have in your PPT to create a scripted app uh presentation. So
John Son
00:24:21 – 00:25:14
in that sense, it’s a yes, but because how A I works at the moment you do will need to modify. We would like to say that upload PPT um export it. You have a complete video, we will do that pretty soon. But this is me trying to be just very honest, we provide the function. This will help you increase the efficiency of creating the actual video. But you will need to additionally add a little touch, human touch so that it’s smooth. But that’s where we’re going. We have a prompt to video, we have a URL to video these features. What we’re trying to do is sometimes people don’t even know what to create, but everyone is trying to upload their own videos and clips
John Son
00:25:14 – 00:26:05
on their social media. They just sometimes just come to a studio and try to build something using the avatars but just don’t know what to write about. Then we, that’s one of the reasons we created a prompt to video, is just give us a prompt for any short sentence and we’ll just create the video automatically for you. That’s one of the features that we provide for our users um which can be um because our avatars are um portrait rights free IP rights for free. We contracted each library model that we have poor commercial uses. So basically a user can come in um to pick an avatar. They wanna um star write a prompt have used like selective templates that we have on our library and generate
John Son
00:26:06 – 00:26:33
and then you’ll have a 32nd or minute long video that you can upload on your social media. So I know this became a little off topic, but that’s what we’re trying to do in A I. We’re trying to generate a good solution for users to create A I generated videos. But also on the way we found out that the users just don’t know how to create good videos. And that’s one of the reasons we’re introducing the features like prompt
John Son
00:26:33 – 00:27:01
to video URL, to video URL to video is very interesting and we personally hope that um it’s more used in e commercial um, industries as well because let’s for instance, say you have a product on Amazon, you have that URL put it in our A I studios, we will crawl the page information, the images, the data and everything and we will create an product video for you. So that
Jeff Bullas
00:27:01 – 00:27:51
Right, right. OK. So, um, I think I’m starting to understand basic issues and also the opportunity a little bit more. Now what with your platform? So for example, um we might have uh the ultimate guide to email marketing. Ok. Alright. It’s 25 pages long. It’s substantial. We do this sort of deep content a lot. Ok. So it’s not like do I look good in a bikini type of influencer video? Right? Which is not a good answer anyway, but the reality is that we do have a lot of deep content so I could give you that URL to that post page and then you could come back to me with a video training essentially out of that. Is that correct? Yes.
Jeff Bullas
00:27:52 – 00:28:15
OK. Very cool. So it would take the images, it would take, let’s say I was presenting it. Uh It would create an avatar of me and then I could talk about that. So we could do like here’s a deep guide for, you know, how to be happy. What are the core elements of happiness, right? And um I’m hoping that, you know, with my increasing age and increasing wisdom, hopefully
Jeff Bullas
00:28:15 – 00:28:53
that my Children and grandchildren could talk to, you know, Poppa over the internet, you know, 100 years in the future and go what’s happiness, Poppa? And up comes a list of how to be happy, which would be pretty cool, right? That means I could actually annoy my Children and descendants for another 100 years as long as I keep the data hosting running. OK. Um So, so basically if I gave you a URL, um we could do that now. So the answer is yes. OK. So we can provide uh and then we can, then we can do that in 80 plus languages. I believe that you guys support me. Is that right?
John Son
00:28:53 – 00:28:55
That is correct.
Jeff Bullas
00:28:55 – 00:29:00
So I could speak Korean. There you go. I actually could speak Korean then. Is that right? Yes.
John Son
00:29:00 – 00:29:04
Again, this is you in a video. But yeah.
Jeff Bullas
00:29:04 – 00:29:38
Yep. So um now there’s, there’s an interesting um snippet that came across my desk talking about A I have a short-term memory. It doesn’t have depth, it doesn’t have long-term memory. In terms if you’re going to chat G BT, it’ll actually learn from what you write in one prompt. Topic is long-term memory. An issue for um A I to keep learning because that’s when it starts to become more powerful. Is that a term that is of interest to you guys?
John Son
00:29:39 – 00:30:30
Well, this is not coming from deep brain A I but personal. And uh as per perspective of this is just long term memories will cost you more. That’s why it is hard for companies to provide such knowledge. Um Deep brain A I doesn’t secure any sort of personal information used with A I studios as you can see unless it’s linked with a large language model that will be um hosted on the customer side. Um If and if this is we’re talking about A B two B customers that own their own large language model, um it will be hosted there. So what we’re doing here is basically again, just being a gateway towards the users. So
John Son
00:30:31 – 00:30:52
it’s a different story from Dra I, but just me personally answering to your question would be just cost effectiveness is the reason that we’re getting short terms, only not long terms. And hopefully the price will go down and we’ll have a more full experience using the A is,
Jeff Bullas
00:30:53 – 00:31:26
Yeah, because the cost of a trial is expensive. The cost of hosting data at scale becomes expensive. In fact, Microsoft invested in uh open A I to the tune of about $11 billion. Uh last year, I think it was uh the reality is that most of that appears to be going into actually the data hosting required to actually answer the prompts of chat GP T. So it’s expensive. Um Yeah. A and I don’t, I don’t think a lot of people understand the scale of the data hosting that’s being rolled out globally. We’re talking
Jeff Bullas
00:31:27 – 00:32:05
in the half trillion dollars mark to actually create data centers. Um So data centers that were popular before are now just gone to a whole new level. And then you’ve got on top of the data centers, you’ve got um the actual chips to actually do the training and actually work well such as NVIDIA. So yeah, it’s, it’s expensive. So um if you’re, do you go out on the web to actually add to training or add to the information that a company maybe providing is, is it a hybrid model between the open web training and also the company’s specific content
John Son
00:32:06 – 00:32:33
for us? Um We do a hybrid, we have a research center in Korea where we have the hardware devices up and running. Some of the GDDX devices are pretty expensive for research, but for our public use and everything, we um partner with cloud companies to host our um I will use them and everything. So, it’s hatred for us in a way.
Jeff Bullas
00:32:34 – 00:33:07
Mhm. Cool. All right. So, let’s lean into different cases. I’ve been banging on a little bit about training and education and content and, uh, so what are some of the most used cos essentially people turn up to you and go, I wanna use an avatar and they’re going, I don’t even know what to use it for. It sounds exciting. I just want an avatar. Right. I um cos I just wanna have an avatar that can have a drink with another avatar at the bar and they just hang out, you know, in cyberspace and just get drunk together. Digitally. No, that’s, that’s not gonna happen. Well, it may actually in the future, but so in terms of
Jeff Bullas
00:33:09 – 00:33:44
we’ll get to this question in a minute. Let’s go straight to that actually. So a lot of people don’t know what they don’t know because this whole A I and chat G bt and chat bots and how you can use them. People are still working that out and they’re overwhelmed. They don’t, don’t know where to start. So let’s go to where you start with using, you know, you know, deep brain A I and the avatar. Where would you do it? You provide guides like this is what you need to do or this is a range of things you can do because you’ve got a hole in some of your use cases, you’ve got here, which I’m just gonna bring up. Um
Jeff Bullas
00:33:44 – 00:34:12
You’ve got training, education, you’ve got e-commerce news media, brand ambassador. In other words, basically a virtual brand ambassador, by the sound of it, social media and customer support. So you’ve got some really interesting uh use cases. Now, do you provide a guide to help people go from confusion to a good idea? In other words, from overwhelmed to actual God, that’s what we should be doing.
John Son
00:34:13 – 00:34:52
So, as I mentioned earlier, that’s one of the reasons we’re introducing a prompt to video URL to video docs. Um We want to provide efficient ways for our users to create videos they want. Now if you don’t have a want, that’s a problem. So we also provide them with lots of templates that they can choose from to create videos. That’s important because seriously sometimes people just don’t know what to start with, but starting with the template, picking an avatar thinking about what you wanna say in this video. It’s
John Son
00:34:53 – 00:35:50
um it’s a flow. So that’s one of the um supportive ways that we um support our users. But I think my question would be more appropriate when I answer it like this. When A I studios first came out, uh one of our first customers were broadcasting companies, they had tons of news articles that they can’t cover within their pro primetime news um time. Um But their news is that’s getting lost um were going up on social media, but it’s not getting the views. So, um what uh NBN is the first one of the first uh users to have their franchise uh anchor created into a studio Avatar. It’s a little different from what we have at the moment. That’s custom avatars. It’s more personalized avatars that we’re providing,
John Son
00:35:50 – 00:36:40
We call them studio avatars. It’s more professional usage. Avatars. We have them, we have the actual people come into our studios, and have a recording session. It takes about half a day or so. Um They read scripts, we use that data, create a sophisticated avatar that’s more like high quality. And these avatars with NBN um are now being used to use the news that’s not touch down on the prime time news, but it goes on to their youtube channels and because the franchise anchor has the popularity and the identify the, the viewers identify her as a famous speaker. Um They get more views and that’s one of the ways broadcasting companies started to use our avatars.
Jeff Bullas
00:36:41 – 00:37:14
Yeah, that’s really cool. So that brings up two things to me. Number one is you can actually produce, you’d have to wait for 12 o’clock news to actually watch it, which is old, which is old school news now, right? We want news now, we want to find out what’s happening with, you know, the US elections and the debate that’s going to happen in a couple of hours, for example. So, so that’s, you know it’s a certain time. But wait, talking is almost overflowing with content that can’t fit into prime time. So this is where it’s basically
Jeff Bullas
00:37:15 – 00:37:41
information uh on demand, streamed to you but done. And this is the other part that’s interesting to me too is that this is a mix of human and machine where the human has credibility. The anchor is basically being replicated as an avatar and then presents that information. So what we have is credibility meets streaming and availability and ease of use. Is that correct?
John Son
00:37:41 – 00:38:41
That is exactly correct. And I do want to add by saying the Cyber University case. Once more, the lectures are created using A I studios. It gives more free hands off time for the professors to research their academic fields more which will enrich the lecture quality again and then it will benefit the students also. So we are, what we’re trying to do is we’re trying, we’re not trying to replace human power. We’re trying to empower them and force them to create a more efficient ecosystem, how the avatars can act on behalf of the real person, providing the real person specialists in this case, more time to do what they’re good at. And then again, feed better information to the avatars so that it delivers to the viewers promptly as it can.
Jeff Bullas
00:38:42 – 00:39:15
Yeah. Yeah, that’s really awesome because I think the challenge a lot of people have is, they think that machine is going to take over? I think that is what we call replacement A I, is your replacement. The reality is that A I should be your enhancement. And what we’re doing is that you might be an expert in your field, you know, the right things to ask, you know, and this is where an expert with the machine at their side and the A I at their side is actually can use that to actually add the latest information on research on, you know, open heart surgery, for example, or something like that.
Jeff Bullas
00:39:15 – 00:40:09
So essentially this is where we get amplification and enhancement of, of the human. And on top of that, the other thing to talk about, which I find fascinating is to do an avatar of the news anchor on a news site, um which then becomes a credibility meets technology and streaming and allows people to get information anywhere any time that can be actually. And, and we go back to the Cyber University thing that actually can be updated in real time as well as new research hits because, you know, for example, there’s maybe 20 universities or 20 research labs or 100 you know, investigating a certain type of cancer or a different treatment for heart surgery or whatever uh or fixing of climate problems, right? The reality is that by
Jeff Bullas
00:40:10 – 00:40:44
using the machine to actually update the expertise, you’re actually really accelerating learning and also adding to the credibility by providing the latest information because, you know, you might have a problem with, I’ll go back to health again that the latest research in America has not been shared with the professor in Australia or career. Whereas there’s some really good treatment. We, in other words, don’t have to die from this disease. You actually, there’s some treatment over here, but we haven’t heard about it. This is where it starts to get rather exciting, doesn’t it? In terms of enhancing the, you know, machine enhancing the human.
John Son
00:40:44 – 00:41:45
If you become my speaker, you read my mind. So one of the things that we slightly covered was the language part. Now we have companies, global companies with lots of operations everywhere in the world and we employ employees locally as well. Sometimes language is a barrier, especially in a retail environment, in a customer facing environment, it’s more localized than the back office. So giving them instructions, trying to train them has been the biggest hurdle for big companies for the longest time and having to create one video which you’ll be updating this one video. And when that video is done, it can be automatically translated into different languages that’s needed will be a time saver. It was a, it’s gonna be a time changer. So
John Son
00:41:46 – 00:42:14
yeah, um that’s what we’re trying to do. We always try to say we’re not trying to replace, as you said, many of the times people just see it as A I replacing stuff, which is I say not true. In most of the cases A I is not still smart enough is what I would like to say, but it definitely helps people in everyday life and it enhances you and that’s what we’re trying to focus on and that’s what we’re trying to do.
Jeff Bullas
00:42:14 – 00:42:54
Yeah. Yeah. This area is quite fascinating and something that I’ve been investigating for a while. In other words, and we, we go into the old areas of creativity as well because we thought that, you know, the machine actually would just do the boring jobs like accounting. Um, I tried to do an accounting degree for one year and I hated it so much. I gave up and moved on to teaching, which I then didn’t enjoy either. So, uh, I became an online teacher effectively by writing. So, I didn’t have to put up with students messing around. Right. I want to teach people to actually want to learn, they’re called adults and they’re grown ups. But anyway, I digress. Uh So
Jeff Bullas
00:42:54 – 00:43:16
In creativity, we’ve had two schools of thought, like you’ve got the creatives that feel that the machine can write, you know, an article better than me, the machine can do art better than me. It can do a video better than me. Then you’ve got, so that’s sort of like, that’s why you had the Hollywood writers’ strike. OK. A I was gonna take over and they just wanted more money as well. So
Jeff Bullas
00:43:17 – 00:43:46
then on the other hand, you’ve got, you know, A I artists and other creators going, I reckon, you know, a I could help me actually come up with some really great ideas and I could produce something really fantastic. That is an intersection of humanity and the machine. Because the challenge I think we have as creators and as humans is we get trapped in our past templates and walls of thinking and we have trouble thinking outside the box, which is a cliche term.
Jeff Bullas
00:43:46 – 00:44:11
But the reality for me is that I think to actually welcome A I as an enhancer and amplifier of our humanity is a better place to start. So you’ve got those who say I’m gonna enhance A I, I’m gonna actually embrace A I and I’m gonna use it to make me better and, and more creative. Whereas the others go oh my job to be lost to A I, I’m gonna cry in the corner. So there’s, it really comes down to the attitude, isn’t it? To A I?
John Son
00:44:13 – 00:45:00
Um I would just personally say yes, but it’s a hard topic to discuss at the moment. But as a person who’s working in the A I field, I hope that everyone feels the way we want them to feel, getting helped by A I because that’s what we’re trying to do. So, um since we’re talking about somewhat controversial stuff. One of the things with deep brain A I, the hardest thing that we had was us being compared with deep bake technologies because we’re creating videos. They think um many of the people think that we are using the deep fake technology, but the basic of technology is
John Son
00:45:00 – 00:46:01
distorting an existing video whereas we’re generating new content and since deep fake videos have been creating lots of issues for a pretty long time and uh creating greater issues recently, um it’s always been a disadvantage for us. So what we decided to do is now we should create a deep, big detector so that people can actually benefit by using our solution to detect if this content was created using debate technology or not. So it’s another thing that recently we’re very delving into. We have the gen A I platform on SAS that people can create contents with. But on the other hand, since we’ve been compared with Deep B so much, we decided to provide a solution that actually detects it for them so that they know if this is a content that’s created by Deep Bank or not. Yeah,
Jeff Bullas
00:46:02 – 00:46:33
exactly. And I think that brings up the uh the, the discussion about ethics and also a use of IR and uh and we can use uh social media as a bit of an example of a social experiment. And uh social media started and still has very few guardrails and the challenge with that is especially starting about 2015 as the algorithms got smarter um basically, the social media platforms measured their success on in how long people stayed on the platforms called engagement.
Jeff Bullas
00:46:33 – 00:47:28
So that was a metric. What they discovered was by putting that algorithm in place. It actually then surfaced stuff that got a lot of people’s attention and kept them online because it drove hate division, separateness. Drama cos that’s what engages people, right? The most base elements of being human were actually amplified. Now, the problem is that that horse has almost bolted, but we may need to actually bring it back to the corral and shut the gate a little bit on this. So there’s a new book out just released today by Yuval Harari who wrote SAPIEN. Um His new book is about human networks and communication over the millennia. And I’d be interested in your thoughts on this. I think creating an avatar of yourself to scale you is not fake. It’s actually just amplifying you anytime, anywhere.
Jeff Bullas
00:47:29 – 00:47:58
But on the other hand, then we come to the problem which Yuval Harari’s worried about, which is basic, which should maybe ban anyone creating a video of anyone else without their permission. In other words, we should, that should not be done. I’d be interested in your thoughts in terms of A I guardrails and uh what you guys are doing and one of them, you mentioned was an A I uh deep fake detector. So you must have a lot of discussions about this in the office. I’m sure.
John Son
00:47:59 – 00:48:59
Of course, it’s a topic that we always talk about. And that’s also one of the reasons. Um We try to provide as many diverse library avatars. Library studio avatars as possible because people still sometimes have frights on creating their own avatars. Um just because it’s not a familiar um part for them yet. So to start with, we started with Studio Avatars and that’s what we’re still providing. We provide 200 plus avatars in different outfits, different races, different nationalities, different languages. Um We try to have as many avatars as possible. That’s number one, number two. Um It’s very hard to just cross check whether they’re using their own picture or not.
John Son
00:48:59 – 00:50:03
Um But the only way currently we’re trying to um bring up uh understanding with users by letting them know that this might affect other people. So um drawing a content with them is actually the only and the best way of having, having a complicated one or two extra steps for them to acknowledge that this um you would have to take responsibility of is the method that we’re taking at the moment. Um It’s, it’s a very hard matter. Um There’s, as you mentioned, like recently, there was a California A I bill that was passed which we’re also digging into. Um but, and it’s a fast changing world with regulations coming up, um, regulations being turned down every like, it’s, it’s like a hand flop. You don’t know what’s gonna be like today and you don’t know what it’s gonna be like tomorrow. But
John Son
00:50:04 – 00:50:23
what we can do as a company and as users, um we’ll have to emphasize the responsibilities that individual person and the company should take on creating such content and us trying to go for the better means than to misuse.
Jeff Bullas
00:50:24 – 00:51:11
Yeah. And, and that’s the challenge we have because it’s moving so fast that actually I’m trying to work out. You know, there’s a phrase which is be very careful what you wish for. In other words, it might be that uh you really want a girlfriend. So you, you meet this lovely woman at a bar and you get together and she’s wonderful for a week and then uh two weeks later, she’s, you know, start a little bit angry saying the wrong things. Um a year later, she’s turned into a horror show, right? Well, he’s turned into a horror show. OK. Don’t wanna be sexist here. But the reality is that uh but you gotta be very c which we don’t know, like social media, you know, it was touted 10 years ago as being an experiment, right? A a huge global experiment.
Jeff Bullas
00:51:12 – 00:52:01
Look for me, it’s still a fantastic technology and used for good, but it’s being used and it’s being baked into the algorithms of um amplifying the worst part of humanity. And this is the thing we don’t know with A I, we’re still, we’re at the early stages what? Cha A I’s been around a while. Ok. Long time. But the consumer use of it, the popular face of a chat G BT has only been here less than two years. And that’s the challenge, isn’t it? Like, how will people misuse this technology as well as how will they use it for good as well? And that’s a burning question, which I think we’re gonna be wrestling for forever.
Jeff Bullas
00:52:03 – 00:52:20
So, um, any other, any other sort of really top interesting cases that people have taken your technology and used it that you’ve been surprised by or thought was a really great use.
John Son
00:52:20 – 00:53:07
Sometimes it’s just amazing to see how many different use cases we can create. So we have business use cases such as broadcasting companies, financial companies and retail. We’ve been to the ones that we’ve been showcasing on our website, I already think it’s a vast variety. But after introducing our custom avatars, which are the personalized avatars that you can custom create on our A I studios. Um, and we sometimes see users upload their contents through youtube or social media and we see them wearing doctor gowns, nurse scrubs or even like
John Son
00:53:08 – 00:54:07
The one that I just thought about was the doc. One of the doctors in his doctorate desk is introducing common sense to people providing medical information and common sense to people in his field. Um That’s something that we hoped uh people use it that way to share knowledge um to the greater for the greater use. But um it’s just interesting. One of the funny parts was um we had uh A I analyst now, this was also to empower the real analyst. He’s a very popular analyst. Everyone’s kind of waiting. He has actually a youtube show weekly and he just didn’t like going on camera and then people wanted information from him. So he created an avatar of himself and he did the show using the avatar. Um
John Son
00:54:08 – 00:54:33
as you were saying, um there are so many use cases that’s not even, that didn’t even come up yet. We don’t know what other use cases there will be. But it’s always as a service provider fascinated to see how people actually leverage the technology and they also inspire us. So it’s always fun to see how our users are actually leveraging A I studios and
Jeff Bullas
00:54:34 – 00:55:32
adding a little bit of a little bit of human creativity to a little bit of uh machine technology. Um And this is where it gets really interesting is that the question I’m intrigued by one of many is that um it’s the intersection of human and machine and allowing us, like for me, I’ve heard someone say that A I almost allows all of us to become a polymath. In other words, understand a whole range of disciplines because you can just ask the oracle chat G BT because the intelligence of the planet and the creativity of the planet is actually recorded in data which is accessible by A I. So essentially, um I can ask, give me some ideas on this topic, uh write an outline on this topic and it’ll actually produce
Jeff Bullas
00:55:33 – 00:56:12
a great outline of what is the best case for, you know, email marketing, right? What are the elements I should be considering email marketing, what’s best use? Uh Give me an ultimate guide outline for email marketing or search engines or digital marketing or social media marketing. The list goes on. So for me, um my attitude at A I is very, how can A I help me be more, how can help me be a better person, how can help me learn faster? Um get up, come up with different ideas, which it does and I continue to be blown away, and overwhelmed at the same time. So, um
Jeff Bullas
00:56:13 – 00:56:42
I’m gonna ask you a question in terms of where we went back to, I’m gonna go back to education a little bit in training because uh it’s a little bit of a personal passion of mine, but more online than in the classroom. But the reality is like if we want to train and do a training course, we could actually go to chat GT and go provide an outline of a training session one hour long um, what topic should I cover? In other words, this is what the experts around the world would be saying . Chat should actually do that.
Jeff Bullas
00:56:43 – 00:57:09
Then you go, ok. Uh, fill it in with information about that, then turn that into slides and then add an avatar on top of that done. You’ve got a training course done, uh, which you haven’t written a course. It’s not sitting there like a cyber university professor, um, who’s actually agonized for 30 years over this, right? Um The machine’s done in 30-30 minutes. Um Is that a, is that a good way to use your product potentially?
John Son
00:57:11 – 00:58:01
Mm We sometimes say it’s a fun way to use our product. Um As I said, this being used as a way to create social media content, having taken out the difficult sophisticated editing features and actually having to be in front of a camera feature is taking up a big hurdle for many users. So um where do you see there’s always a good side and a bad side? But mm what I want to say, uh one of the fun use cases that we actually have uh with A I avatars with education is language tutoring. And
John Son
00:58:02 – 00:59:06
Because in Korean learning English is a big thing and you also have to call English tutors, which is actually just having 10 minute calls every day to use English and having short, free conversations. And to actually learn a language, you have to speak it to learn it. So that’s a very good way to get English to you. With avatars, you have facial interaction, you have a conversation with an avatar which will help you practice language and it will ask you questions, it will answer. When you provide an answer, it can also evaluate you around. Hm, correct you having but also as a user having a target to talk to would just be one of the best things to have. So, creating chatty pity generated lectures, creating videos using that and
John Son
00:59:07 – 00:59:39
selling it as a lecture. Is it ethical? I would have to say no to that. But that’s also a way to, for people to gain. Um, that’s one of their skills that they can leverage. But there, I would also like to focus on many other, more better ways. And that’s again what A I studios is trying to provide, I have good ways to use A I and share it for greater use. But
Jeff Bullas
00:59:39 – 01:00:37
I, in other words, you in an unethical way and uh the trouble is the ethical um, Becomes rather murky because a lot of the, I even, I think Mustafa Suleyman said the web is open range and free to be used to do anything. And he’s actually the A I Head of Microsoft. So that was rather interesting. Um So, uh anyway, so last question, um if someone came to you and went, OK, I want to create an avatar of me, what’s the process with you guys? I know you’ve got different levels of, of uh creating avatars such as bring people in the studio, record them. Can you actually um use online videos of them such as me on a podcast to actually train the avatar um and other videos as well. And then, um so
Jeff Bullas
01:00:38 – 01:00:45
actually answer that question first, then we’ll, then we’ll talk about how do you, how would, how would you work with someone if they come to you and go, we wanna start,
John Son
01:00:47 – 01:01:37
we have for custom avatars to start with, we have studio avatar and the personal avatars, the personal avatars, you can record yourself using a webcam on your computer and also upload a short video file that will be used as you. Um It, we will try to use that video clip and just modify the lip sync of that so that you can provide other scripts for that typical video to speak. Otherwise the professional avatars that studio avatars should be um done in a studio which does take up about an hour a month or even two. Sometimes depending on the quality of avatars that we want to produce. And
John Son
01:01:38 – 01:01:55
yes, so sophisticated the studio avatars, you actually need to be in our studio. You have to have at least one or two recording sessions to create one avatar or the easy way in which we call the instant custom avatar, which is doable with your laptop.
Jeff Bullas
01:01:55 – 01:02:25
Right? Cool. So that’s the first question. The last one I’m curious about is um you’ve got people that turn up confused and overwhelmed going. I just want to create an avatar cos I think it’s a good idea. Then you’ve got someone to come to you with a particular case like customer support avatar or whatever. So in terms of pricing models, I think it’s starting with $24 a month, billed annually and $72 a month bill annually. And then you’ve got enterprise custom pricing as well. So um
Jeff Bullas
01:02:27 – 01:02:50
how, how would, what’s the process of onboarding and creating an avatar? And then things like, how do you, how do they train from? How do you learn from their content? For example, they bring their own box of content to you. Just like, I would have like three or 4000 articles, 7 million words and 223 videos. What’s the process to on board someone and create an avatar in at, at a simple level,
John Son
01:02:51 – 01:03:54
at a simple level. So it’s a I avatar uh A I studio as a tool. We have different tiers depending on the possible usage. You can upgrade your tier as you go. But basically, it’s based on how many minutes of videos that you’ll be creating. And we have personal and team plans where you can collaborate with other team members with those same work spaces. Again, it’s dependable on how many minutes of video that you’re expecting to create. Otherwise the platform is there for you to use, you can upload as many um powerpoints or documents that you have up there. It’s just the end synthesized minutes of videos that we count. Um So basically, if you have a paid account, you’ll have a monthly limit and then you can create as much video as you want.
Jeff Bullas
01:03:55 – 01:03:55
OK.
John Son
01:03:56 – 01:04:00
That’s the easiest way to leverage our A I studios at the moment, right?
Jeff Bullas
01:04:01 – 01:04:05
So is it an onboarding cost followed by the monthly or is it just the monthly fee
John Son
01:04:06 – 01:04:48
Is it on board? Um Is something that we can provide in cases where our users want, but normally just as a SAS product, it’s not included officially. But if you request that is something that we can provide. That’s one of the differentiators that we have with our competitors as well. Whereas our competitors seem to lack a little bit of human touch in their customer services. That’s where we try to um shine a little bit more by providing our CS agents, customer success managers actually become talking to the users.
Jeff Bullas
01:04:48 – 01:05:09
Cool. All right. Um That’s fantastic. And uh the other thing I just want to know is um like just curious. So you’ve been around six years now. What’s the size of the team? And um are you raising any more venture capital funds to fund growth? Uh Just a little bit about the company and where it is at the moment
John Son
01:05:10 – 01:05:45
you bring a um started up from South Korea. Um We have a little under 100 head count at the moment. It’s been bigger, but currently it’s resized a little bit. We have offices in Seoul, Korea, Beijing, China and Palo Alto, US California. So we have three operations, Beijing and Palo Alto are mainly sales and marketing operations. Um Our global functions mostly are managed through the Palo Alto office here.
Jeff Bullas
01:05:45 – 01:05:45
Cool.
John Son
01:05:48 – 01:05:49
Yes. Um That’s
Jeff Bullas
01:05:50 – 01:05:58
ok. So are you looking to grow faster by raising more funding or investing in more technology?
John Son
01:05:58 – 01:06:11
Yes. Um We’re currently funded up to CB um it’s total above about 55 million, 200 billion in valuation at the moment. So, how much
Jeff Bullas
01:06:11 – 01:06:13
is the valuation about a billion, is it or
John Son
01:06:14 – 01:06:17
200 million?
Jeff Bullas
01:06:18 – 01:06:19
200 million valuation? Ok. Cool.
John Son
01:06:20 – 01:06:42
Right. And so we’re currently preparing a PRE IPO and we are targeting an IPO maybe by the end of next year or something like that. That’s, that’s our current plan, nothing fixed, but that’s our road map at the moment.
Jeff Bullas
01:06:42 – 01:07:22
Cool. So John, thank you very much for sharing um your insights and uh what the company’s doing and how to do it. I think it’s been a fun conversation for me in terms of exploring, I suppose the downsides and the upsides of A I and uh what it can do. And I think uh hopefully we’ve made more sense of A I. So people are less afraid of A I so they can use it properly to actually enhance their humanity rather than descend into a dystopian disaster, which a lot of people are worried about because, you know, we’ve got to blame Hollywood for that uh machines taking over the world and spacecraft and everything. So,
Jeff Bullas
01:07:23 – 01:07:33
So John, thank you very much for your time. It’s been an absolute blast mate and uh look forward to maybe bumping into you in California or in Silicon Valley one day.
John Son
01:07:33 – 01:07:45
Sounds great. Just so much. Thank you for having me here. I had so much fun. I hope this podcast is, I don’t know.
Jeff Bullas
01:07:47 – 01:07:57
It’s well for me, it’s been great. I hope, you know, the people who view it in a few weeks actually think it’s good too. But, um, look, I don’t care, I’ve had a fun time, so that’s all that matters really.
John Son
01:07:58 – 01:08:01
All right. Thank you so much. I had so much fun.
Jeff Bullas
01:08:01 – 01:08:02
Thank you.