The Exchanges

Q Okay. Wonderful. Um, so much great thoughts here and, and, and different directions to unpack this, but maybe you mentioned some use cases. So maybe, maybe let's, um, talk about this. So they, there are two ways people can buy the product. They can sell server online or the, or there is an enterprise, uh, kind of, um, uh, version. Uh, how do people use the product in both cases?

A So I, I'd say on the self service plans, we enable a lot of like smaller businesses to make video content. So this is a group of, of customers who generally haven't had any kind of budget, uh, or capability to do video before, right? This can be anything from, I think recently I was looking through some of our latest signups, like, uh, a barber in Amsterdam and one in Brazil, who just making videos for their Facebook page, for example. Uh, all the way to maybe just like a smaller team in a big company who's just like trying out the product at first. But what I generally say is that for where the technology is at today in terms of the realism and capabilities, what we're seeing is that instructional video content, it really excels at, right? It's great for like information dissemination. So you're making a video to someone who has to understand something, learn something, and you want to do that in a better way than just sending them a long word document. So that could be customer support. It could be a customer success. It could be sales enablement. It could be, you know, training or frontline workers, things like that. For those types of use cases, the product really excels. Um, as the technology gets better and better, I think we'll see much more of kind of like storytelling, um, emotional types of video content start to take off. Um, but I think, you know, the, I think the h…

AI assessment note: “we enable a lot of like smaller businesses to make video content”

Q you, how do you think about this, uh, right now? I mean, obviously that's the, uh, the, you know, that could be a good reaction that people could have when they see avatars and imagine that you can make, uh, Tom Cruise speak in your voice and all the things. How do you, what safeguards have you Put in place, and how do you think about the topic right now?

A Yeah. So I think AI safety is, of course, is a huge topic, uh, right now and something that, that we've been thinking about for almost seven years since we found the company on our ethical framework. I think there is, uh, you know, I mean, there's lots of things to unpack here. I think for us, or for me, there's sort of two, you know, angles I look at this through. One is like, how do we secure Synthesia for not being misused by malicious actors? Um, I think that one is, that one is very much around like consent. So we'd never create an avatar of someone without their full actual Consent, right? You can't just go in and upload video images of someone that you don't know. Um, and the second one is content moderation. So that's something we're investing heavily in trust and safety. Uh, basically we take a quite hard stance on like what kind of content you're allowed to create on Synthesia and what kind of content you're not allowed to create on Synthesia. As with anything content moderation, um, we're, we're not perfect. We're always improving. Um, it is really difficult to ascertain if someone is making, Great videos about how blockchain works or they're trying to push you into a get rich quick scam, right? That's hard to do automatically. But those are some of the things where, you know, we go in and actually moderate the content at the point of creation, right? So you can't ev…

AI assessment note: “we'd never create an avatar of someone without their full actual Consent”

Q And for now, the, um, the avatars, like the speaking figures, are based on real-life actors, right? Like you have a whole library of, of, of people, um, and then the idea is to then create synthetically avatars, so each company could have their own, sort of like, Spokes, figure type, type person. Is that the, is that the path?

A There's two streams here. One which is already used by roughly 80% of our enterprise clients today, which is that you can create a real avatar of yourself. That could be someone from the leadership team or management team, for example, or it could be someone who's like a brand representative. This process is something we're working on, on scaling a lot right now. It is quite an easy process today, but we definitely believe that we're all going to have a digital representation of ourselves Um, kind of an avatar of ourselves that we can use for creating video, maybe even Zoom calls, stuff in the future. So that's one stream. It's like taking you and making a digital version of you with your voice. You can speak any language. You can do PowerPoint presentations live. You can maybe even do these conferences one day by just typing in text. And then the other stream of it is what we think of as synthetic humans. This is where you can, some people might have seen the MetaHumans approach from Unreal that came out recently. But this is essentially where you go in, kind of like when you start a computer game and you create a character, you can brand that character, uh, with the logo and put on, you know, a hat if it's a fast food chain or whatever you want to do, and then you can create these kind of artificial characters that can be, that can represent your brand, and that's also quite …

AI assessment note: “There's two streams here. One which is already used by roughly 80% of our enterprise clients”

Q And so that's learning. And so the, the use cases are mostly B to B enterprise, your platform targets, large, uh, corporations around the world. And, uh, so you just to double click on this, uh, some of the use cases are learning and training where people need to, um, learn to be able to do their job and also onboarding. Um, that's, that that's correct, right? Across different industries.

A Yeah, for sure. I think From a general perspective, video is, is a much, much more effective medium to communicate, even more so in a remote world. And what we're seeing our companies use is all these kind of things, which traditionally would be, would be text, right, which could be like onboarding documents, manuals, training and learning. Those are the ones that we're working primarily with right now, but we also slowly started to see the first kind of use cases with external content, like marketing, for example, right, where you also want to have And video assets instead of text assets. And then there's the kind of very interesting cross-section of those two, which I would call customer experience. We see that a lot here, right? So let's say that you're a bank, for example. You have FAQ, help desk articles, and you have a lot of them, because you have a very complex product. For most users, they're simply not just going to read through a long page of text explaining how insurance works, or how credit check works, and things like that. And they're now starting to use videos generated on our platform. And to communicate those kind of things as well. What we also slowly see is not just turning text into video, but data into video. So the big idea that Synthesia is built around, right, is that video production and media production in general is going to go from something that we…

AI assessment note: “Yeah, for sure. I think From a general perspective, video is, is a much, much more effective medium”

Q workflow and collaboration. And then there's another layer around verticalization and user touching, um, kind of like features. Do you, Do you, I mean, would you agree with that? Is that like the, the, you know, one of the most, uh, likely to succeed kind of way of building a company or, or are there, uh, other sort of, uh, potential models for success in, in building an AI company?

A Yeah, I, I think I agree. I think it depends a bit what, what your kind of decide outcome is, right? I think if you're bootstrapping and just want to build a great business for yourself and want to have 15 people employed, probably you could build a great just like wrapper company, but if you're building just a wrapper company, you probably shouldn't be free on the people and raise a lot of VC money, um, that, that might not end out the way that, that you envisioned. Um, I think in terms of building a full stack AI company, I agree. I do think that an added dimension to this is if you really want to build something big, I think it's quite important. That you're not just building a company which is kind of an existing product with some AI bolted onto, right? Because I think incumbents, um, also as easy as it is to build something on OpenAI API for someone who's just starting out, as easy as it is in theory, at least for a big company to do the same thing, right? So I think it's really important to think about the product that you're building, building that AI first, and thinking deeply about what are you doing differently than incumbents? And how does AI change not just like one feature in the product, but the kind of value proposition of the product? So if you take something like customer support, for example, which is an obvious use of LLM, right? It's sort of fairly low stake…

AI assessment note: “Yeah, I, I think I agree. I think it depends a bit”

Q Very good. And Victor, how do you guys connect? Maybe walk us back to the origins of the company and, um, uh, how the, the, the four co-founders are you, Stefan, Matthias, and Dordis connected?

A Yeah, sure. So I'll give you the short version. Um, I think it's a most funny story. It's, it's a long, complicated story. But, uh, me and Stefan had worked together, um, back in Denmark at a venture studio, uh, like, I think it was almost 10 years ago now. And we had a great energy with each other. I think we had the same level of ambition. Um, and, um, so we kind of had a good partnership going on there, but we decided to go, like, two different paths. Stefan went to Sambia. To work in private equity, and I went to London because I had figured out that I loved building things, but I was more passionate about science fiction technology than I was about building accounting programs or kind of business tool type of software. So I went to London and I started working on, um, AR, VR, um, type of technologies, which I'm still very excited about, but I think the market is, is, you know, is, is still growing. Let's put it that way. And during my work there, working with these AR VR technologies, um, I, I met Matthias. Uh, I had also spent some time at Stanford. Matthias spent some time at Stanford, and I sort of started looking at these technologies, um, like face to face with Matthias's, I think, probably most well-known paper in, in, in the space of, um, of, like, what we're doing, and I just got very interested in, like, how these AR, VR technologies, three-D, uh, computer vision,…

AI assessment note: “me and Stefan had worked together, um, back in Denmark at a venture studio”

Q One of the questions that came up in the, in the chat, and maybe for either of you, uh, is around languages. How many languages does Synthesia support and how does that work behind the scenes?

A Yes, we support 55 languages right now, and that is a feature that I think more or less all of our clients are using. And the way that it works that we have a very broad selection of, of text-to-speech voices, which, which drives the avatars. And so on the back end, that's how it works. The, the kind of core technology that, that drives our avatars is that we can take an audio signal, we can turn that into a video. From a user experience perspective, it's Uh, it really is quite easy. You have this text box where you, where you type in the script. If you type it in English, the video will come out in English. If you type it in French, it will come out in French. If you type it in Italian, it will, uh, it will come out in Italian. So that's how it works right now. But I think the very interesting thing about synthetic media to me is that, you know, we're building this videos of this technology, which itself is a very hard problem, but there's all these other technologies which are all complemented. Complimentary and will work as, I think, false multipliers in this space. So everybody knows GPT-free, of course. What's that going to mean for translation? What's that going to mean for automatic content creation? And I think, so I think it's right now it's, that's how it works, but in the future, I definitely foresee machine translation becoming 10 X better than what it is today, and…

AI assessment note: “we support 55 languages right now, and that is a feature that I think”

Q and I first met, uh, that was the term did not exist. Uh, and, uh, almost by definition, you had started the company, uh, you know, probably a couple of years before that. So what, what was the vision at the time? Because in retrospect, it seems like a completely crazy idea to start a, uh, generative AI company at a time when generative AI did, did just not exist.

A Yeah, back then we called it synthetic media. We were hoping that that was going to be the term that would latch on. Unfortunately, it wasn't. We still have some decks where we say genitive AI, like, 19, I think. Should have stuck with that. That'd bring great Versio. Um, so the origin story, um, the very kind of, uh, I guess, like my path to starting Synthesia. So I grew up in Denmark and Copenhagen that that's where the accent's from, um, and figured out in my late teens that, uh, I love building products. Um, and that I wanted to start my own company after doing something like four or five years in the Danish startup ecosystem. Um, but what I also kind of figured out during that time was that I just wasn't super passionate about like building accounting software or business process tools, which was sort of mainly the things I've been involved with at that point in time. I'm a huge nerd in my spare time. I love science fiction. I love the kind of weird, wonderful edges of technology. And I've just always been sort of drawn to, to that in my life. And I want to see if I could Combine that with, uh, also building a great business and making awesome digital products. So to make a very long story, very short, uh, I decided to move to, uh, to London back in 2016. I knew I wanted to start a company. I knew I wanted to do something with deep tech, and I basically just spent nine to …

AI assessment note: “the origin story, um, the very kind of, uh, I guess, like my path to starting Synthesia.”

Q that's great. You anticipated my, my next question, um, here. Like, how do you, I guess, balance, um, uh, you know, moonshot kind of research with deliverables, uh, for an AI team? Do you, and, and, and do people, I don't know, get together and agree on roadmap on what might be feasible? And how do you know when to push or cut your losses on an AI? Research project.

A It's really, it's really, really, really difficult. Uh, just want to preface that saying that I don't think we have the right answer, but I think the kind of mental model that we're sort of moving towards now is basically looking at teams in terms of like, um, how far removed are they from product, right? On the one hand, you have blue sky teams that are really doing things that might not hit the product before two years, because this is betting on like the next really, really big step change. In the middle, you might have something which is kind of trying to advance the capabilities of what you have right now. And then you might have a team that's almost like more kind of AI engineering, which is like taking more kind of stuff that already exists and quick, quick wins where like in three months you might be able to kind of ship a feature, right? And then looking at those three things, figuring out like how much resource do you place where? How do you manage the kind of interaction with the product engineering teams? What do you do when a piece of research is done? How does that get put into the product, right? You want to try and paralyze those two things to ship faster. But it's also really difficult because with, when you're doing AI research, nobody really knows if it's going to take three months, or six months, or nine months, or 18 months, or if it's even possible. So it'…

AI assessment note: “mental model that we're sort of moving towards now is basically looking at teams”