This article was originally published on ParlayMe.
Based in London, Synthesia has been named one of the 100 fastest-growing startups in the UK and Ireland!
Generative AI is a transformative technology that is helping people create synthetic media on par with images, audio or video produced with traditional means.
January Barnes, founder of ParlayMe, sat down with Alexandru Voica, Head of Corporate Affairs & Policy from Synthesia at the recent TEDAI Vienna conference in October 2024.
Voica reports directly to the CEO and co-founder to craft and implement communications strategies that effectively articulate Synthesia’s leadership in generative AI. Voica works closely with internal thought leaders and external partners to bring the company's story to life and execute Synthesia’s communication strategy across earned channels, including media relations, public policy and key events.
[January] I want to talk to you a little, well, I did a bit of Googling. I didn't use any chat GPT in this, so I'll just let you know. You currently live in the UK, you were born in Romania and you admit to speaking fluid AI, which not many of us do, so you're ahead of the game already.
You own your own content agency. I'm telling you a lot of things you already know, but you also work in comms now, obviously, and you have for a long time.
And maybe most importantly you like Guns N' Roses, but I'm not going to hold that against you 
I'd love to start this interview off with more info on where you started.
You studied electrical engineering, electronics and computer science.
I know this interview is about Synthesia, but I'd love to know a little bit how you got involved in this AI world.
So take us back to kind of when you were doing your electrical engineering degree and then what made the transition to marketing and then to where you are today.
[Alexandru] Yeah, when I was growing up in Romania, I was very good at math and physics and I would go to competitions with other students and sort of be moderately successful. But I also was fascinated by language and so I studied English and I got very good at English.
I had this kind of like dual brain of doing a bit of both. But when it came down to sort of choosing what I wanted to study. I went and studied at university. I went to study electronics, which is very kind of math-heavy, and computer science. But I still had a sort of a passion for literature, and I always knew that I wanted to sort of combine the two. So after a few years as an engineer, I saw this opportunity to kind of go for a job in communications and when it was for a technology company, in semiconductors.
I was like I'm gonna try this out for a year. Yeah, if it doesn't work out I can always go back to being an engineer, right? They're always in demand, right? There's a high demand for engineers, but it kind of worked out. So that's what I've been doing for the last 13-14 years.
[January] This is amazing and I think this gives you an advantage in somewhat because you know, you're a technical marketer.
You've got the engineering side of the brain and you've also got the marketing side of the brain, so I guess that brings us to kind of your current role, do you see this as like putting yourself obviously at an advantage?
If so, how much of an advantage? Especially when you're dealing with AI, like, do we need people in your position to have this kind of background of understanding? Is it good enough to just, you know, I don't know, being the marketing director of say, a studio, let's say, well, Netflix is a technical studio, but let's talk, you know, old school, like 20th century Fox and then step into an AI situated kind of platform.
[Alexandru] I know that from the outside, we see them as magical, and they are to some extent magical in the sense that their abilities, we perceive them as magical, but underneath the magic they are just complex mathematical systems that are working to produce what we want them to do.
I'm not saying that everyone should have the same background as me to kind of work in this type of role, but what I'm saying is that by having this background and understanding how these systems work, not only do I try, and hopefully I'm successful, to explain how they work, but I can also understand what they don't do. Their limitations, and I can, I hope and I argue that I can be more realistic about the future.
Whereas maybe for someone who doesn't understand how they work and just kind of looks at what they can do today and has maybe a more sort of creative mind, they can either exaggerate in terms of the good or exaggerate in terms of the bad. I feel like it kind of keeps me grounded to still have the sort of vision of where we're going, but understanding of what's holding us back.
[January] The limitations.
[Alexandru] Yeah, absolutely.
[January] Broadly speaking, what kind of key developments in AI have you found most fascinating this kind of transformation era we're going through? And you can say AI video, because that's what you guys do. But broadly speaking, I'd love to know what's kind of fascinating you at the moment?
[Alexandru] I remember more than 10 years ago, I was working for this company that made the graphics processing units inside mobile devices, especially Apple devices. They were the NVIDIA of mobile devices. NVIDIA puts their GPUs inside your laptop, inside your computer. This company put their GPUs inside your iPhone, inside your iPad. And I remember at the time we were contemplating this idea of could this graphics processor that's very good at gaming, at generating frames of games, could it also be used for compute? And I remember the first application of this was for self-driving cars. And we all thought that self-driving cars would become a huge market for AI on the GPU.
So I was very excited about that and then I realized the limitations of that approach and made me sort of understand that even if you can get from zero to 90% very quickly, you can get from 90% to 100%. It's very, very hard to do that, and we still haven't done that.
I think that was sort of the first time where I got excited about AI in this kind of self-driving space. I thought about it, it was less about robo-taxis, and it was more of safety. So many people die needlessly of car accidents, especially in North America where cars are huge. By adding these systems, you can make cars safer.
Then the second time around was, you know, with ChatGPT. I think there, what fascinated me, and I think you don't talk a lot about, is how the technology to build ChatGPT was always there. It was there for a number of years. It was just that no one thought about really putting it into practice. So the theory behind ChatGPT been at Google for a number of years, it was just that they said, oh, it's an interesting theory, and we'll just kind of leave it there. OpenAI was the first one. Like most ideas, right? You've got to do it. It's an idea until it's done. for them was that they took that idea and really invested a lot of engineering effort into putting it into practice. And that's how we got eventually, you know, a few years later, we got Chat GPT. I think there, what that demonstrated and what kind of got me excited was that research is not enough. You need to sort of take research and put it into a product.
It takes a lot of creativity to do that. Creativity is not just coming up with the idea, it's actually taking your idea and putting it into product. And obviously now it's this idea of going beyond text and what we can do with video and audio.
Sort of looking into the future, I think what I'm most excited about is like the possibility of applying, because in addition to my role at Synthesia, I also work for a university that does AI.
I think what excites me there is sort of the real world applications of AI beyond, I mean, here we're just talking about content creation, but there are so many other things in terms of healthcare and climate change.
This is where I'm very careful about hype. So a lot of people will say, we'll live 200 years. And I don't know if that's true. It might be true. What I'm more excited about is, for example, one of the things my university is doing is predicting the likelihood of you reacting badly to medication based in your just by looking at your health records, right? So just just looking at your sort of medical history being able to say if we give you this medication. Yeah, I sort of react badly to it right which is something that happens in hospitals You know when when people get sick and doctors sort of rush to administer a drug you might you know be causing some accidents. These are the kind of things that sort of get me excited.
The possibility of living 200 years, I don't know, maybe.
[January] Do we want to? That's the question. But yes, then it brings other problems, right? And opportunities.
So you worked at Meta for years. You also worked at NetEase. I mean, these are big, big tech companies. So I guess my next question is, how do you go from working at such big tech to a startup? You're still scaling, you're bringing out product. Like what are the advantages that you bring from coming from big tech? And what are you finding a bit like constraining from being, you know, in a smaller realm? So to speak.
[Alexandru] Yeah, I think one of the ways in which I've tried to contribute has been, we you know when we released social media products into the world the original hope and to some extent it was kind of proven in practice was that these technologies and products will be a net benefit to humanity. And by connecting more people we would get more societal benefits and I think initially that was true because I think that the vast majority of people are inherently good. But, perhaps, not perhaps, but surely a lot of times those companies didn't think about the negative impact as well. And so they rushed to achieve the positive and to some extent neglected or didn't necessarily foresee the negative. I think it's kind of like, I've tried to approach it from that point of view to ensure that when it comes to AI, we kind of are better at detecting those risks. Checks and balances in place. We don't rush to release technology just because we can or just for the sake of it.
The other thing I think is, one advantage of smaller companies is that they can move faster, and it was one of those things that I missed. Working for very large companies, Meta has over 70,000 employees. It's difficult to put your ideas in practice, not necessarily because of regulation otherwise, it's more just a bureaucracy. So I kind of missed the ability to be like, okay. I have an idea and I think we should do it. Yeah, just let's test it and get it done.
[January] Well, that's exciting, right? That's kind of where innovation happens when you know you have to change things rather than like it ain't broke don't fix it kind of which a lot of these tech companies get it's doing well and don't change. Interesting. So, you know, Synthesia is doing incredibly well. I don't need to tell you. I think 60% of the world's Fortune 100 companies are using it. You guys just raised $90 million. You've had a strategic investment from NVIDIA. So things are looking good. The goal, I guess, that I heard from Victor, the co-founder, is that he started off with wanting to democratize video creation. So my next question is, how much is that is still the end goal? Or is it I mean, we live in this world where AI is changing by the millisecond, right? Like, does it have to be reflexive at some point and responsive to the market? Is that the end goal? Or is that shifting slightly? Can you elaborate on that a little?
[Alexandru] Yeah, so you're right. The original vision was to make everyone sort of able to create video. But since then, we've sort of expanded our view. So right now, I think our stated mission is to bring the magic of video to all communication. What that means in practice is it's not just about creation, it's about distribution as well. So one thing that you might have heard him say in his TED Talk is that so far we've been distributing AI video with traditional. So we went from, we were creating video in a traditional way and distributing it in a traditional way.
You and I would film ourselves with a camera. We would upload it on YouTube, and, hopefully, a lot of people would watch the video. Right. We've we've now replaced the traditional, way of making video with AI. So we're making video with AI, but we're still uploading into YouTube or TikTok for lots of people to watch it. So we think that we are or we're working to also replace the distribution, part. So instead of distributing in any sort of classic traditional way, we wanna create a better experience.
We could talk about what that means, but yeah, that's sort of where we are now. Then looking sort of towards into the future, if if all of this technology, evolves on on its current trajectory, we think that there will there's gonna be a world where, millions and billions of people interact with billions of AI agents. At the moment, those interactions are happening through text. You know, type in a prompt, and we wanna be sort of at the interface between people and AI.
We wanna be the face of AI, essentially. Right. So Big goals. Yeah
So we're gonna be at that kind of, boundary between you and and an AI agent, interacting with, with a familiar face.
[January] Interesting. So I guess that brings me I wanna kind of, build upon this concept of democratization of content. I wanna also talk about the kind of freedom of speech, if you will, like Elon Musk. I mean, what he calls freedom of speech. That's debatable. But how much so, like, is what you're doing, like, you're not not having any political content from what I understand, although that may have changed slightly. You're not doing news from my understanding. So you're not engaging agency in fact. So why is that? Why is that stance?
Or is that shifting slightly now? Is this what I'm hearing from you? Or is it steadfast, like no political contact, no news, or can that change?
[Alexandru] So we do allow political content and we do allow news. We just allow it with some controls in place. So for example, if you wanna make political content, you have to make it with your own avatar. You can't use someone else's likeness. It's only fair that if you have some political views you wanna express, you should use your own one right.
Your own avatar to see. We allow the content from, with an enterprise account, because with an enterprise account such as an organization that sign ups with our platform, you go through a process of onboarding. We know who you are. We know what your, what your intentions are. So there's checks and measures in place.
Just like when you get into a car right. You need a driver's license. You need insurance. You need seat belts. It's the same thing. Same for news content. So you can make news content, but you have to use your own avatar and you have to have an enterprise account.
So it's not like we have these kind of, like, complete bans. It's just that for specific areas where we feel like there's a risk. We wanna make sure that you are guided a bit.
We wanna make sure that there are some some limits in place. The way the best way to think about is that there are essentially three categories of content. There's the content that we all want and we all seek.
We can all probably agree as a society what what that content should be if you wanna learn about how a car works or if you wanna learn about AI or, that's, yeah, that's perfectly normal. Then there's a content that I think we all, again, at the societal level, we agree it's not good. You know, hate speech or sexually explicit content, non consensual, that that type of stuff.
And then there's the the thing in the middle where it's a very it's a very difficult problem and we're still having debates at a societal level. Some people say we should allow more of it. Some people say we should allow less of it. And those are things like, you know, what is misinformation? What is the difference between someone educating you about Bitcoin versus someone, fooling you into buying like a fake coin.
Right? What's the difference between someone selling you a new drug that can cure your disease versus someone selling you some fake pills that give you, you know, bad side effects? So this is where, the way we approach this issue is. We're not here to fact check. We're not scientists. We're not researchers. We take a stronger stance on if you wanna make this type of content, some of it will be restricted. So just make it with a normal camera. Right.
Some of it, you have to get an enterprise account. You have to get your own, avatar. So there's like I said, rather than trying to kind of decide we deciding, what is misinformation, we're saying, okay. Well, maybe if you wanna make health care content, if it's educational use your own avatar and make your, you you know, onboard with an enterprise account. If you're trying to make sort of, like, pyramid schemes at multilevel, like, gambling content, go somewhere else.
[January] And are they, like, humans looking into that or, like, bots or what is the is it a combination of Combination?
[Alexandru] We use, AI, based filters as well as human we employ human moderators.
[January] I wanna just talk quickly about the pros and cons of, AI generated video because there's a lot of them. But, primarily, there's a lot of people that, you know, there's big advocates for what it can do, and there's, on the other side of the coin, ones that say, well, there's homogenization of content, and we've all created the same content. It kind of, you know, it's an age old problem, like we've had with every technology that comes.
It's like, I remember when the earphones came in. It was like, oh, if everyone's walking around with their earphones in, listening to music, they're not creating art, you know, all these crazy things because they, you know, you're not interacting with the art around, you know, the world around you, which essentially is how you make art. Right? So there's lots of pros and cons. What what I wanna kind of ask you is, like, what your platform does, it gives other opportunities that perhaps you wouldn't do as, like, you know, in traditional forms of production. Is there a happy medium? Is there one where just your your guys' platform is gonna be the place, the only way you can do it? I mean, I know Vic just spoke about once, like, he wanted to create, you know, Hollywood star movies, a click of a button.
Is that is that really the end goal now? And if it is, is it gonna happen? You know, what's your what's your thoughts around that?
[Alexandru] The way we see it is that, people will still be able to express themselves with traditional tools. You can still use your camera. You can still use, you know, Adobe Premiere Pro. To edit films. You know, big studios will continue to do movies with visual effects. Marvel productions with, like, green screens and so on.
What we're trying to do is we're trying to bring some of those capabilities to people who don't have access to a movie studio. To, you know, Premiere Pro and these kind of, like, professional tools that are, expensive. That's what we're after. We're not trying to replace Hollywood with Synthesia.
Hollywood can continue to exist and I'm sure they will and I'm sure they do very well. We're trying to help, you know, your, someone like your average small business or your average sort of office worker that doesn't have $50,000 to send on a video. Right. So it's another tool, in fact. It's another tool to the bow, like, you can use.
You know, it's not one or the other. And I think there, the the idea is to, not just improve access but to sort of open up, the creativity that people can have. Because let's put it this way. If we go, bankrupt tomorrow, if we disappear tomorrow. All of these we have over 1,000,000 users. All of these 1,000,000 users will just go back to text. Because none of them or the vast majority of them were not video creators. They were people who are making, PDF or PowerPoint Yeah. And now they're making videos. So for us to disappear it would mean that people would just go back to text.
Our focus is it's not about converting existing necessarily existing video creators into Synthesia users. It's about people who are, kind of using text still and making them video user. You could argue you're almost humanizing it because there's a debate that, you know, it becomes dehumanizing if you watch just avatars in videos and where's the human element.
[January] You know, to what you're saying, if you're, you know, at marketing agency or whoever you are, you're doing training videos and you just used to put a sheet in front of people and say, here's your checklist of how you get through the day at this place. But instead, now you're creating some kind of humanized element. You're almost create putting a human touch on things, which I think Victor spoke to yesterday. He always wants that. Or you guys wanna have some kind of human touch element to what you guys are doing, which I think is interesting and, obviously, important.
I'll just talk ethics really quickly, and then I'll let you go. It's a it's a big one, obviously, but, you know, they they spoke yesterday, you know, the whole Word document kind of scenario, which I never really thought about myself. Like no one's saying in Microsoft, well, you know, you have Word - so you're responsible for what people type up or, like, you know, how they use it. How do you guys see that? It goes back to what we talked about earlier. But, you know, what are your checks and balances? So say, you may or may not be able to tell us, but someone say is you know, you do become, the next Facebook or whatever it is, and you're creating all this content and something happens where, I guess, the transparency of what you guys do and who's accountable?
So what's your thoughts on that? And it's changing daily and, obviously, with policy and everything like this?
[Alexandru] The the way we view it is that because AI is a powerful technology that is, to some extent, not just a tool. The point of view of with a Word document, you have to sort of sit there and write the whole thing.
If you're doing a video, you have to sit in front of the computer and put together the entire video, in kind of traditional sense. So that you're limited by your, sort of, creative capabilities and you're limited by the time you have. They're only twenty four hours in the day.
The difference with AI is you're not limited by your creative capital release. You can make, content that goes beyond your your limitations and also to scale. You can make a lot of this content very, very quickly.
So from that point of view, with those sort of just those two elements. There are some other considerations, but just those two elements in mind. We believe from the very beginning of the company that we have a responsibility to put some controls in place, to limit bad outcomes, and but still allow people to obviously pursue the the the more positive applications.
I think from that point of view, that's how we kind of did that. This was seven years ago before everyone anyone was talking about regulation in in the space. And that has helped us now because it's very difficult if you if you've built your company with no sort of responsibility and no sort of, ethical guidelines. It's very difficult to don't then go back and try to refit yourself around regulation.
If you have some really, sort of strong foundations in terms of, this the, you know, responsibility, it's much easier than and so what we've seen is, there's a new standard that's very technical. There's a new standard called ISO 42,001. Right. And this standard, takes some of the regulation principles, like, from the EU AI Act. It's built the standard for responsible AI. If you're a company who wants to develop AI responsibly Right. You can sort of get it certified for this standard. Okay. And we were the first company to be certified because, you know, we were able to kinda sail through because we had all of these, sort of, ethical, principles in place.
I would imagine the reason why some of our competitors will struggle is because they don't have these things. Like, they don't have content moderation. You can make deepfakes, nonconsensual deepfakes. So, yeah, it sounds strange, but it becomes almost like a business advantage.
When you have these foundations, you can move faster. You can, with businesses, you gain trust.
[January] So it's not just, you know, morally and ethically a good thing, but it can help you with business deals. Of course. I mean, you guys are in a really great position.
Like, you have a product. So many AI companies are still in the research and the you could revenue, not just a product.
It's very interesting your model, and you see some businesses do it and some don't. I I don't wanna compare you to other ones, but, you know, Canva CEO recently said that they really listen to their customers. You know they reiterate their product, constantly for feedback.
Do you think a Steve Jobs model would apply, or do you think there's a there's a balance? Or how do you guys approach it? Because I feel like from what I've heard from Victor is that you guys really do take the feedback of what people your customers essentially giving you and you take them on board? Or or is that not the case? You tell me.
[Alexandru] I think when when you sort of the danger is to sit at the extremes. So the danger is to build your product, solely based on what the customer wants, and then you end up with a mess because no two customers will have the same feedback. Right. And then the other experience, you completely ignore what the market demands and just, try to build. And there have been examples where, you know, companies have been able to create entire new markets with products, that maybe, that there was no demand for them, solely based on building something cool.
I don't know, maybe you wanna be somewhere in the middle where you can still innovate, try new things, create a market. But once you've created a market, once you've created sort of a customer base, you do have to bring sort of the customer with you. Right. And we have this, thing internally about the voice of the customer and we always, study the, that data very carefully.
Yes, it's easy when you're kind of creating a whole new category of category of products, to just kinda go alone and kind of, move at speed. But once you do, you have to kinda slow down and start start sort of listening to the people who use it.
