Episode Summary
In this episode of My Space, host Manav sits down with Laodius, the founder of Argil AI. The conversation dives into the cutting edge of AI-generated avatars, exploring how Argil creates hyper-realistic digital humans that could revolutionize video content creation. Laodius shares his fascinating journey from filmmaker and novelist to tech entrepreneur, touching on his experiences in fundraising, his path into Y Combinator, and how storytelling has been the common thread through all his pursuits. The discussion covers the deeper implications of AI in content, including job creation, democratization of video, and the ethical questions around deep fakes and content authenticity. Listeners will also get practical insights into the technical and creative processes behind building Argil, how to use the platform, and what the future holds for both AI-generated content and human creativity.
Transcript
Laodius [00:00:00]:
Some people are losing their minds because the video below would be AI generated. I'm also a deep fake. And you should take care when you watch a video on the Internet.
Manav [00:00:08]:
In this video of emerging founders with Manav, we will be interviewing Laodius. He's the founder of Argil AI and they're building AI avatars that look insanely human. Like, he's one of the most interesting people I've actually talked to because he was a filmmaker. He's really good at fundraising, and he's also got into Y Combinator. So he tells me all these stories in the podcast. And check this video out. This is an actual video made by Argil. It's going to blow your mind.
Laodius [00:00:32]:
I'm a deep fake version of Zuck made by Argil AI and I'm truly convinced that we need to raise the awareness of people in the Internet.
Manav [00:00:45]:
I'm really excited to have you on the show, Laudis, because I really want to use your product for myself.
Laodius [00:00:49]:
Thank you for having me. Mana.
Manav [00:00:51]:
Amazing. So you're one of those multifaceted people who has kind of like, proven himself in multiple domains. You were in the movie industry. You've helped founders raise a lot of money. You've actually hired a lot of engineers. You've joined Y Combinator. So I feel like it's very interesting to talk to people like you, because you're not only building, but you're, like, doing many different interesting things. So how did you get here? Like, how did you get into so many interests?
Laodius [00:01:15]:
I think there's a common thread in what I've always been doing is storytelling. I think both building a startup, getting clients is telling a story about a product and solving a problem that they need. I think both the creative work I've been doing, like writing a novel, doing movies, it's also about telling a story, a compelling story. How do you do that? Even helping founders fundraise. So that was the first masterclass I did in the 2020. I think it's also when you help a founder raise money is you help him basically craft a story for the investors for them to follow them and believe in them. So I think the common thread is I've always revolved around storytelling, but there's many ways to tell a story. Either it can be by building something or just like actually writing something. And so I've always managed to take my time in order to really. For example, my novel, I wrote a science fiction novel. It took me nine months, but I only did that. I'm not the kind of Person that would be able to, like, have hard days of work and come back home and then write in that mindset. So I always managed to really put my time, have hard focus periods of at least a couple of months, two years on different projects, which is the case today with what we're doing at Argel. But also I think there's a period between your 18 to maybe 30 something years old where the cost of opportunity is super low for trying stuff. And when I was very young, one of my goals was like, by 30, I'll retire and be a millionaire. And then it really evolved in my 20s because I was like, at this time, I can work hard, but I can also do a bunch of stuff that I may not have the energy at the time to do later today. And this involves creative projects like writing a book, doing a movie is stuff that you don't necessarily have enough time or energy to do later in your life. So I think making the most of that period in your life is super important.
Manav [00:02:48]:
That's a really good answer. Actually. A lot of people have this goal of being a millionaire by 30. I think I made my first million when I was like 27, and it felt amazing. It just gives you more room to follow your creative adventure endeavors. I want to talk a little bit about how you perfected the pitch decks because you said you helped a lot of people. You really understood what investors are looking for. Can you summarize what investors are looking for? Because I've seen investors, like, skim through the pitch deck in like less than 20 seconds. So how can you grab their attention in under a minute?
Laodius [00:03:20]:
Yeah, I mean, the most important thing to understand when you're raising funds globally is you have to understand the job of the investor. And most of the masterclass I spent just explaining what's in their minds. Investors, like the way the math works. If you're a VC or an angel today, you need to invest in a bunch of companies, and every single one of these companies need to be able to return the entire fund, which means you need to invest in basically, definitely moonshots. It's the only way for that business model to work at the end of the day. So first of all, understanding that, what specifically are they looking for? A moonshot in terms of founders character traits or business or whatever it is is super important. So understanding the investor is one thing, and second of all is crafting a story that will really pinpoint those things they're looking for. What I always tell founders is whatever you're doing, these investors have seen 10 companies that do basically the same thing than you in the past month. So you can be good at everything, but you're not going to stand out. It's much better to exceptionally like to stand out exceptionally well in one thing, even if you suck at the rest. Because that's a better case for a moonshot investment. Meaning you need to know what part of your team or your business or whatever you're doing is an outlier compared to other companies. It can be because your team is a veteran in that industry. It can be because you're having an insane traction. In our case, for example, we had two things. First, the tech. We managed to build the tech in like six months. That rivaled companies that built like two or three years of tech with tens of millions of dollars. So the efficiency was one thing. The second thing was really the founder market sh. Because as we're doing a product in the creator economy, since I've been a creator myself for years, having a TikTok account and newsletter and a bunch of content, it made a lot of sense for the investors that, okay, this guy gets it. Like he was doing Reddit front pages in 2010. So he knows this ecosystem, which is not the case of everyone. So I think the most important is if you skim your pitch deck, whatever is the outlier of your company and it's always unique to your company. It can be just applied from a template or whatever and needs to stand out extremely well. It can be a single number, it can be a market size, it can be literally the first slide. Like this market is worth 10 billion and I'll concur. But you need to know that and articulate your deck around that.
Manav [00:05:22]:
I used to work for a startup accelerator and I reviewed like more than 200 pitch decks there in Palo Alto. And when you're investing that early stage, I noticed like the team is like one of the most important slides because everyone's just trying to make sense of the story. They're like, is this person, like, trustworthy? Do they have the track record to like stand out and persevere? Like, you know, that's what we were looking for. Like, when you're investing that early stage and then obviously it changes. Like seed round and series eight, that changes completely because you've already have revenue, but pre revenue, you're like, you're just betting on the person itself. How do you end up looking for co founders? Has it just been through, like the people you met irl or have you done a cold outreach and Metro co founder that way?
Laodius [00:06:02]:
No, Luckily I met my co founder in the first startup I worked for 10 years ago, which was a French scale up. They were hiring like really high grade engineers, like the best in the French market. And that's where I met my current co founder who was coming back from the Valley by the way, which, which is quite uncommon in France. And yeah, we had a good match quite early. I mean even in 2014. We were talking about bitcoin at that time. So that's like put me on the mindset of he's also intending to do to build something pretty big someday we might stick together. So we built the first company in 2016 that didn't work out and then we both went our ways and like basically our tracks met again a bit less than two years ago.
Manav [00:06:34]:
Amazing. Okay, so I want to know the founding story of our Gil and what problem are you solving?
Laodius [00:06:39]:
So since I've been doing a lot of video content for the past years, either super creative, like when I was working on my movie, but also as I said, video masterclasses and TikTok accounts, the big issue with video is always the production cost. When you think about it, it's a very elite media, meaning it's super expensive to get into video regardless of what you're doing. Not even going to talk about like Hollywood or TV channels, but even any YouTuber with a sizable audience will spend a lot of money doing YouTube videos. So it's not that easy to get into video cameras, lighting, you need to know either to know to have enough money to pay for post production and editing. And video ended up being the only market where you cannot run a lot of experiments because whatever you do is so expensive that it has to work. And if it doesn't work, you'll have just abandon it or take the loss. So we were working on a bunch of AI projects about a year and a half ago. And then at some point it became kind of obvious that the biggest problem I wanted to solve for myself was actually how do I make more TikTok videos and more educational videos without spending too much money? And so we decided to dig on that model. We looked at the market, we saw that in the market some players were doing avatars, but mostly for different use cases, for example enterprise learning. And we're super focused on how like my goal was always avatars are not only going to stick for internal use cases for enterprise clients and corporates, they're going to be used as well ultimately for social media. And at some point in the future it will make no sense to pay for a $5,000 camera if you can actually generate the exact same content by typing it on a platform. And so when you go far enough in the future and you know this is going to happen, inevitably you can backtrack to what we're doing to it today is like, how do we build the first foundations of this tech and the platform to get there?
Manav [00:08:20]:
First of all, I'm really bullish on this idea. Again, like, there's a lot of people who also find downsides in this. Oh, we're replacing humans. And some companies will think of it as like, oh, we're not even going to hire creators, we'll just use avatars to pump these short form content. But I obviously think we're still going to need humans.
Laodius [00:08:37]:
By the way, these companies do not spend money to do video. What I mean by this is from my lessons is that nobody is replacing anyone because those that never did video start doing video with avatars, but would not have the budget to pay for normal videos. And the ones that already do video usually just enhance their workflow. So so far I think it's job creating than job destructing.
Manav [00:08:59]:
So a little bit context for the people watching. There are current tools like Heygen and Synthesia, but they're targeting more enterprise companies and their avatars, they look very robotic. But what Laurius has built with Argil, it looks very real and I was actually shocked when I first saw it because it looks insanely real and you can be like walking on the street. It just looks very organic. I kind of want to know, how did you achieve that? Because that's like really impressive.
Laodius [00:09:24]:
Yeah, I think one of the things, you know, when you start a company is what kind of secret do you have about an industry? I think within the companies you cited, a lot of these founders are really, really good technical profiles. I do think, however, that none of them were content creators before starting that endeavor. So I guess it's more of. It's not so much of a technological breakthrough as in an approach that we decided to have that avatars need to be good enough to crack views and likes and subscribers on social media. So everything like every decision we'll make will be done in that sense. So that's why in the choice of avatars, in the features that we have, for example, you can manage body language, body movements, and you can say hello and all of these. And even on the way our editor work, which goal is really to generate a fully edited video in like two minutes with captions and backgrounds and whatever this whole flow was thought on. I want a performing video at the end. And so it's a bunch of different components essentially.
Manav [00:10:17]:
Interesting. Okay, so I actually have used the product. I know how the setup works. Can you explain how a person can sign up on Argil and basically how long does it take from the start process to like get their avatar up and going?
Laodius [00:10:31]:
Yeah, so it really depends on the grade of the avatar that you want. Like what I mean by this is content creators that are already doing content at a good grade have the material, have all the setup, will spend maybe one or two hours doing like really an absolutely perfect flawless avatar because they want to replace their real self by their avatar and so it has to be really, really good. So they will basically try to have the same setup than the actual videos they do. They will do every body language, gesture and the training video so it's reflected on the avatar. But I would say for average person that just wants their avatar for any use case you really need three minutes of you speaking to the camera with a decent lighting and a decent, decent sound. You can do two or three versions of your avatar by the way, if you want like different outfits or different backgrounds. But in general the process from start to end, if you start from scratch, it can be like 20 to 30 minutes to have the first avatar training on the platform and then it takes us a couple of hours to train the avatar. Once you have it, a video will take like a couple of minutes to be generated.
Manav [00:11:26]:
Will you guys be integrating the voice model inside the software or would you have to be always has to to do it separately on 11 lab.
Laodius [00:11:33]:
So right now you can generate the whole package on the platform. If you have an elevenlab account with specific voices that you want to use, you can plug it to Argel. In the future we'll probably do our own model but right now it's very convenient to use 11 apps because they have all the languages already supported.
Manav [00:11:48]:
Got it. I am really excited to see how you guys progress because this is a big challenge because filming every day, it's such an inconvenience. You have to set up so many things. It's so time consuming and a lot of people just that technical barrier can make them overwhelmed or they'll procrastinate or they'll not post. That inconsistency will be the reason they quit because they burn out and they don't see the results. But we all know like podcasting, YouTube, TikTok, Instagram is such a long term game.
Laodius [00:12:16]:
Exactly.
Manav [00:12:16]:
Mrbeast. He didn't see results for like seven years. Imagine if Mr. Beast quit, he would not be like the world's biggest creator right now. So it's great that you guys are building this. I want you to talk about, like, what's the next thing for Arjun? Like, what are you guys focused on? And have you guys already like gone to the market? Is it like still in the beta stage? Like, where are you guys right now?
Laodius [00:12:35]:
Platform is live. We have already a lot of users, a lot of paid clients, ranging anywhere from a couple dozens of bucks a month to thousands of dollars a month, depending on the size. Because we have an API that people can use for high volume use cases. There are a lot of use cases that demand the API. A good example is a job board that wants to transform every one of their job offerings into a video at scale. The vision I have is, in the future, creating a video needs to be as easy as tweeting, essentially, because if we want to attract a new generation of content creators that were so far not willing to do video because of the effort, because they don't like what they look like in front of a camera or for several reasons, but they would use any other platform like LinkedIn or Twitter just to create content because there's not this barrier, we want to remove that barrier to make basically video as easy to do as any of these. So the goal is keep doing a model that is more and more realistic and keep doing the platform to create and pre edit videos that are going to perform on social media and have this social loop, this loop of I create a video on the platform, I post it, I'm getting a bunch of likes. Ideally, in the future we'll track the data of all of that to help you like draft videos. Long term, we want the platform to be an assistant as well. What content do you want to exist? I mean, the way I see it is in like two years, the average local plumber carpenter is going to take a picture of whatever, you know, work he just did and it's going to be posted on all social media because it's going to be a really cool video generated and it's going to bring in like local clients. So I see that as everyone will have an incentive to create video and we want to help people do that.
Manav [00:14:00]:
Okay, so this is kind of like a question I'm guessing you think about a lot, but how are people going to differentiate? Is there going to be a watermark situation on videos in the future? I mean, because it's not just your platform, there's so many now AI generated video platforms, like, how are we going to differentiate in the future? Or is there going to be a need for that? Or no?
Laodius [00:14:18]:
Yeah, I think your last question is the right question. Is there going to be a need for that? My intuition. So if I were going to predict what's going to happen, I would say that for the next two to three years, we'll need to water more videos because we're still at a stage where people feel cheated in a way, when it's an avatar. Most of the work I'm doing is trying to explain people that I'm not cheating anyone. I'm still doing the content. I'm just saving energy on actually filming myself. Energy and money. But at the end of the day, I'm trying to bring as much value as I would if I had filmed myself. So we need to go through that stage. But I think especially because the new generations are much more comfortable with AI avatars and AI in general, I would imagine that in three to four years the need for watermarks will just disappear because I think the audience will not care whether the content is AI or not. And I'm pretty sure this is going to apply to everything.
Manav [00:15:01]:
I 100% agree with that. I think AI is just raising the bar of the quality we look for. And it's so funny right now, if there's a bunch of ads that were created just by AI, they're actually outperforming the human ones because they stand out more. Like, because not everyone is using AI. So those AI generated ads are like outperforming the normal ad, which is like, in my opinion, crazy. I was actually going to ask you to make a prediction, but you already kind of did, so I'm really happy about that. Well, long term, what's the vision? Are you going to be back in the movie industry? Like, do you still write? I want to kind of like know a little bit about your writing, writing process, because I'm actually trying to become a writer myself. And I want to know, like, how do you get into that state of writing? And are you using like AI, like Claude or ChatGPT to write some of the stuff you write?
Laodius [00:15:44]:
So it's interesting. So I wrote a novel five years ago, it's in French, but there was.
Manav [00:15:48]:
No AI back then.
Laodius [00:15:50]:
That was exactly. There was no AI back then. And so the first thing I was like, what I've been trying to figure out in the past year or so is if I had to write that book again, could I leverage? Because so one thing is for sure, right now, you cannot just unleash Claude and tell him, like, write a book.
Manav [00:16:05]:
It's crazy.
Laodius [00:16:05]:
Going to be nonsense. The style is not going to be consistent. I have seen though, I need to check that. By the way, there's someone on Twitter that I saw posting. He basically built, I think it's like 18 agents to write a book, but each one has a specific role. Like, one is for character consistency, the other one is for location consistency. The other one is responsible for the global style. So this, I think could actually work very well. I could imagine that a bunch of agents criticizing each other to write a book would work well. And the question is where they. That the human fit in the orchestration of that. I think if I were to write a book right now, I would mostly use it. Like, if I were to do that process again, I would probably come much earlier to conclusions that took me weeks or months to take. Like the role of a character in the, in the story, the potential endings, the like, the coherence of the whole story. So I think I would use it to be, to challenge like, okay, I'm thinking of doing that. Would that make any sense? Does that character make sense in that context? I want to use it as a sidekick. I think. I do think that within, again, three to four years, the value of a human trying to write a novel is going to quickly go to very close to zero. Or to say it in another way, I think most average writers will be out of the market completely. Because honestly, if you look up at the kind of content that comes on, even Netflix or some of these TV shows, I think ChatGPT is already good enough to challenge most of that content. So, yeah, I think the outlier writers will stand out and will like use AI for a bunch of stuff. I'm not making predictions for more than five years because honestly, we have no idea if we're going to hit a higher threshold in quality.
Manav [00:17:34]:
Yeah, I think my prediction is two things. That AI will probably never be as good as us is comedy. Comedy and fantasy. Like Game of Thrones or Lord of the Rings or Harry Potter. I don't think AI will ever be able to do that. But one of the biggest question I want to understand is, is AI good at reasoning or not? Because large language models, they just predict the next word. There is no cause and effect. So I get it. AI is impressive at acing the test exams, but these models are trained on words, not experiences. Is like they know language, not life. So I want to understand, like, if AI would be able to reason, like us or not? What do you think?
Laodius [00:18:09]:
I think it's a super philosophical question because the underlying question of that is what is reasoning and are we really reasoning or are we also just predicting the next token? And I think in a way it's not impossible that we humans are maybe kind of predicting the next token based on context and a lot of knowledge and a bunch of connections. There's something that could be somewhat deterministic in there. My feeling is that honestly the answer to that question doesn't matter because what I'm seeing recently from even 01 is sufficiently like O1 with an infinite memory. Because I think some key AI figure recently announced that Google had cracked infinite memory or close to infinite memory. So a really good model like O1, even if it's just token prediction plus a really good memory. I've been a believer from day one.
Manav [00:18:52]:
Can you explain what infinite memory means?
Laodius [00:18:54]:
Yeah, of course. I mean today when you ask a question to ChatGPT, like if you have a conversation with ChatGPT or any model, they have a limited amount of input that they can use as a memory. So it used to be very short, like, like it would forget anything further than four or five questions. Now it's able to take like two or three whole length book books worth of memory. Now if you have very, very long memory, it means that. Let's go back to the example of writing a book. I could like use that as a sidekick all day long and it could literally use all the knowledge of all the discussions we've had in the past, plus its knowledge to keep being accurate about something about a character. A good example is a character I would give ChatGPT in the older versions my book and it would forget who half of my characters were and I'd have to re explain them again and again. You don't have to do that. And if you have a bunch of very, very specialized. I think the approach of that guy on writing a novel is really good. I've always believed a lot in swarm of agents doing very specific roles because it's much easier for an agent to be like, my only role is to check every chapter and make sure it fits a specific style. And my context is super short. This is the style I'm aiming for. I'm going to read that and see if it works. And the other one will be like, my only role is to make sure that character is consistent with the previous chapters. I think if you have a big enough swarm of agents that are very specialized, then we can really break the question of Is it reasoning or not? Because technically it's an orchestration that mimics reasoning.
Manav [00:20:12]:
It's so identical to a real process. If you see shows like Seinfeld or Curb youb Enthusiasm, it's like the writer's room. You know, they're arguing, they're criticizing each other. It's basically you're mimicking the same thing. But that's such an interesting concept. I think I'm going to look for that.
Laodius [00:20:27]:
Yeah. Another example, if you're going to do an AI news channel, which I think with our technology is going to be doable, actually, pretty soon, again, you would have the editor, the journalists, you have like a bunch of roles of different people to assess what kind of script is going to go live. And then you have the presenter, which could definitely be an AI avatar, reading that with expression and emotion. But that takes like 7, 8, 9, 10 people thinking about the content, challenging what's going to go live, what you can say, what are the recent news? Some people to try out check if it's a fake news or not. Like, for example, you'd have a single agent whose only role would be like, what is the latest news? Can I check sources and make sure it's true? And then I'm going to transform that into something that can be aired again. This process would be achieved. If you think about I'm only doing one model model, and it needs to do that perfectly. You're gonna bet, like, it's gonna happen in 10 years. If you think about it as I'm gonna do 25 very small versions of that model that will work together, suddenly the timeline is 6 to 12 months.
Manav [00:21:21]:
Yeah. I see these models being like the iPhone, you know, like, up until, like, iPhone 11, the innovation was going crazy. And then eventually it kind of like, slowed down, you know, now it's like you don't even have a reason to change. And that's how I feel it's gonna change. But it's crazy because. Because ChatGPT search feature, I don't even use perplexity anymore because ChatGPT just. It's like the Apple or Amazon copying the same product that they know, which is cannibalizing the whole industry. But yeah, thank you so much for coming on the episode. I really appreciate your time. I know you're an extremely busy person and I really appreciate you coming on the show and, like, giving your thoughts. I think there's a lot of wisdom of nuggets in the podcast and yeah, I'm really excited and rooting for you guys because the quality is impressive. There's no other company that's doing this with the level of quality you guys are doing. And it seems to be on, like, a great trajectory. So I'm rooting for you guys.
Laodius [00:22:10]:
Thanks a lot, Manav.
Manav [00:22:10]:
Thank you. Loudest.
Related episode
