Victor Riparbelli
Appearances
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
I love disappearing into these worlds that were different than the physical world around me. And that, of course, only was amplified when I discovered the Internet at around 10 years old, which opened up an entirely new world of ideas, of music and people. The Internet information was free, and I saw firsthand how technology didn't just change the distribution of content.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
It changed the content in itself. The difference between a lively forum and a book is massive, right? A blog and a newspaper, and so on. In music, which is my big passion outside of work, I saw how software instruments and sampling and drum machines gave birth to entirely new genres that had never been possible before.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Not only that, but all the people around the world making new, exciting music could share it with the world without the middlemen of labels. The video games I played got connected to the Internet. You formed communities with people around the world that you didn't know. That's what my parents thought, at least. I started my first business when I was 13 years old in World of Warcraft.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We were 50 people slaying dragons together online. And I would later learn in life that it's actually not that different than running a startup. And those early years of my life sparked a lifelong interest in media and technology, how they change the way that we create, consume, play, and communicate.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And in 2016, I discovered a research paper called Face to Face by Professor Matthias Niesner and his team. They built a system that using neural networks could produce really, really photorealistic video. And when I saw this for the first time, I felt like I'd seen magic.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And I was convinced that in 10 years, you're going to be able to create a Hollywood film from your bedroom without needing anything else than just your imagination. That took three years, and I think that's actually going to hold up. I couldn't get this idea out of my mind, and eventually I ended up founding Synthesia, an AI video company.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
along with Matthias and Lourdes and Stefan, my co-founders. And this really was a way to kind of marry my interest in media and technology and sci-fi. We started the company with the vision of making everyone in the world into a Hollywood director.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And while that's definitely still a very exciting vision, as the years kind of went on, we realized that as exciting as AI-generated Hollywood films is, it's only the tip of the iceberg.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
The really exciting potential about these technologies is that they're going to enable every single piece of content, from text messages to novels to boring corporate training materials, to be brought alive in video and audio. But before we talk more about video, let's talk about text. Text is the original way of compression for human communication.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We encode our thoughts and knowledge into symbols that can convey meaning across time and space. That hasn't always been the case. Thousands of years ago, text didn't exist, and the only way we had to share information was orally.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
In 1500 before Christ, the first alphabet was invented, and it essentially took these very complex writing systems and simplified them into just a few characters, laying the groundwork for the modern society that we know today. In 1440, Gutenberg invented the printing press, and for the first time in history, we could mass-produce written content.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
It would take until the mid-20th century before reading became something for the masses and people slowly became literate. If we fast-forward to today, text is ubiquitous, and it's impossible to live a life without being able to read and write. But as great as text is, it's an imperfect technology. It's very efficient and it's very scalable.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
But it's a very, very lossy method of compressing information. It lacks all the nuance and additional information that we get when we speak to someone in real life. Your tone of voice, your body language, where you are, all those things matter for that message, right? Text can be interpreted a million different ways depending on the receiver. But we invented emojis to make it a little bit better.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
But even emojis are not perfect. And if you think about visual communication, it's a very intuitive way of consuming information. What if I shared this with you instead? A text description of the image would have taken you 30 seconds to read it with a high cognitive load to take those pieces of symbols and turn them into an image in your mind, right?
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Your grandchildren will be the last generation to read and write. I know that sounds strange, almost unthinkable. Text is everywhere around us. We use it hundreds of times every single day, and it's woven into the fabric of our daily lives. But today, I'm going to make the case that humanity's relentless pursuit of better ways to convey ideas and preserve knowledge doesn't end with text.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Once we add a time dimension, as in video, this problem massively compounds. So it's not that strange that since the invention of text, we've been innovating towards richer and more intuitive ways of exchanging information. We invented radio, TV, internet, VR, social media, and now AI. In 2024, it's very obvious that people want to watch and listen.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
If you look at the digital economy, TikTok is the fastest growing social network. It's also the fastest growing search engine, which is very interesting. Video and audio is everywhere in the apps that we use. We send voice notes on WhatsApp. Now we're on dating apps. When we shop, we watch product videos. And my thesis is that the more we consume video, the more bored we are by text.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And I have to say, this is also true for me. even though I love reading. When I learn something new, I usually start on YouTube, on TikTok, listen to a podcast. And only if I'm really, really invested in something, I'll take the hours out of my day to read a 200-page book. It just doesn't feel that much like it's worth it anymore. And for a lot of you, you probably feel the same.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Do you want to learn music theory from a long book or from a video on YouTube that has audio? Do you want to listen to the news on a podcast on the way to work or fold out this physical piece of paper somewhere? Most people feel like this. But we all have this guilt. I have at least. I feel guilty when I watch videos and I listen to podcasts instead of picking up a good old-fashioned book.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
You hear the commentary on this. Young people are unable to focus anymore. They need constant dopamine hits from cheap content that they scroll through on their social media apps. They don't get outside their room anymore. Exactly the same thing parents have been saying for 200 years. I have a prerogative idea. What if we're all just tired of overly dense, slow information?
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Books with too many pages. Newspaper articles with filler. What if we become much more sensitive to the quality and the conciseness of the content that we consume because we now have infinite choice? What if the current generation of kids are able to learn and absorb information much faster because of technology, not despite it? Is the problem us, or is the problem text?
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We still read a lot, right? We just read from many different sources every single day, not just a book and a newspaper. We read our messaging apps and our social media networks and our blog. We listen to long-form podcasts. It's very growing, right? But we still have this idea that books are morally superior in some sense. And even though I'm giving you this talk, I still very much feel it.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
I don't know why that is. I'll let someone else do the research on that. But it's definitely very interesting how we psychologically attribute more value to the written word. So why is so much information still text if you really prefer video? It's pretty simple. The answer is cost. Cost in time and cost in money.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Today, we have to pick between the speed and scale of text or the accuracy and engagement of video content. And so there is this underlying economic incentive where only the content and the ideas and the knowledge we deem important enough gets converted into video and audio formats.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And so in the entertainment industry, this means that we get Fast and the Furious 278, rather than avant-garde cinema from film students. In the corporate world, this means that the Super Bowl ad is a video, but the fire safety video is a long, boring document. And This is about to change in a big, big way. AI will change that equation completely.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
With AI, we can get both speed, scale, accuracy, and engagement. AI can create highly photorealistic content digitally. Computers can learn what the world looks like, and they can replicate it and remix it in amazing details. This is going to usher in a new wave of creativity, and it's not going to be driven by Hollywood.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
It's going to be driven by YouTubers and young people with great ideas who take these tools and tell amazing stories. At Indesia, we focus on AI avatars, digital humans that sound and look like us. They can even be us if you make a clone of yourself. Today, our avatars already interact with millions of people every single day.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
They teach school subjects, they onboard restaurant workers, provide health guidance, and sell products in more than 130 different languages. And they're getting really good. Very soon, they'll be very difficult to distinguish from reality. So with these technologies, we can create anything without the need for cameras.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We can bring our imaginations to life without the traditional barriers of skill and cost. But what's more exciting about all these new technologies is that they're going to make it extremely easy for anyone to create content. We've already seen how keyboards and computers turned all of us into writers. We've seen how PowerPoint turned all of us into designers.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And with AI, everyone is going to be able to be a director producing Hollywood-grade video without needing any training at all. At Syntheser, we've already turned more than a million people into video creators, and it's really fun to see all the things people make videos about, but they don't have to ask their boss because they can just log on and make the videos themselves.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
I think we're at the dawn of a new era of AI-enabled communication. And I think that future generations will slowly replace text with more intuitive forms of communication, like audio, video, and eventually immersive technologies. And one day, I think we'll look back at reading and writing as historical artifacts, like we do with papyrus scrolls or hieroglyphs or cave paintings.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Why do people make videos? It's not a very well researched area yet, but we've been working with UCL in London to figure out how do we learn differently with video and AI video versus text. We did a study with 400 participants. It's still ongoing. We plan to publish the results sometime next year. But there's some really interesting initial findings.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
When you just ask people, do you prefer to learn with AI video or text, the results are pretty astonishing. 77% of people prefer to learn through video. So faster and easier content creation will be transformative. But with new technologies, we always invent new media formats. And right now, most AI-generated content is what I call a bridge genre.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We're using it to create old formats with new technologies. Just like the first newspapers, sorry, the first websites looked like newspapers on a screen, AI videos today are linear. They have a beginning and an end. They essentially emulate what we can record with a physical camera.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Once we combine AI video with reasoning systems like language models, we're going to unlock an entirely new type of media that's going to be interactive and personalized. It's going to be able to think and narrate and personalize content for us. If you're learning music theory, you'll have an assistant that knows your skill level, knows your taste in music, and build a curriculum around that.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
All of your kids maybe will have their favorite celebrities teaching them math in school. And they'll do it in a context that's interesting for your kid. Maybe that's soccer or sci-fi or whatever. Education is going to be turbocharged with these new AI systems. Entertainment is going to change. It's going to be shaped by the viewers and by the world around it.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
We'll probably have interactive films. We'll have TV series that never end. It sounds weird, but in many ways, this is what a social media feed is, right? It's just an endless story. Your TikTok feed is also an endless stream of personalized content just for you.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
And when we combine this with new form factors like AR, VR, brain-computer interfaces, media is increasingly going to appear like real life. Maybe we'll be transmitting messages to our friends just via thought in the future. And our inbox is not going to be text. It's going to be videos of all your friends and your relatives' AI avatars talking to you.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
It sounds weird, but I don't think it's that far off. And I promise you that your fire safety video at work will be a Spielberg-esque masterpiece in the not-so-distant future. These technologies open up a Pandora box of questions. There's ethical questions, there's political questions, there's design questions, and there's commercial questions. I'll leave you with a few provocative questions.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Do you care if our content is AI-generated or not? Do you care when a film uses special effects or a green screen? Do we care that actors in commercials are paid to play someone else and endorse products that they've probably never heard of? Do we care if actors are real or generated? Are you a fan of Harry Potter or Daniel Radcliffe? Does it matter that Daniel Radcliffe exists in the real world?
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
You'll never meet him, probably, right? What about Mickey Mouse or Pixar characters? Can we feel something for them despite them being AI-generated? Will we trust AI agents, or do you still want to call the call center to speak to a human? Can we be entertained by computers interacting with each other? Would you want to play computers, play soccer, or play chess against each other? Maybe not.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
The future is ahead of us, and while a lot of this sounds like sci-fi, it really isn't that far off. And it's up to all of us in this room and society to make sure that we build a really awesome future. I think we have the technologies, and I think we have the people to do it. Thank you.
TED Talks Daily
Will AI make us the last generation to read and write? | Victor Riparbelli
Don't get me wrong, I love reading. This is not a personal vendetta against text. Some of my fondest memories is walking around my local library back in Copenhagen, picking out all kinds of books, mostly science fiction books, returning home to read them, so I could go back there again and pick out the next set of exciting books to read.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
This is 20VC with me, Harry Stebbings. Now, six or seven years ago, this Danish entrepreneur walked into my office and presented his vision for the future.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But from the outside, it looked like, oh, it's just like cute UK company doing these avatars. It's kind of fun, but like most people probably don't use it for anything else than just like making a fun video for their mom. And so for many years, we had the market to ourselves. And then slowly people started to realize that, okay, actually this is maybe not just like a flash in the pan cool demo.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Maybe there's something real under the hood. And so we started getting a lot of competitors. We have one particular competitor that, I mean, literally verbatim copied our mission statement, like everything I say. They like just copy everything that we do.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
They know who they are. But they copy everything that we do from the way we talk about the product, we talk about the market. And of course, that's annoying, right? But that's capitalism. That's good. And then eventually, you start to learn from your competitors. They do something right. They do something wrong. And it's actually very helpful.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's really difficult to pioneer space because you have no feedback loop except for whatever you do and how the market reacts to that, right? And seeing what competitors are doing and seeing what works and what doesn't work is extremely powerful. I think for us, at least, that's been a very powerful thing for us the last 12 months.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think the reason that we had a massive re-acceleration in the second half of the year, and a big part of that is actually because our competitors did a whole bunch of marketing for the category. We clearly had the superior product. So that actually helped us quite significantly. I don't think you always have to be the first to do something.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
If you're good at being a fast follower, that's also very powerful. It's a different way of running the company when you have a lot of feedback signals in the market around you. But it's much more fun. And I think it helps you get to the right answer for your customers much faster, which ultimately is the most important thing.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think that's pretty spot on. What I see a lot in the enterprise is buyers don't really know what they want. A lot of people have been told that they need to have an AI strategy, they need to execute on AI strategy, which means they're very willing to have conversations. They're also very willing to spend their innovation budgets on doing things.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
but they don't really know what they actually need and want. They don't understand the technologies well enough to kind of themselves figure out what do they need for their business. And that is both an opportunity, but I think it's also a problem for a lot of AI startups that doesn't have that customer obsession.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's great because you have a lot of budget available and people are extremely willing to do things and sign up for pilots and POCs because they want to deliver to their boss that AI strategy. But when they don't know what they actually want, it's very difficult to prove the RIs for them, right? So what is the problem?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think the problem is that to be successful today as an AI company, I think you need to be extremely customer centric. You need to deeply understand your customers. What are their problems? How can your product help solve whatever specific problem that they have? Now that sounds like pretty obvious, right? That's not like AI specific.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But the issue is that today there's a lot of companies who do a lot of like cool technologies and they go out and they kind of convince the customers that whatever they're building is the right thing. It'll help solve some like big problem for them, but maybe it doesn't really work or they haven't understood the customer's problem deep enough.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And so now you have this thing where the AI startup kind of thinks that they're delivering real value, but actually it's not a good signal because on the other side, you have a buyer who's like, just spent a lot of money on doing something. And they'll like say, yeah, it works really well because they just spent $100,000 on doing it. But eventually they will churn, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Because it doesn't actually work or doesn't actually do the thing that you wanted to do. And so that's a problem because the AI startups think they're doing the right thing, they think they're delivering value, but at some point you're going to be hit by that wall of churn where all your 12-month contracts begin to phase out because you're not actually delivering value, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Are we seeing that wall of churn moment now? I think a lot of companies are seeing that big time, right? And I think what's very unique about it about AI the last couple of years is that people are extremely willing to part ways with their money. People don't mind like paying on a consumer level, like $30 a month to like try out something that looks cool.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Enterprise level, like sign for 50K pilot to do something. But the real signal is not that you sign a contract. The real signal is renewal. And I think there's too many AI startups who optimize or have optimized too much for like closing new contracts, not for the renewal, right? If you optimize for the new contracts, not the renewals, you
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Unless you've hit the right thing, which of course some people do, then you're in for a whole bunch of trouble.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I mean, I think we're definitely in a bubble, right? I don't think that's necessarily a bad thing. I think that's how capitalism works, right? For a lot of money, a lot of products at the world, and you try to do like a million different things at once and you kind of see what sticks. And that's the right way to innovate, right? That's Darwinistic by nature.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But I think there's a lot of money that's going to go up in flames in AI products that either just aren't that valuable or things that eventually become features in some of the big cloud providers.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Buzzwords always take me out. When people say they're building AI agents to do all sorts of different things, that always lights up my bullshit detector a little bit. But I just feel like when you're overly obsessed about the technology and the latest buzzword, that's usually a yellow flag for me. Maybe I'll give you one concrete example. I really, really hate it when people call AI employees.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think it's so dumb. I think it's not helpful to build useful technologies that people want to adopt. And I just think it's the wrong way of thinking about that you're going to have AI employees doing all sorts of different things for you. These are algorithms. It's a piece of software.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You wouldn't say that Miro or Figma is an AI employee that sits and takes people's design and put it onto something, right? I understand that it's because people think these things are going to be making decisions autonomously, but I just don't think it's that different from software that we already know.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think in general, in the tech industry, we're very good at setting really high expectations and then not always meeting them. I always try with these hype cycles to stay positive but rational. What does that mean? Well, it means, yeah, sure, probably at some point that will happen.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I care most about what I know to be true today and in the next couple of quarters, and I'll adjust our strategy based off that. I think it's very clear we're going to have very capable AI that can produce software. I also think when I speak to a lot of developers that
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I mean, we're not at the point yet where you just sit down and say, hey, build me, get the next social network with these functionalities. It just comes up with it. I think building software is a lot harder than that. And I think what we see a lot with, I have seen the last two years, a lot of cool demos. And that's great. That's a good start. But what we really want to see is this applied.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
mass scale in production because you can prompt like a tetris game the browser or something like that that's great but if you look at most software out there today it's very complex and it's not just technology right a lot of humans it's customers it's feedback loops and i don't think we're anywhere near being able to um to automate those things but i definitely think that a great software engineer will be way more productive in the next 6 12 to 18 months but
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You know, there's some podcast on point, especially 12 months ago. Also, friend, it's that, right? In the future, it's going to be, you're just going to go on to whatever LLM. You're just going to say, make me a copy of Monday.com. You're going to deploy it. You're going to sell it at one-tenth of the price. It's going to be great. And all the existing companies are dead.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think that's a really, really naive view of what it's like to build a business. Technology and lines of code is super important. But as you know, right, you've built your company yourself. It goes way deeper than that. And I think it really... For me, that's things that people who've never built operate a company.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
For sure. I think what it'll come back to, as it always does, like distribution is king and great products are king. And of course, there will be new LLMs that'll be better and more powerful as everyone else excited to see what GPT-5 kind of has in store. But I do think we are seeing the commodization of the text generation layer, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
For most of the use cases that we'll see LLMs transform the world as we know it today, I think the current generation technologies are good enough. It's about, of course, improving the base models, but it's a lot about like building the product, the scaffolding around it. I definitely think we are seeing that, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Like if we're seeing like X and Elon Musk catching up like pretty quickly and Frobic has a really great product. A lot of people prefer those models over OpenAI's. OpenAI has a huge distribution mode and I think they've really managed to capture like the consumer version of the world here. And that's definitely going to be really valuable.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But I think it's a lot of distribution of product from now on.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I would do X because I think there's the most asymmetric upside if Elon delivers what he usually does. We've all heard of that data center that he built like 10 days or something, right? I think I would just never bet against Elon. And I think the upside potential there is huge. And I also think that the fact that he owns X is really, really powerful.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
OpenAI clearly managed to capture the consumer as in the destination that you go to to use an LLM. I think we'll see LLMs being a part of many different apps.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think owning X along with building the LLMs is actually really powerful, especially for the real-time information that will be able to feed into the models directly from X. And the fact that X, of course, already has hundreds of million users that in theory at least could start using their LLMs rather than going to
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Clearly, there are still gains to be made from scaling, I think, especially if you look outside the text domain, but like in video, audio, etc., 3D. So you agree that scaling laws will continue? I think they will continue, but I don't think it's just going to be linear, like whoever has the most compute is going to win. The world rarely works in those ways. Something will happen, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Someone will come up with an algorithm that is like 100 times as efficient as what it is today. I think compute is important, but I actually also think data is... Actually, I would say compute algorithms and data, maybe. I think in algorithms, what everyone is trying to build into more or less any AI system today is control, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We've proven these things are extremely capable at replicating the real world, producing video that looks real, audio that looks real, text that sounds real. But what we all really want to do is get a deeper level of control over these things. In my world, it's like you have some of the big video generation models.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I kind of delineate what we do versus like what Zora does or Runway or something like that. Zora and Runway, these models, extremely capable, extremely powerful, right? You type something in, basically anything you type in, it'll actually spit out. And that's really, really powerful. And it's a great demo.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But to really be able to use this, you have to be able to say, I want this same character in a different scene. I want the character to say this particular, like there's so many layers of control. And those are very hard to build in because every time we try to build them in, it decreases the overall fidelity.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Because all of a sudden, you're trying to make the model do what you want it to do, rather than just replicate whatever a real kind of looks like. And so I think getting those layers of control in is going to be difficult. I think that is going to be a lot of algorithms figuring out, like... To what extent is Grok's lack of layers of control a problem or a benefit?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Well, there's controls like content moderation, and then there's like controlling the output, right? So the thing that I'm talking about here is like the slot machine kind of thing. You type something in, you get something out. Then you try and change the problem, but you pull the slot machine again, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And it's very frustrating if you're trying to get to something specific because the model just doesn't really put it out. You want to have some degree of control in there. I think on the moderation, that'll be interesting to see how it's going to pan out, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think we're definitely in the middle of like a vibe shift with Zuckerberg also pulling back moderation, which I think in general, I'm in favor of that. I don't think that like humans moderating content is the right way forward. I think community notes.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I've seen the statement. I don't have the actual details of it, but I think it is directionally correct that having humans sitting down and evaluating content is not the right way of doing it. I think what we've seen with products like Wikipedia, for example, is that the collective power of people working together to arrive at some sort of truth is really, really powerful.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think community nodes is kind of like taking that Wikipedia way of thinking about the world and trying to implement that into every single piece of content. And it's not easy and it's not solved yet, but I do think that is the right way for us to have some degree of control of what people do and say. Where it gets really messy is this kind of like... gray content.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We have the problem at Synthesia as well, right? We have basically a product internally for content moderation. The hard thing is you have what we call the green content. The content that everyone agrees is great. That's 99.9% of the content. You have the red content, hate speech, violence. Most people will agree that that's bad as well.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Then you have all the gray middle, and that's where it gets difficult. I think this is the content that Sock is talking about here, where he's going to let off controls. This is essentially like you're making cryptocurrency content, for example. When is it a really enthusiastic entrepreneur who just started a new coin that he or she definitely thinks is going to change the world?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
When is it an outright fraud of trying to get well-meaning people to dump their money into something they think is going to go up 10x? Those lines are very difficult to draw, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Yeah. So today we are, and that's a decision that we have made. We had a lot of discussion about this in terms of how do you, what's our approach to this problem? And for me, again, it comes back to the customers, right? We're an enterprise product. It's important that the avatar that we have are not seen to be used in all sorts of like wacky content online.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We don't, as a company, have, we don't see ourselves as having to uphold any kind of right of free speech. And frankly, from a business perspective, having someone pay me $30 a month to create very questionable conspiracy content, it's just not good business. So for me, all of those things kind of lined up, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Our enterprise clients, they don't want us to be affiliated with content that doesn't kind of like match the brand. And economically, it just doesn't make sense. So we've taken a very strict approach. Which means that we are actually going and being the arbiters of truth. And we definitely have people who are unhappy with that.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think this is like classical VC brain. Too much focus on the technologies, right? Like, you know, you speak to a... I think there's a big misconception about Synthesia is that we're like an avatar company. And it's kind of a fair assumption to make because you go to a website, if you like, you know, see about us online or like... Advertising is a headline feature for the company, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's one of the things that made us where we are today. If you go and talk to our customers, they don't buy us for the avatar model. They buy us for the workflow. Like the way we've taken the entire video value chain. You get an idea, so you make a draft for that video. You don't have to use a camera or a voiceover artist, right? Because we have the AI models to do that.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We give you a great editor, like using PowerPoint or Canva. We give you a distribution mechanic with an AI video player that's made to serve multilingual content. The list could go on and on, right? Talk to our customers today. The reason we have, you know, million dollar plus contracts is not because people are like, oh, I want to make an avatar video.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's because they want to convey a message to someone and they want to do that with the highest efficiency and engagement and in the best workflow, frankly speaking. And that is so much more than just the models. So, of course, we want to win on the models, and we are going to win on the models. And our specific niche here is humans presenting content, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Whereas all the other companies, yes, Sora, Runway, all these guys are doing amazing AI video things, but they have a much, much kind of wider merit of what they're doing. They're trying to build true foundation model video models. And maybe we'll be customers of those one day, maybe we'll be partners. The only thing I care about right now is voiceovers and humans presenting to the camera, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think in that niche, and if you take the entire platform as a whole, That's how we win. I think that's something we're going to see the next couple of years is that the winning companies are going to be the ones that are not overly obsessed about the AI component of their platforms, but who build for workflow.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think, you know, as I move into going to 2025, models super important for us, but expanding across the value chain, we increasingly need the publisher of videos for our customers as well, which means that instead of like using, you know, YouTube or another kind of enterprise video hosting platform, they use us directly.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think that's where you're going to see a lot of the value being created. There's going to be a few foundational companies that'll be very big. They already power us with, of course, we use OpenAI, we use Anthropic. What do you use most and has that changed over time? We're mostly using OpenAI.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Yeah, we've shifted some workloads to Anthropic. We're also using Gemini actually. I think in general sense is that it's a lot about like pricing. Some models are definitely like better at things than others, but we have like a lot of like workloads that were big workloads. Moderation, for example, is one of them. So I think pricing is important.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And again, like the scaffolding infrastructure around it.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We only want to build models if we think we can be the best in the world at it, which in general for us means that it's a narrow domain and that it is directly tied to a bigger workflow. And so we always work backwards from what our customers are trying to do.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
If we think we can be the best in the world at it and we think there are compounding advantages of building that model, then it's great for us. And what that means very specifically is that for humans presenting to camera, to dialogue-driven content, that's what we want to be the best in the world at.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think there will be some concentration for sure in the big companies. But it's kind of interesting because like specialized models, right, it's still very technology centric. I think there'll be many, many, many different products that people use in their daily lives to run their businesses.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And a lot of those will probably use models that are like an open source thing that's kind of like tuned specifically for something and integrated into like an overall platform. And if you think of that as a specialized model, then yeah, I think we'll see a lot of specialized models, but it's very like tech industry to talk about models all the time, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Like I think most people, we will interact with models like we do in ChatGPT or whatever today for sure. But a lot of like the things that you won't notice as a consumer, when you're just using someone else's platform or product. There'll be a lot of workflows in the background that are also driven by LLMs. You'll definitely see a lot of companies building their own specialized LLMs, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But I think we should think more about products than models.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
it's going to be incredibly difficult to do anything that can mirror the chat GPT moment when it first came out. As humans, I think we saturate incredibly easy. We often forget how crazy just like GPT-free actually is when you use it, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So my sense is that when they release something, it'll probably be good and there'll probably be a whole bunch of tech nerds will be very excited about all the things you can now do. And maybe I'm wrong and it's just like super intelligent thing. They'll just like self-improve and save the world. But I think it'll come out and it'll be like good.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It'll be definitely better than the previous things. But I think most people will not like care that much because the bar for someone to really care that much is incredibly high, right? And there's all this talk about this is like AGI, it's not AGI. And I think all those things kind of riddled with like, well, what's the definition of AGI, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
If you took ChatGPT 200 years back in time, right, in England, and you showed someone this magical thing you've built, you'd be burned at the stake. If you did it 50 years ago, people would definitely think, this is AGI. I'm talking to a computer that knows everything about the world. This is absolutely batshit, right? And then today, people are like, ah, it's just a stochastic parrot.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I don't know what's true or what's not true there. I think it'll be very powerful. I don't think it'll have as much as kind of like the cultural moment that ChatGPT created initially. And as I said earlier, I think elements of this that there'll be like a small group of people who really care deeply about what benchmark does it work against?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And it's like slightly better in this area and in this area. And I think that's awesome. But I think for the average business around the world, the current models are already pretty good and can solve a lot of problems.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think what has happened in very broad strokes is that we democratized distribution with the internet. There are no gatekeepers. Anyone can make a website. Everyone can share that content online. That's been really, really powerful. Then we saw, to some extent, the democratization of content creation because all of our smartphones have cameras in them.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
The price of camera equipment dropped and all those things, which, I mean, this setup like 30 years ago would probably have been like 100x the price of what it is today. What's about to happen now is that we're truly going to democratize content creation.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We're going to change the world of making content from something that you need to where we capture things with sensors in the physical world to being able to generate everything digitally, right? And that's going to have huge implications. When you think about text, that's just like a fabric of society. We don't really think that much about text, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But text went through this journey of like, you know, the printing press is like one of the first really big inventions around text. Then we invented the keyboard, the computers. And now text is everywhere. Everyone can create content online. It's entirely digital, right? And that's really powerful.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
With something like music, we've somewhat also seen that go from being something we generally do in the real world and capture with sensors to something we can do digitally. So most music today is not made with real instruments. It's made with software instruments. I can recreate almost any song I want without real instruments. I can synthesize that on my computer.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And that changed the music industry in many, many, many, many different ways, right? And I think we're going to start to see the same with the video and audio industry, where you are going to be able to bring your ideas to life without needing much more than just your imagination. And this is going to have a whole bunch of effects.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Most obvious one is that the price of creating content is going to go to zero, just like it costs you nothing today to write a book. Everything is free. In the not too distant future, the price of creating a Hollywood film from a purely technical perspective, I think is going to be zero as That's going to mean we're going to have a deluge of content, an absolute deluge of content.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It means that people like yourself, YouTube creators, everyone else can compete with Hollywood in terms of what they can bring to life. Today, there's like still the separation of like Hollywood content. So like all the Netflix series that we watch, the really high production value things. And then there's like the YouTubers who can do like prank videos and podcasts and kind of like fun things.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But like even for you, right, it'd be very difficult for you to make like a Hollywood-style production because it'd be very, very, very expensive, right? That's going to like flatline, which means that the best ideas and the best content is going to win. And I think that's going to be pretty.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I totally agree with you. I think a very important delineation is when I say that you're just going to need your imagination, that doesn't mean that it doesn't require skill, right? It means that the kind of sort of technical barriers, you don't need to have a camera anymore. You don't need to be necessarily like amazing at visual effects.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
A lot of stuff will be able to be done automatically, but it'll put more emphasis than ever on being a great storyteller, on understanding what's important, on asking the right questions, on if you're creating fictional content, combining things that work in the cultural side, guys, but still stands out a little bit enough that it's like interesting for people to watch.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
right so i think all those things will be more important than ever what would be less important well knowing the right people to distribute your content having spent like eight years and becoming the world's best visual effects artist because you can get a lot of help with that from ai systems but the storytelling piece is going to remain exactly the same
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I also think what we are going to see is that the value of real content is going to be higher than ever. Because we have a lot of AI-generated content, and I think that'll be a genre of itself. But just like we still like to watch people play piano or know that the artist we're watching is actually a real person, I think that will carry over to this world as well.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Definitely AI-made content, for sure. But I think what's actually more important is what content surfaces, right? Because if you go to YouTube today, I mean, it's like, I don't know what the status, but it's probably like 95% of all videos have like less than 10 views or something. So the content is there. It's a matter of like what actually gets surfaced.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I'm sorry for you that we're meeting under these circumstances.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think a lot about this, and I think TikTok is actually a really great product, potentially a bit of an unpopular opinion. When I look at my TikTok feed today, I would say it's 90% highly informative, educational, really great content. It's about music production. It's about the music that I like. It's about politics, about tech. I see your face once in a while.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think it's an amazing feed of interesting people. What TikTok has done really well is that they built the graph based on your interests, not on your social connections.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
of that as well and i think that model is actually extremely powerful what i do want to see is a bit more control over your algorithm right so today it's basically like i think everyone has this like intuitive thing which is like if you see content you don't want to see more if you swipe really quickly because you know the algorithm will kind of notice that you swipe quickly past that so you're kind of almost trading the algorithm yourself i think that's like an awesome way of doing it but a bit more control maybe you can say like these particular topics just don't show it to me i'm not interested in it
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And then I also think what TikTok has done really well is because they have this like really great interest graph, they know a lot of people who are very interested in specific areas. When you post a piece of content yourself, they basically show that content to like a couple of hundred of people and they quickly sense, do those hundred people like the content or not like the content, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And if they don't like the content, you go in the trash. If someone kind of likes a little bit, they'll show it to a bit more people and they'll try to figure out what's actually great content. I mean, if you look at TikTok's usage, right, I think that's very clearly working really well. Should TikTok be banned? I tend to believe that it's a moral panic more than anything.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Clearly TikTok has built an amazing product. I don't think it's a brainwashing machine. I don't think it's a propaganda machine.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I totally agree. And I think that's the paradox because probably for you and maybe for me, like we built our algorithms to be productive to our lives. We're kind of starting to find our way into like, how do we build a really awesome media ecosystem that is driven by citizens? and experts rather than journalists and politicians.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And where we still have some of the control that you can argue has been good around the previous generation of media where a few corporations owned the newspapers and there's this chain of command of before content goes out the door, it's been reviewed by six different people. And I think what just always happens in the world is that we invent new things as humans and we learn as we go, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And there's definitely some negative consequences. I don't think anyone would disagree with that, but I think we are finding our way. I am actually very optimistic about the future moving into
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We just raised a Series D, which is a super exciting big milestone for us. We raised $100 million, led by NEA, with participation from all of our existing investors. So we're very excited to get into 25 with a big war chest and just get to escape velocity, shut down the category that we're in, and win.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Yep, exactly. Exactly. I did a TED Talk like a few months back, but I actually talk about this. But I actually talk about this because I think this- So you did a TED Talk?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
No, what I should talk about there is that my kind of like a provocative opening statement is that we may be the last generation to read and write, right? And I think there's a lot of like moral panic about TikTok, for example, like, oh, people's attention spans like decreasing massively. And maybe those are real problems. But what if we're just tired of overly dense, long, boring content?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And that actually a two minute long TikTok video is a better way of understanding a topic than a 20 minute long YouTube video, which is a better way of understanding a topic than a 300 page book or
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think we're dopamine junkies for everything. That's how we govern our lives most of the time, right? I mean, I totally get the analogy, but is it all the dopamine addiction if you do sports every day? Whatever thing you do that makes you feel good, I think most of the things we do is to try and make ourselves feel good in some way.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I do think we need some degree of identity verification for some of the big spaces that we use. I think we need to know somewhat where content comes from. Who created it? When was it created? How was it created? So I think we need identity verification to some extent. But I also think we need to take that all the way down to kind of like the content level.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So if you imagine that you fingerprinted, you know Shazam, the app on your phone. Of course. Imagine you had a Shazam, but for every single piece of content on the internet, it could tell you, hey, this picture was originally uploaded five years ago. It was uploaded the first time by Harry. Maybe in the future, it'll say this was partly made with Synthesia.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You're watching a dubbed version, so this is AI for translation. Whatever kind of signals we can give around content, if you had that and identity verification and community nodes, every time you scroll past a piece of content, you would see if it's verified or not verified. Right. Today, everything, well, it's changed a bit now, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But it used to be that only celebrities had verified accounts. Why was that? Well, that's because a lot of people tried to make an account pretending to be some celebrity. And so as a consumer, when you see that trusted, verified sign, that tells you that that's actually the celebrity posting that content, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think we should try and flip that around so that everyone and all content is kind of verified. If we have the provenance trail of it, we know kind of where it came from, maybe how it was created and some other kind of things around it. And that'll give you like a green check mark on that content. and it should stick out like a sore thumb is something that's not verified.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So if there isn't a fingerprint for the content, if we don't know where it's from, that should be the thing that sticks out, not the kind of verified sign on a celebrity's page. Does that make sense?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
No, because I think right now, is it real or is it fake? We're already in a world where most content that we consume today is to some extent fake, right? Most pictures on Instagram have been touched up in some way, shape or form. When it leaves the sensor, they already do color corrections and a whole bunch of other things. And so I think this line will just increasingly blur.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think it'll be less about how you made something and more about the content itself. You could make a video of you talking about something and in a couple of years, no one will be able to tell if it's actually real or not. But it won't really matter that much. What matters is that you made it, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And that whatever you say in that video is something that you actually wanted to say more than how you made it. It's a bit like we don't care if content uses green screen today. We don't care that much if when you go to see a Hollywood film, right? I know that's fiction, but you don't care how it's made. You care if it's a good film or if it's a bad film. Yeah.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You should evaluate content based on the content itself. And in here, how it was produced certainly is a signal, but it's also a signal when it was produced, who produced it. Are you watching the original version? Are you watching an edited version of it? It's really hard, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's not because it's like right around the corner, but you can imagine a world where you can begin to build a chain of like, where was content originally created? What happened to it since? Are you watching an edited version? Maybe you're watching a 25-second outtake from Harry's podcast show and you want to watch the full segment. Click here.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You can trace it back to actually watch the full episode. We can begin to build these kind of things. And the reason I mentioned Shazam is that Shazam exists because this was a commercial problem for essentially YouTube, right? Record companies were telling them, hey, people are uploading videos on YouTube, but they're using copyrighted music.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We're going to sue you out of existence if you don't fix this problem and pay us some money. I said, what did they do? Well, okay, we have to detect if something is actually copyrighted, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So when you upload a video to YouTube, as I'm sure you know, it'll scan it and if it's copyrighted music in it, it'll either take it down or most of the time it'll just put an ad in the middle of it and it'll pay something back to the license holders. Now what's actually happening here, right, is that you're uploading a piece of content
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And in, I mean, milliseconds, that piece of content is held up against a database of billions of songs. From that little short clip, like a four second clip of a song, it actually identifies in milliseconds what song it is. And if it's a copyrighted song, they'll put something in there.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So if you imagine that happening as well, you upload a video and I upload a video on my channel with maybe with this interview, right? But I've cut out 10 minutes of it. It'll just say, this is actually a piece of, this is a clip from Harry's original podcast, which was released two weeks prior to Victor putting this on his YouTube channel.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And if you want to watch the full thing, you can click to it here. So begin to build this like chain of content and we can identify where things came from. That does require we have a centralized database, decentralized ideally, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Where you just write every time you make content, no matter if you're making a video on Synthesia or taking a picture with your camera, you registered in this, I kind of hate to say it, but it could actually be a blockchain. where you register the first time a piece of content was created, and you then upload content to different social platforms, distribution systems.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You check it against that just to see if disinformation, one of the big problems about what people mostly do, right, is that they take a picture from a war five years ago, and they say, this happened yesterday in whatever country. And if you had a system like this, if I tried to do that, actually what it would do is it would say, well, this picture was uploaded five years ago.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So this is back in 17. So it's a long time ago now. And the very short story is that we're a bunch of people. We had this sort of idea that generative AI, which back then wasn't really a term that most people thought about. But generative AI would change how we create content. And the big shift was that back in 2017, when most people thought about AI, it was about analyzing data, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So when you're looking at this picture, It'll have, you know, a little block of notes on the community notes, right? But automated version of community notes saying this picture was actually taken by a BBC journalist like seven years ago. A lot of this stuff is hard, but we actually have a lot of the piece of the technology to get there.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think when we think of like content identity verification, we think of moderation, content creation, I think we are beginning to see the birth of an entire new content ecosystem that will look very different than what we have today.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think the very technical roles like camera operation, those types of things, they'll probably like fade away. And I think a lot of emphasis will be put on creativity, storytelling, being relatable, all those things that makes great content, I think will be more important than ever.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I actually think that most of those roles will have a natural transition to something different, assuming that you're open-minded enough to do it. So visual effects, for example, right? As we talked about before with making clips, AI can do a first pass as it, which may be better than random, but it's way worse than what you can sit down and do yourself, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
but you plus ai can probably be really really good right because if it's a two-hour episode it just figures out like all the right places that could be interesting and shows you that you can kind of choose yourself if you're a visual effects artist probably what it means that you'll just like a software engineer right means that you'll probably be 100 times as efficient and instead of relying on 20 other people to create a digital clone of me by like modeling my face you just prompt it or control the river i think if you're good at that you'll be even more powerful in this new world
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think we'll have much more content creators than we have today. Just like if you look at pre-internet, pre-computers, how many people wrote things? Very few people, right? Like in the 1930s or 40s, most people didn't like create text. There was like specific people like secretaries and then typewriters later on. Now all of us produce content all the time. And I think that'll be a big trend.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think most of the rate limiting and how quickly AI becomes a part of our daily life is actually not the technology. As I said earlier, I think we have very, very powerful LLMs that they can do a lot of great things. But the rate limiter is people, right? Because people have to use these things, have to trust these things, have to pay them, buy them, integrate them.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I mean, with technology, it's always the case, right? It takes a long time.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
The honest question is because I couldn't get to the US. I mean, I didn't have an amazing CV before I started Synthesia. The US, basically, you need to get a sponsorship if you want to go to the US, right? I couldn't really get that. I wanted to start a company. I knew a few people there, so I just kind of moved on a whim, actually, and spent a year.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Making decisions. That's kind of like that era of AI. But there was these early kind of GANs, which was basically a neural network that could kind of produce new data instead of just analyzing existing data. We thought that this was going to be a world-changing technology. We thought it was going to change everything we know about how we create content from video, speech, audio, music, whatever.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I still do think that the chance of success is higher if you're in the US on the West Coast. I think there's a lot of pros, especially like the last six, seven years, got a lot better to build here. But I do still think that there is a lot of advantages to being in the US. But I think Europe is catching up pretty quickly. What are the pros of being here?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
One big one, which is talent, kind of goes both ways in some sense. But if you look at the ecosystem in the Valley, one thing is kind of like the price of talent, which is just abnormally high, right? But there's also things like loyalty, for example, that I think are less spoken about.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think in the US, you have a lot of people who are very driven and they basically view their careers like building a portfolio of like stock options in different companies. And hopefully one of them will kind of explode and they'll be very rich, which is a rational way of thinking about it. And it makes a lot of sense, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But it also means that it's a more transactional relationship I think people have with their employer. People are much easier to jump ship, right? You have one or two bad quarters. You may lose a lot of good people because, hey, there's this other cool thing on the block over here that I'll go to and work at instead. In Europe, I think people, for better and for worse, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
They think less like that, right? As you of course know, like stock options in Europe, people still view it as like a random lottery ticket, not really as part of their compensation because no one in Europe knows anyone who got rich off working in a startup because we don't have that many successes here. Are you seeing that attitude change? I am seeing the attitude change a little bit.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
It's going to take a long time to change because the way, no matter what I tell people who work at Synthesia, no matter how much they can kind of read the news and see all this stuff, what people care about is like, do they know someone? Do they have a cousin who made $2 million of being early in a startup?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
In the US, especially in the Valley, everybody knows someone who made $2 million of working in a startup, and they know several of them. That makes you feel like you can do it as well, right? Here, it's very kind of abstract, like, oh, yeah, people, these companies made.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But before we really see that, and that just requires, like, a bigger ecosystem with more exits, more big companies, and so on, I think it'll get there, but it'll take a long time. But what I think you get in Europe is you get more loyalty in the sense that people care a lot about, like, working a place that they like working, working on interesting problems, having great colleagues.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think that's a pro that's, like, less spoken about. Do you think people work as hard? Honestly, I don't really have a strong sense that people work way harder in the Valley. I know a couple of great friends who've moved over there and opened offices. I have not heard from any of them that, wow, people just work so insanely hard.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
There's probably a percentage of people there that work really, really hard, but I think we have those people here as well. I think that's more like a personality trait that's dispersed all across the world. But I don't think if you take an average software engineer in the Valley that they work harder than we do in Europe.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Yeah, I mean, I think it's kind of interesting to see what the new government is going to do, right? I think they have a lot of great cards on their hand. Like we have a lot of great companies that are actually building here. And I think what's important or what we're seeing in the London ecosystem is that
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But we were focusing on video. I went out to the world with a great PowerPoint deck and, we think, a great vision. But I think, understandably, most people thought we were pretty crazy, right? We basically went out and said, look, in 10 years, you're going to be able to make a Hollywood film from your laptop, needing nothing else than your imagination.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We're actually building global leaders, not just regional winners of expense management or something like that. You have Eleven Labs, who are leading in voice. You have us, which is leading in enterprise AI video. We have Wave, which is definitely doing self-drive. We have a lot of great companies that actually become global leaders.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think it's really important that the incoming government doesn't screw this up. How could they screw it up? I think they could screw it up by over-taxation, all the kind of usual things that I think everyone is talking about, right? I think what they can do to make it better, I think the UK has a brand problem. What do you mean?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think we disagree on this, but I actually think the UK is a pretty good place to build in general. I come from Denmark, right? That would have been a terrible place to build a company. That would have been impossible, basically. I think with SEIS, EIS... I think there's a lot of good things that have happened in the ecosystem. You have all the talent.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think the venture ecosystem has gotten better, even though it's still very PE-centric. I still think it has gotten better. There's also still a lot of ways to screw it up, right? And I see a lot of people who are leaving, especially after the recent administration. My phone.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
The obvious things like corporate tax, entrepreneurs relief, all those things make it really attractive to build here because we are in stark contrast to Europe, which in many places are terrible. I think it's different because you're from here. You're British, right? I'm from Denmark. I think we always, wherever we're from, we're much more negative.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I moved here and I've seen a lot of the positives around it. But I think don't screw up the fundamentals, right? Don't raise the taxes too much. Make sure we still have great relief schemes for people who great jobs, great economic productivity, and those sort of things.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
On the brand problem, that of course starts with the fundamentals like crimes on the rise, living costs, all those things that I think you've also been very vocal about. Those are big, but those are big problems, right? Sorry. But what I actually would say though is that
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
No, I agree. I totally agree on that. And I think that comes a bit back to the brand problem because I actually do think that as much as we have a lot of people who want to leave to the US, one of my learnings has been that actually a lot of people want to come to London as well, especially from the US.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
In our journey, I've seen that it's actually easier to move people from the Valley to London than from the Valley to New York, which feels kind of counterintuitive. But a lot of people, want to come and experience the European lifestyle, right? They like the cultural heritage. They like the fact that London is a cultural melting pot.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
That wasn't a pitch that landed particularly well, especially not in Europe. We were based in London at the time. Why do you think that was? Because I think in Europe, the VC industry is still dominated by people who used to work in private equity. And that's a very different mindset than technologists, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
All the things I also love about London on a personal level, right? If you worked a bit more on the brand, I just think you could attract a lot of great people to also come here and build. But of course, it starts with the fundamentals, right? Now, if you look at California versus London in terms of like taxation and things like that,
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I'm not sure you can say that there's like a massive difference, right? California is also like- Super high taxes. It's very high taxes and pretty rough. So I wouldn't over-index on that, but I think definitely a case of like double down, right? There's a lot of good things happening in the UK.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think if we don't screw it up, we just need one or two companies to become plus $10 billion companies headquartered in the UK. Then I think we'll really start to see the ecosystem grow.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
No, but it's a joke. I mean, the London Stock Exchange, like... They need to fix that. They need to put some liquidity in the system. I don't know. I'm not a financier, so I don't know what they need to do to fix it. But, I mean, that is a disaster, right? And I think we've seen, unfortunately, great companies really suffer under that. 100% you have done. Yeah, it's not very attractive.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I agree.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I agree. I don't think that's my top three of things needs to be fixed. What is your top one that needs to be fixed? Keep the reliefs for entrepreneurs, for people who create things, maybe even make them better because that's very attractive. I think there's also a whole bunch of things you could do in terms of subsidizing GPU costs, for example, that would incentivize for UK data centers.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
That would be interesting, right? That's an actual competitive advantage. Obviously, don't over-regulate. We haven't seen much regulation yet. The census, that is a little bit of a wipe shift from the last administration to this one in terms of regulation. Don't look at Europe as an example, right? Europe has all the AI regulation and none of the AI companies.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Staying the course to some extent and not screwing it up, I think is really important.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So when you used to work in private equity, you're like a financier more than you're a technologist. And this was not a pitch you could understand for an Excel sheet. So we got turned down by, I mean, basically everyone, I think like 80, 90 investors, something like that. Including podcasters.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think that playing computer games is probably one of the best things a kid can do for like future success in their life.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
That's exactly what I did. World of Warcraft. No, actually, what I think it is, is that we're probably the first generation that really had access to like, especially a bit more complex strategy games. Basically, what you learn throughout life is like decision making to some extent, right? And computer games is a microcosm of the universe, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And you can run so many simulations and it may not look like it for parents if you're playing like Warcraft 3 or Red Alert or whatever game that you're playing, World of Warcraft that you're playing back then.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But actually what you're doing is you're training yourself to make a lot of decisions really rapidly and to understand what the implications of each decision that you make are in some kind of game that you're playing. But life is a game. Careers is a game. Building companies is a game to some extent, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think we're just the first generation where you actually can sit by yourself and get so many iterations in that would have otherwise been impossible. You can't sit down before the computer era and simulate lots of small decisions in how to run a business. You just couldn't do it, right? You would have to actually build a business and run a business. Here you can play rollercoaster tycoon.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And of course, rollercoaster tycoon is not like running a real theme park, even if it's just like 20% of being the CEO of a real theme park, right? I think that teaches you a lot. And so I think the decision-making and kind of sense of strategy you build from that is, I think is highly valuable because it's a two-way feedback mechanism, right? You can sit down and read books about things.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But I think what's powerful about computer games is that there's this kind of constant feedback loop.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I am a generalist in the purest sense of the word. And when I was younger, I wanted to be an artist, wanted to be a great computer programmer, developer. And I just like... I never really excelled at any of them. I still have a dream of making the music improve one day. But I think what I'm really good at, I know a lot of things about lots of very random things. I'm just extremely curious.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I'm not an expert in anything specific, but I have a very, very wide range of interests. And I think my neural network has been trained on
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
so many random things like how elevators work to black metal in Norway to like how people make technology like all these kind of different things and I kind of used to be a bit annoyed at myself like why can't you just sit down and become like you know the world's best music producer or like the world's best programmer or AI researcher or something like that and I think now I've realized that my power is actually the fact that I'm like decent at like a lot of things
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think especially in my role, right, what you do is you make decisions and you get ideas and you can't force those things. I think I really truly believe that those come from your subconscious. And the more food and diverse things you put into your brain, the better you get at like making analogies and pattern matching and those kind of things.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So I did a degree in kind of split computer science and business at university. And I really kicked myself not just taking a STEM degree, like a pure CS degree. Studying business, controversial opinion is like mostly useless. There's basically like 10 ideas in business. And if you learn those, you'll be pretty well covered.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
The first round was a million dollars. We ended up doing it at 5 million post with Mark Cuban.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I should just do like a pure CS degree or pure math or engineering, something like that.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I think that one of the things I've changed my mind on in this whole AI hype cycle, I think when you're in the middle of these cycles and you see all these things getting funded, all these competitors pop up, you get a lot of noise and a lot of like, oh, should we do these things? Should we listen too much to what the market says? Someone is doing this over here.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I think I've just really learned the lesson again, although I think we've stayed the course really well, to never do things out of guilt of other people doing it or looking too much at competitors. or what the broader kind of AI landscape is doing. Listen to the customer. They're the ones who pay your bills, and ultimately, they're the ones that you need to make happy.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I had a period where I maybe over-indexed a little bit too much on what other companies are doing as the AI space in general, what's happening there as the most valuable signal, as opposed to the customers. Maybe more like a relearning of a fundamental lesson in building startups.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I can't answer that.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
That's a hard one. I would take Mark Cuban again because he truly believed when absolutely no one else believed in us. I'll be forever grateful to him for doing that.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And with today's valuation of 2.1 billion, that's... He has like 15% today. Something in that range.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
My parents' generation, they saw computers as this like inherently socially isolated thing of like you're just sitting and like dopamine addicted to like clicking buttons. Exactly what people say about TikTok today, right? The reality is usually much more nuanced than I think.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
For me, when I played computer games as a kid, first of all, it was a very social thing, both because you play with other people online. And I think our parents have a hard time believing you can be friends with people you haven't met in real life. Mm-hmm. And second of all, that is like a waste of time, right? As I said, why do you think you're playing strategy games?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I would sideline that with like solving math problems or like reading STEM books. Of course, everything in moderation. That's one thing I'll definitely encourage my kids to play great computer games.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Not having to worry about like financial stress and having everything you own in one stock, I definitely think frees you up to make better decisions. So I definitely, I mean, everyone in Angel Investing, I definitely encourage like take secondaires off the table when you can.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I'm 100% certain that if we had taken no secondaires at all, we would definitely have thought about selling much earlier and you might have made more rational decisions, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So that was the later round. So that was back in 2017. We did a million at five. That kind of got us started. We went like 12, 18 months, as you do. Went out to the market, felt like now we had a working technology. We were back then focused on AI doping, not the kind of avatar tech that we mostly have today. Again, learned a hard lesson.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
We went out and we're like, okay, we built this great technology. We have a great team. Let's raise $8 million. That completely failed. And for like nine months, which is like dragging our feet. I made all the mistakes you can make as a founder. I dragged out the funding process over nine months, like different data points, different investors. It was a big shit show.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And we actually had to rewind because we were running out of money. And we ended up raising $3.1 million. And that kind of took us through to the series A, which is when we had actually found product market fit and had like a sustainable business. But those two first rounds were very much rounds we raised based on story.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And it was a story back then that just didn't resonate that much because it was very hard for people to see what we had, how that could extrapolate that into everything that became today.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So we talk a lot about this, and I don't think we would. I think if we'd raised eight, we could have done a lot more things, right? And we would have done more things. We would have built deepfake detection, which everybody wanted us to build. You have to remember, this is in 2018, right? At this point, AI video, to everyone outside of Synthesia, it was just deepfakes.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
People thought that that was going to be the big thing. And that is, of course, it's a problem and it's a real thing, but we always thought this is a subset of AI video. Most people are going to use these technologies to create awesome creative content. But if we got that money, I think we would have built a team to do that. And that would have meant we've lost our focus.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I actually think that operating under the constraints that we had at the time really just focused us on our customers and selling the product. We were like ferocious about charging people from day one, even if it was like £500, right? And I think that focus and working on those constraints definitely help us to get to the point where we actually found product market fit.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I'm a big fan of working under constraints. I really do think it forces a lot of discipline that easily gets lost the more money you have because you have more options, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Too much money too early is not healthy. Maybe some people are good at really having that discipline, maybe like second time founders, but it's very tempting to spend on things that you shouldn't be spending money on, especially when things are not working right. I think once you clearly have product market fit,
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
then a lot of things, a lot of decisions becomes a lot easier because they're more obvious. But before product market fit, it's really dangerous, I think, to be developing two, three, four things at a time, having 15 people working on your team when you're still at idea stage. And I think a lot of people make the mistake of raising the money and then using it too quickly.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
One thing you cannot use money for or buy away with is product market fit, right? I think learning about your market, about your customers, it just takes time. And having a team of 20 people instead of five people trying to learn that, I think actually slows you down. You have to own that as a founder.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I don't think we could have learned the same amount about our customers faster if we had more money back then. It really did take us like two years to just get into the minds of customers, understand video from like first principles. And I think people try and then you hire like a product manager to help you fix the product market fit problem because they're a product person.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Or you hire like a salesperson because you think you have product market fit and you just need to be someone who's better at selling it. Unless you have product market fit, like you shouldn't hire product managers and salespeople. That is your job as a founder.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
100%.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I totally agree with you, and I think a successful company is a long series of product market fits, but you need that initial spark, right? Once you have that, you have something to build on. The bigger you get, the bigger company you're building. You need new products, new product market fit, essentially.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And I actually do think that that's one of the things as a founder that you should always be focusing on. What is the next market? What is the next product you're targeting? But in our case, I would say we have a bunch of product market fits already. And those, I think you can hire super smart people that can run with it, make it great lines of businesses, make it great products.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But I think pushing to those next product market fits and where the company needs to be in two or three years, I think that remains to be the founder's job.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Well, I think when you have the money, you do spend it, right? I think for us, it's been more a matter of just, we've always been obsessed about actually building a business, having great unit economics, and making sure that we're a business that generates money and revenue, and that we're always in control of our own destiny, rather than being tied to a VC parachute.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And so I think we've always just spent conservatively, but it's also very clear, right, that if you know where to spend the capital, it is an amazing asset, right? So building a great go-to-market team, for example, I mean, that does cost money and that there's like a cash flow thing of like you have to hire like a bunch of people that are really great. They have to train.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
They take nine months before they wrap up and then the investment is worth it. You want to have capital, right? You want to have a really healthy balance sheet so that you can chase any opportunity that comes ahead of you. For us, like we want to build like a really, really big company. I think there is Easily a $50-100 billion company can be built in the space that we're in.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And we want to win that. You're not going to win that by bootstrapping all the way, right? I think that's a myth.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
So I think we're in the early stages of a shift in how we communicate. If you think of most communication today, it's text-based, right? Emails, text, we read things. And text is a great technology. Built the world up to where it is today. But it's actually like a pretty bad way of compressing information.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
You lose a lot of context when you transform your thoughts into something that's written down in a document, right? As humans, we're much better at consuming visual content. We like to hear things, we like to see things, we like to feel things in the physical world. We can't do that yet.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But it's very clear that higher fidelity content, like video and audio, is a better way of training, informing, and entertaining people. The reason that we're using that much text today is because text is the only scalable way we have of essentially storing information and sharing information, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But that's changing now because the more we don't need cameras and microphones and capturing things in the physical world around us, the more we can get video creation, audio creation to be as scalable as text. And once that happens, there isn't really any reason for us to use text anymore.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
This sounds a bit crazy, but I actually do think that maybe not us, but maybe our kids' kids are going to be one of the last generations to read and write as a default way of communication. I think we'll increasingly just consume everything via video and audio. I think you have all the trends. You look at TikTok, right? I don't know if you're on TikTok, but
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I do exactly the same thing. I think TikTok is a great example of how video truly can be the default of how we consume information, right? There's almost no text left in the interface, even comments, right? People respond to comments with videos. And so they've really started to build out this graph of how we communicate with video as a default medium. There's a long way to go.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
But if you believe that that is true and that not just your TikTok scrolling at night, but if you're trying to buy a software product, you actually don't want to read a whole bunch of stuff and jump on calls with people. You want to watch a video. And at some point, you probably want to watch an interactive video, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Where you just, you know, with your voice say, hey, can you show me how this functionality works over here? And the video will just like switch over to that. When you do custom support, you're not on the phone, you're not reading long knowledge articles, you're watching videos. Again, probably they'll be interactive. Then all the world's communication is the market.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
When we started the company, we always talked about, and it's a very important part of our thesis, is that our market is not video production. Because all video production today, that's an interesting market, but all of our market is text. Text and slides is the market that we're targeting. And that market is infinitely big.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
And if you manage to capture just 5% of all the world's text communications and turn that into video, I think you have probably more than $100 billion company.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I mean, I think it can backfire if you don't know what to use the capital for. You start doing stupid things, right?
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
I mean, sure, it's a part of it, right? But I think more than anything, this round is raising the capital to build the best product in the category. So, I mean, every time you raise the signal, like hiring people, you send a signal to competitors and VCs, et cetera. But I've never made a decision to these based off what our competitors do or don't. I think that's a bad way to run a company.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
Do you not pay attention to competitors? Of course you pay attention to competitors. And now we have lots of them, right? I think it's great to have someone to kind of play ball against to some extent. One of the interesting things about Synthesia is that when we raised our A to B to C round, it was very much a secret how fast we were growing and how much people loved the product.
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Why Scaling Laws Will Not Continue | OpenAI vs Anthropic vs X.ai: Who Wins and Why | How Far Will Model Providers Go Into the Application Layer | The End State for Models: Many Specialised or Few Generalised with Victor Riparbelli @ Synthesia
A lot of people looked at us from the outside like, oh yeah, there's like a few people in learning and development and training that thinks it's like a cool thing to spend time on making these AI videos, right? And what we actually saw from the inside was holy shit, this is huge. And training and learning was the first market we targeted, but this is clearly just the beginning.