Mar 27, 2024 · 29m · mad
Hollywood Producers FREAK OUT over AI | Cris Valenzuela, CEO of Runway
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this live recording from Foursquare HQ, Runway CEO Cris Valenzuela joins Matt Turck to explore the generative AI creative revolution, discussing Runway's technological evolution, model architecture breakthroughs, shifting industry adoption, and the future of multimodal storytelling.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 13.6% of the talking time here. How this is scored →
Matt as informed peer 3.2 Guest teaching 3.3 Guest disagreement 0.3 Matt pushing back 0.4
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 15:22 Reframing company mission over modality
Cris explicitly rejects the host's framing of Runway as a text-to-image or video niche company, reasserting that Runway is defined by creative mission rather than modality boundaries.
Hardest push from Matt ▶ 14:20 Probing core model architectureMatt presses Cris on technical specifics, directly asking whether Stable Diffusion remains at the core of Runway's stack or if Gen-1 represents an entirely distinct product model.
Biggest teaching moment ▶ 12:35 Latent domain innovation explanationCris delivers a clear technical breakdown explaining how operating in the latent domain rather than the pixel domain fundamentally lowered compute requirements for fine-tuning diffusion models.
Matt holds his own ▶ 19:52 Referencing Sora launch and market dynamicsMatt demonstrates high domain expertise by quoting Cris's specific tweet upon OpenAI's Sora release and linking it to broader market movements including new DeepMind alumni funding.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Origins and Founding Vision of Runway ML | 1 | 3 | 0 | 0 | Cris provides a monologue on the founding history of Runway at NYU, highlighting early tinkering with convolutional neural networks. Matt lets the guest establish the background without interrupting or pushing back. | |
| Early Conviction in Generative AI Before the Hype | 2 | 5 | 1 | 1 | Cris corrects Matt's timeline assumption about Transformers by clarifying that LSTMs, RNNs, and GANs were the dominant architectures when Runway began. He then reframes generative media using a historic camera versus paintbrush analogy. | |
| Redefining Creative Software 2.0 | 3 | 3 | 0 | 0 | Matt asks how generative software differs from traditional tools and inquires about Gen-2 product capabilities. Cris collaboratively walks through user control features like Motion Brush and camera controls. | |
| Latent Diffusion, Open Source Research, and Stable Diffusion | 4 | 5 | 0 | 1 | Matt demonstrates technical awareness by bringing up Latent Diffusion and Runway's co-creation of Stable Diffusion. Cris educates the host on operating in the latent domain versus pixel domain to democratize model fine-tuning. | |
| Beyond Text-to-Video: Multimodal Storytelling | 2 | 4 | 1 | 1 | When Matt asks about transitioning from image to video generation, Cris politely rejects modality-based startup classifications. He emphasizes that Runway's core identity is centered on general storytelling rather than single media formats. | |
| Compute Scaling, Data Strategy, and Getty Images Partnership | 4 | 2 | 0 | 1 | Matt specifically cites Runway's partnership with Getty Images to address AI training data provenance. Cris explains how licensed enterprise data allows studios to fine-tune proprietary models safely. | |
| Competitive Landscape and OpenAI's Sora | 5 | 2 | 0 | 0 | Matt showcases strong industry context by referencing OpenAI's Sora release date, Cris's reaction tweet, and recent competitor fundings. Cris welcomes the validation and growth of the video model ecosystem. | |
| Hollywood Adoption, Creative Users, and Go-to-Market Strategy | 4 | 3 | 0 | 0 | Matt names prominent user examples like Madonna and ASAP Rocky to ask about customer adoption. Cris outlines the shifting sentiment among Hollywood executives over a five-year timeline. | |
| Building Research Culture and Startup Soul | 4 | 3 | 1 | 0 | Matt quotes Cris's recent tweet about startups having 'a body, but no soul.' Cris explains his philosophy on building vision-driven research cultures versus chasing short-term hype. |
Statements from this episode (17)
Valenzuela: Runway has millions of users, including Madonna's creative team
“We have millions of users who work with creative teams across the board from like Madonna's creative teams to filmmakers that are making like blockbuster movies to creators and like Tiktokers and YouTubers and podcasters maybe.”
Valenzuela: Generative AI is a new medium, not automated legacy media
“I think there is a common misconception that the first way to think about images and videos and content and audio and all sorts of like multimedia things that you can generate with these models is that you're going to make the previous thing that we make, but …”
Valenzuela predicts AI will soon generate rich multimedia in near real-time
“If you compound how fast things are moving, we're going to get to a world really soon. We're going to be able to generate rich multimedia formats and audios and videos in almost real time.”
Valenzuela: Modern creative software primitives were invented 20 years ago
“Creative software really hasn't changed. If you think about, like, everything we use these days from non-linear editors to vector designs to image manipulation software were all primitives that were invented 20 years ago.”
Valenzuela: AI requires new creative interfaces beyond timelines and pixel moving
“If you think about this new category of things that you can do with computers, you can interact with them in natural, with natural language, you can think with them and reason with them, perhaps in the same way that we reason, then the type of interfaces and s…”
Valenzuela: AI creative tools will not matter without granular user control
“If you don't have control over the way you're using something, then it won't probably like matter because it will be just like a system creating things for you without actually you being the one with the agency.”
Runway hired an engineer who fine-tuned Stable Diffusion on a 4090
“There's one engineer, for example he fine-tuned a version of stable efficient with, like, his own Forty-Night-y incredible work, and he figured out things that no researcher had figured out at the time. Of course, we hired him.”
Cris Valenzuela: Single-modality AI specialization is a transitory local maximum
“I think in the speed of how things are moving, that's very transitory, and it's like a local maximum. Like, it won't matter how you generate things. It will matter that you can generate things.”
Valenzuela: AI startups repeating Runway's early experiments will fail
“Now I see a lot of companies, we're trying the first things we tried five years ago. And we know where we're gonna go, and we know they're gonna fail because we were there before.”
Valenzuela: Runway works mostly with media companies and Hollywood studios
“We work mostly with media companies and studios in Hollywood”
Valenzuela expects video AI models to develop an ecosystem like LLMs
“Actually, for me, LLMs happened or had a similar momentum, like, two and a half years ago, where, like, we started also building and seeing infrastructure being built around training and deploying language models, and then we expect the same thing to happen wi…”
Valenzuela: Runway targets enterprise sales at companies with organic user clusters
“That's how we We've worked with some of these companies. We just see who else is using it. Sometimes we have, like, dozens or hundreds of people, and then just reach out to them and, like, sell them, like, a bunch of licenses.”
Valenzuela: Kanye West used Runway to create a video
“Kanye West made a video runway to make yourself done it for like movies.”
Valenzuela: The best AI research often comes from pragmatic software engineers
“Sometimes the best research actually comes from engineers who want to step and try this thing called research because they come from with a much more pragmatic view of how the world works.”
Valenzuela: AI researchers face existential crisis as model scaling solves problems
“A lot of them are going through an existential crisis of sorts. It was maybe a lot of things you were trying to solve and work are just going to be solved by training larger models.”
Valenzuela: The best AI ideas came seven years ago before intense commercialization
“And actually, the best ideas I've seen were, like, from seven years ago, where people were not too, they were less risk averse in trying new things and weird things, because there wasn't, you were not trying to raise money for a company, or you weren't trying …”