I once poured $250,000 into an AI startup that promised to write Hollywood-level movie scripts. The idea was incredible. We had a team of brilliant PhDs, a massive dataset of classic screenplays, and a belief that we were on the cusp of changing filmmaking forever. Eighteen months later, we shut it down. The scripts it produced were grammatically perfect, structurally sound, and completely, utterly soulless. That failure taught me a lesson that has shaped my entire investment thesis for AI in creative fields: we’re focusing on the wrong problem.
Everyone is chasing the dream of a fully autonomous AI storyteller. We see the power of large language models and imagine a future where AI generates blockbuster movies, bestselling novels, and chart-topping songs from a simple prompt. I’ve seen hundreds of pitches from startups claiming they are building just that. As someone who has backed companies like Anthropic and OpenAI, I understand the appeal. The technology is advancing at a dizzying pace, and it feels like anything is possible.
But my own experience, both as an entrepreneur with MovieLaLa and as an investor in over 200 companies, has shown me that this approach is a dead end. The real reason most AI-driven storytelling efforts are failing is that they are trying to replace the artist. They are building tools to automate creativity, when they should be building tools to augment it.
The Seductive Promise of AI Storytelling
I remember the early days of AI-powered creativity. It felt like we were on the verge of a new renaissance. At MovieLaLa, we experimented with using machine learning to predict a movie’s box office success based on its script. We analyzed everything from character archetypes to plot pacing. The system was surprisingly accurate, but it couldn’t tell us why a script worked. It could identify patterns, but it couldn’t understand emotion.
This is the fundamental misunderstanding that I see time and time again. We treat storytelling as a data problem. We feed the machine thousands of examples of what has worked in the past and expect it to create something new and original. But stories aren’t just a collection of data points. They are about human connection. They are about empathy, vulnerability, and shared experience. An AI can’t feel those things. It can only simulate them.
I fell into this trap myself. I invested in a company that was building an AI for game design. The system could generate entire game levels, complete with characters, quests, and dialogue. On paper, it was a game developer’s dream. In practice, the games it created were boring. They were technically impressive, but they lacked a point of view. They were a collection of mechanics without a soul.
My First Big Failure: The AI Screenwriter
The AI scriptwriting startup was my most expensive lesson. We thought we had cracked the code. We used a sophisticated neural network trained on the best screenplays of all time. The AI could generate dialogue in the style of Aaron Sorkin or Quentin Tarantino. It could plot a three-act structure with perfect precision. It could even create character backstories.
But the scripts were hollow. They were like a beautifully crafted piece of furniture with no purpose. The dialogue was witty, but it didn’t reveal character. The plot was intricate, but it didn’t create suspense. The characters had detailed histories, but they didn’t have motivations. We spent a quarter of a million dollars to build a machine that could mimic the form of storytelling without understanding its function.
We tried everything to fix it. We hired a team of professional writers to work with the AI. We created a complex system of rules and constraints to guide its output. We even tried to teach it about human psychology. But in the end, we couldn’t overcome the fundamental problem: the AI didn’t have anything to say. It was a mirror reflecting the data it was trained on, but it had no life of its own.
The "Tool" Trap
That failure forced me to re-evaluate my entire approach to AI in creative fields. I realized that I had been thinking about it all wrong. I was looking for a magic box that could spit out finished stories. But that’s not how creativity works. Creativity is a process. It’s messy, intuitive, and deeply personal.
The most successful creative tools are not the ones that try to automate the process, but the ones that empower the artist. Think about the electric guitar. It didn’t replace the musician. It gave the musician a new voice. It opened up a new world of sonic possibilities. That’s how we should be thinking about AI.
We need to stop building AI storytellers and start building AI instruments. We need to create tools that help writers, filmmakers, and game designers to be more creative, more productive, and more expressive. We need to build AI that can be a collaborator, a co-pilot, and a muse.
Lesson 1: The Human is the Storyteller, AI is the Intern
The most successful AI creative tools I’ve seen are the ones that understand this distinction. They don’t try to do the creative work. They do the grunt work. They can help a writer with research, a filmmaker with storyboarding, or a musician with arranging. They can generate ideas, but they don’t dictate them.
I recently invested in a company that is building an AI-powered research assistant for journalists. It can sift through thousands of documents, identify key sources, and even generate a timeline of events. But it doesn’t write the article. It gives the journalist the raw materials they need to tell a compelling story. That’s a powerful application of AI. It’s not replacing the journalist. It’s making the journalist better at their job.
Lesson 2: Data is the New Ink, But Garbage In, Garbage Out
Even with the right approach, the quality of the AI’s output is only as good as the quality of the data it’s trained on. This is especially true for storytelling. If you train an AI on a diet of generic, formulaic stories, it will produce generic, formulaic stories. If you want to create something original, you need to train it on a diverse and high-quality dataset.
This is a huge challenge. The best stories are often the ones that break the rules. They are the ones that defy categorization. How do you teach an AI to do that? How do you teach it to be surprising, to be subversive, to be human?
I don’t have all the answers, but I believe it starts with curating data with the same care and attention that a museum curator brings to a collection. It’s not just about quantity. It’s about quality, diversity, and a deep understanding of the art form.
Lesson 3: The Uncanny Valley of Narrative
There’s a concept in robotics called the “uncanny valley.” It’s the idea that as a robot becomes more human-like, it reaches a point where it’s almost human, but not quite. And that’s when it becomes creepy. I believe there’s a similar phenomenon in AI storytelling. When an AI-generated story gets close to being human, but misses the mark, it can feel unsettling and emotionally hollow.
This is the hardest problem to solve. It’s the subtle nuances of human interaction, the unspoken subtext, the emotional resonance that makes a story come alive. These are the things that are incredibly difficult to quantify and teach to a machine.
I’m not saying it’s impossible. But I think we’re a long way from an AI that can write a story that will make you cry. And maybe that’s a good thing. Maybe there are some things that should remain uniquely human.
Where I'm Placing My Bets Now
Despite my skepticism, I’m still incredibly optimistic about the future of AI in creative fields. I’m just more realistic about what’s possible. I’m not looking for the AI that will write the next great American novel. I’m looking for the AI that will help a human write it.
I’m excited about tools that can help with world-building in games, creating vast and detailed fictional universes for players to explore. I’m interested in AI that can create personalized news experiences, tailoring the content to a user’s specific interests and knowledge level. I’m bullish on co-writing tools that can act as a creative partner, suggesting plot twists, character ideas, and even lines of dialogue.
I believe the future of AI storytelling is not about automation, but about collaboration. It’s about creating a partnership between human and machine, where each brings their own unique strengths to the table. The human brings the creativity, the empathy, and the life experience. The machine brings the data, the processing power, and the ability to see patterns that a human might miss.
The Path Forward
If you’re an entrepreneur or a creator working in this space, my advice is this: stop trying to build a better storyteller. Start building a better tool for storytellers. Focus on a specific problem that a human artist faces and build a solution for it. Don’t try to boil the ocean. Start with a small, well-defined problem and solve it better than anyone else.
And most importantly, never forget that storytelling is a human art form. It’s about connecting with other people on an emotional level. No matter how advanced our technology becomes, that’s something that a machine will never be able to replicate. The heart of the story will always be human. Let’s build AI that helps us to share that humanity with the world.
Frequently Asked Questions
How long did it take to see results?
Most meaningful business results take 3-6 months to materialize. Anyone promising overnight success is selling something. The companies in my portfolio that grew fastest were the ones that stayed patient and consistent.
What was the biggest challenge in this case?
Almost always, the biggest challenge is people and alignment, not technology or strategy. Getting the right team focused on the right problem is harder than any technical challenge I've encountered.
What would you do differently looking back?
I'd move faster on the things that were working and cut the things that weren't sooner. Most founders, myself included, hold onto failing strategies too long because of sunk cost. Speed of learning is everything.
Can these results be replicated?
The specific numbers will vary, but the underlying patterns and principles are transferable. The key is understanding the context behind the results, not just copying the tactics. Every company has unique constraints that shape what works.