Loading...

Creative AI tools have come a long way in recent years — but most still just follow commands. What if artificial intelligence could be more than just a tool? What if it could be a true creative partner?

In this episode of "Approximately Correct," Amii Fellow and CIFAR AI Chair Matthew Guzdial talks about his work on making creativity more accessible to everyone through AI.

Guzdial focuses on how machine learning (ML) can support humans in creative work, particularly in areas like video game design, where the technical barrier to entry is high. Someone might have a brilliant game idea but lack the skills to build it.

"If one of us wanted to start writing a fiction book, we could do it. That's not true for games," he says.

"There's something that feels wrong to me about that. I want anybody with a cool idea to just be able to start working on the thing.”

To do this, Guzdial says, requires AI to act not just as a tool, but as a collaborator. He calls these mixed-initiative systems—AI that engages in a creative back-and-forth with users.

One example is his MorAI Maker project, which lets people create Mario-style game levels by placing platforms, enemies, and other elements. As they create, the program would analyze the person’s design style and start building it alongside them in a true collaboration.

This collaborative work does ask the question: Can AI truly be creative, or can it only mimic what people are doing? Guzdial’s research offers insights into how and why people create, and in determining their creative intent.

The impacts of Guzdial’s work extend beyond the creative fields. Building collaborative tools has led to a greater understanding of how AI tools can learn from human expertise and how ML can be most effective at supporting human goals. He describes how his work in computational creativity, specifically in machine learning that can make effective use of very scarce data, has been used in areas such as the financial sector and healthcare.

These applications might seem wildly different, but Guzdial sees a common goal: building AI that better supports human ambition. By working directly with artists and creators and creating a dialogue to better inform and build the tools they need, Matthew envisions a future where AI helps smaller teams achieve big creative visions.

Listen to the full live interview with Matthew Guzdial to learn more about how people create and how AI can support that creativity.

Approximately Correct: An AI Podcast from Amii is hosted by Alona Fyshe and Scott Lilwall. It is produced by Lynda Vang, with video production by Chris Onciul. Subscribe to the podcast on Apple Podcasts or Spotify.