Podcast: AI and Video Games (with Julian Togelius)

The episode explores the intersection of artificial intelligence and video game design through a conversation between podcast host Eric Seufert and NYU professor Julian Togelius. The discussion centers on how games provide ideal sandbox environments for AI research, offering rich learning signals and complex reward structures that mirror real‑world challenges such as financial markets. Togelius highlights the limitations of traditional game AI benchmarks—chess, Go, and early platformers—and argues for broader, more open‑ended research that includes content generation, player modeling, and general game playing. The dialogue also addresses the growing use of generative tools in production pipelines, noting industry hesitation due to consumer backlash and concerns over authenticity. Togelius suggests that AI can accelerate development cycles, enable procedurally generated worlds, and create personalized player experiences, while also raising questions about human oversight and the evolving role of designers. The conversation touches on world models versus language models, with examples like Google’s Project Genie and OpenAI’s Sora, underscoring their potential for rapid prototyping and interactive content creation. Overall, the episode presents a nuanced view of AI’s transformative potential in gaming, balanced by practical considerations around consumer perception, workflow integration, and the future of game design.