NYU Researcher Explains Why AI Models Still Struggle to Play Video Games

Summary

Julian Togelius, director of NYU's Game Innovation Lab and co-founder of Modl.ai, discusses a recent paper exploring why LLMs and AI models struggle with playing video games despite their rapid improvements in coding. He argues that coding is a "well-behaved game" with clear rules, while video games require general intelligence, adaptability, and real-time decision-making that current AI lacks. The article uses this gap to highlight the broader limitations of AI in 2026, challenging the perception that AI is close to human-level general intelligence.