Togelius on how the game industry actually uses AI
What happened
Julian Togelius, a professor at New York University and co-founder of modl.ai, appeared on The Information Bottleneck podcast to discuss AI and games with host Ziv Ravid. He said studios use more machine learning in practice than they publicly admit — for example, the goalkeepers in EA FC are reportedly networks trained with reinforcement learning. He also discussed the limits of dialogue trees, and how NPC responses turn strange once an LLM is put into a game. He further argued that for years the industry has been building machines that imitate humans, rather than machines that help people with things they are not good at.
AI-written from the coverage · updated 20 d ago
Coverage timeline
Follow the coverage to see the story from different sides.
- Julian TogeliusJulian Togelius joins The Information Bottleneck podcast to discuss AI and games
Julian Togelius, an NYU professor and co-founder of modl.ai, joined host @ziv_ravid on The Information Bottleneck podcast to discuss AI and games. The episode notes that studios use more machine learning than they publicly admit — for example, EA FC's goalkeepers are reportedly networks trained with RL — and covers the limits of dialogue trees and how NPC responses turn strange once an LLM is put into a game. He also argues that the industry has spent years building machines that imitate humans rather than machines that help people with what they are not good at.