Amazon’s AGI Lab Head David Luan Unveils ‘Gyms’ for Reinforcement-Learning AI Agents
In a wide-ranging Decoder interview, Amazon AGI lab head David Luan—an OpenAI and Google AI veteran—explains why “agents”—AI systems that learn by trial and error in simulated workflows—are the next big step beyond chatbots. Amazon is building “gyms” of digital work environments (from CRM to CAD) for reinforcement-learning agents to master real tasks with high reliability. Luan left his startup Adept for Amazon’s massive compute scale, betting that training true action-driven AI agents, not just language models, will redefine computing and unlock vast new productivity gains.
Also mentioned in:
- TechCrunch — Amazon AGI Lab Head Luan Backs Reverse-Acquihires, Eyes Tens-of-Billions-Dollar Clusters for AGI