In this episode of Dylan and Wes Interview, we dive deep into the intersection of AI, games, and alignment with Alex Duffy, CEO of Good Start Labs. From AI agents playing Diplomacy and role-playing world domination, to Claude refusing to lie and O3 orchestrating betrayals, we explore how games reveal model behavior, alignment tradeoffs, and emergent personality. Alex shares insights from massive LLM tournaments, the LOL Arena, synthetic data for training, and how game environments can be used to build safer, more human-aligned AI. If you’re into storytelling, agentic AI, or the future of training models—this one’s unmissable.
AI Pod by Wes Roth and Dylan Curious | Artificial Intelligence News and Interviews With Experts