David Luan has been at the center of the modern AI revolution: he was the ~30th hire at OpenAI, he led Google's LLM efforts and co-led Google Brain, and then started Adept in 2022, one of the leading companies in the AI agents space. In today's episode, we asked David for some war stories from his time in early OpenAI (including working with Alec Radford ahead of the GPT-2 demo with Sam Altman, that resulted in Microsoft’s initial $1b investment), and how Adept is building agents that can “do anything a human does on a computer" — his definition of useful AGI.

Full show notes: https://www.latent.space/p/adept

00:00:00 Introduction of David Luan, CEO and co-founder of Adept
00:01:14 David's background and career trajectory
00:03:20 Transition from reinforcement learning to transformers in the AI industry
00:05:35 History and development of GPT models at OpenAI and Google
00:13:08 Adept's $420 million funding rounds
00:13:38 Explanation of what Adept does and their vision for AI agents
00:19:20 Reasons for Adept becoming more public-facing
00:21:00 Adept's critical path and research directions (Persimmon, Fuyu, Act One)
00:26:23 How AI agents should interact with software and impact product development
00:30:37 Analogies between AI agents and self-driving car development
00:32:42 Balancing reliability, cost, speed and generality in AI agents
00:35:11 Adept's unique positioning and advantages in the AI industry
00:37:30 Potential of foundation models for robotics
00:39:22 Core research questions and reasons to work at Adept
00:40:57 David's closing thoughts on the AI agent space and industrialization of AI