Researcher Nathan Lambert describes the gap he sees between closed frontier work and public literature, especially in complex agent training. He proposes shared research infrastructureAI compute infrastructure is the hardware, facilities, networks, power, cooling, storage, and software used to train and run AI models. including checkpoints, training recipes and environmentsAn AI agent harness is the software framework that packages a model with tools, instructions, context management, execution controls, and user interaction. that others can study.
The interview explores funding and compute access, institutional collaboration and differing perspectives on model openness and safetyGovernance sets responsibilities, rules and oversight for decisions about how AI systems are used.. Lambert’s views about research incentives and international approaches are attributed.
Watch on YouTube
