Jacob Effron, Ari Morcos and Rob Toews discuss coding agents that can handle longer assignments and the growing importance of human reviewHuman-in-the-loop AI keeps a person involved in reviewing, correcting, approving, or guiding an AI system's work.. The panel considers whether application scaffolding and workflow design can retain value as underlying models become more capable.
Jacob Effron, Ari Morcos and Rob Toews compare the economics of open weights and closed APIs. Their suggestions about APIs losing importance are predictions about possible business models, rather than a claim that current services have disappeared.
Jacob Effron, Ari Morcos and Rob Toews examine constraints on faster progress, including inference demandAI inference economics examines the revenue, cost, utilization, pricing, and margins involved in serving trained models., semiconductor supply and the time needed for research experiments. Efficiency improvements and smaller models may help, but do not make physical capacityAI compute infrastructure is the hardware, facilities, networks, power, cooling, storage, and software used to train and run AI models. irrelevant.
Jacob Effron, Ari Morcos and Rob Toews also consider recursive improvementRecursive self-improvement is the proposed process in which an AI system helps improve its own capabilities, then uses those improvements to support further advances., robotics and adoptionAI adoption is the process by which people and organizations begin using AI systems as a sustained part of real products, decisions, and workflows. in organizations. The conversation distinguishes technical potential from the slower work of earning trust, integrating tools and changing established practices.
Watch on YouTube
