The podcast examines the trend of Meta employees competing for token usage, questioning its effectiveness and potential cost implications for Anthropic. It explores whether "token maxing" is a misguided metric, referencing Goodhart's Law and contrasting it with Meta's AI-driven successes in ad targeting. The conversation analyzes the accuracy of claims regarding Meta's spending on Anthropic tokens, providing alternative calculations based on OpenRouter data and different usage scenarios. Additionally, the podcast covers Intel's partnership with Elon Musk's Terafab project, discussing its significance for securing American chip manufacturing and addressing potential chip bottlenecks. It also touches on cybersecurity concerns, AI safety, and the challenges of model commoditization.
Sign in to continue reading, translating and more.
Continue