Brendan Foody, CEO and co-founder of Mercor, an AI company, discusses how AI models are trained and evaluated, particularly in subjective domains like poetry, where experts grade the AI's output using rubrics. He highlights the AI Productivity Index (APEX), which measures model improvement at economically valuable tasks, noting a 25-30% annual improvement. Foody also addresses the challenge of AI models switching between tasks and integrating multiple tools, predicting significant advancements in this area within six to twelve months. He touches on the potential for AI to transform labor markets by creating new jobs focused on training AI agents and building reinforcement learning environments, and also shares insights from his entrepreneurial journey, including his middle school doughnut business.
Sign in to continue reading, translating and more.
Continue
