Scaling and the Road to Human-Level AI | Anthropic Co-founder Jared Kaplan

Jared Kaplan discusses scaling laws in AI, focusing on pre-training and reinforcement learning phases, and how they drive AI progress. He touches on the increasing flexibility and capability of AI models, emphasizing the importance of organizational knowledge, memory, and AI oversight for achieving human-level AI. Kaplan also advises building ambitious projects that push the boundaries of current AI capabilities, integrating AI for AI development, and identifying sectors ripe for rapid AI adoption beyond software engineering. The conversation transitions into a Q&A session, where Kaplan answers questions about Claude 4's features, the role of human-AI collaboration, greenfield opportunities for AI builders, and the influence of his physics background on his AI research, including the application of scaling laws and the pursuit of precision in understanding AI trends.