Anthropic co-founder on quitting OpenAI, AGI predictions, $100M talent wars, 20% unemployment, and the nightmare scenarios keeping him up at night | Ben Mann

In this episode, Benjamin Mann, the co-founder of Anthropic, discusses the rapid advancements in AI technology and the crucial need for safety measures as we approach the potential development of superintelligence. Mann provides a timeline, predicting a 50% chance we might reach some form of superintelligence by 2028. He reflects on his experiences at OpenAI, where concerns around safety led him to prioritize it at Anthropic, stating, 'We felt like safety wasn't the top priority there.' Mann discusses the recruiting challenges in the AI landscape, mentioning how mission-driven teams at Anthropic retain talent despite aggressive offers from competitors like Meta.

The conversation deepens into the anticipated economic impacts of AI, with Mann noting that up to 20% of jobs could be affected, and emphasizing the unpredictable nature of capitalism as AI technology evolves. His focus is on aligning AI with human values, which he believes is critical for preventing potential catastrophes associated with AI. Mann proposes the 'economic Turing test' as a metric for assessing whether AI has become truly transformative. This test assesses if an AI can competitively perform jobs equivalent to that of a human.

The episode further emphasizes the urgency of preparing for an AI-augmented future while maintaining societal stability, with Mann advising on how individuals can stay ahead, underscoring the importance of adaptability and lifelong learning. Through discussions of scaling laws and AI alignment mechanisms, including their approach called 'RLAF' (reinforcement learning from AI feedback), Mann expresses optimism about safely integrating advanced AI into society and encourages listeners to engage with this pressing topic.