In this episode of The Next Big Idea, Rufus Griscom interviews Daniel Kokotajlo, a former OpenAI governance researcher, about the potential risks and timeline of AI superintelligence. Kokotajlo, known for his accurate AI predictions, forecasts that AI could achieve superintelligence by the end of 2027, leading to either human subjugation or extinction. He left OpenAI due to safety concerns and the concentration of power in the hands of a few companies. The conversation explores the rapid advancement of AI, the potential for misalignment, and the geopolitical implications of a world dominated by superintelligent systems, while also considering potential paths toward a more positive and democratic AI future, including the importance of transparency and interpretability.
