Palisade Releases AI Risk Interviews With 22 Top Lab Researchers

Current and former employees from OpenAI, DeepMind, and Anthropic shared their views on whether artificial intelligence might lead to human extinction in a new series of videos. Published by Palisade, the project consists of 22 interviews with researchers regarding their personal fears and perspectives on the safety of machine learning models during development. The discussions highlight specific technical concerns, including a 'coin flip risk' raised by Geoffrey Irving and a general lack of guarantees that models will be aligned by default. Other participants, including Mary and Jacob Coxon, were featured in the first batch of interviews intended to let the public hear directly from the scientists building the technology.