DeepMind co-founder urges AI progress stay ahead of safety measures

Shane Legg, a co-founder of Google DeepMind, unveiled the DeepMind Institute to foster discussion on the deployment of artificial general intelligence. He emphasized that the accelerating capabilities of AI—such as autonomous software creation and task execution—risk outstripping current safety mechanisms. Legg dismissed recent assertions that AGI has already been achieved, labeling them premature, yet he maintains a 50% probability of attaining a minimal form of AGI by 2028.

He highlighted Anthropic CEO Dario Amodei’s suggestion to slow, rather than halt, the rollout of cutting-edge models as a worthwhile consideration. The institute will explore topics ranging from science and policy to philosophy and human flourishing, beginning with essays on AI safety, economic policy, and a vision of “new utopianism.”