DeepMind Launches Institute to Broaden AGI Debate

Google DeepMind has launched the DeepMind Institute to foster open debate around artificial general intelligence, led by co-founder Shane Legg, Demis Hassabis, and James Manyika. The institute's inaugural collection of four essays addresses economic disruption, AI transparency, human flourishing, and model evaluation frameworks. Notably, Demis Hassabis proposes a US-led frontier AI standards body that could require models to pass evaluations before deployment, with potential for coordinated developer slowdowns if safety falls behind. Separately, DeepMind safety researchers argue that shrinking AI transparency is not inevitable and urge developers and regulators to directly confront the trade-offs.