Google Launches DeepMind Institute on AGI Safety, Expands AI Agent CC for Family Coordination
aiinsider ·
Google and Google DeepMind have officially launched the DeepMind Institute to foster broader discussions on artificial general intelligence safety. The institute is directed by key figures including Shane Legg, James Manyika, and Demis Hassabis, aiming to surface diverse perspectives within the research community. Its inaugural essays address critical topics such as economic policy for AGI disruption, the necessity of transparent model reasoning, and evaluation frameworks for frontier AI. Notably, researchers Rohin Shah and Anca Dragan argued against inevitable opacity in AI systems, while Hassabis proposed a US-led standards body for independent model evaluations. In a separate development, Google expanded its CC AI agent into a family coordination tool designed to manage household tasks like meal planning and school schedules. This new feature supports up to six family members with controlled information sharing and runs on an isolated cloud system powered by Gemini. Currently, the service is limited to US-based Gmail users aged 18 and older, excluding teens and school accounts.
aiinsider