DeepMind establishes AGI research institute, chief AGI scientist says Astra is not yet AGI.

Beating AI News Flash: Google DeepMind announces the establishment of the DeepMind Institute, dedicated to studying how AGI will affect science, the economy, and society.

DeepMind co-founder and Chief AGI Scientist Shane Legg will lead the institute's content direction, with CEO Demis Hassabis and Google Senior Vice President James Manyika also involved.

Legg also disagrees with recent claims that "AGI has already arrived." OpenAI President Greg Brockman and Nvidia CEO Jensen Huang have both previously linked GPT-6 Astra to the arrival of AGI. Legg believes that although Astra is highly capable, it still falls short of OpenAI's own definition of AGI—namely, being highly autonomous and surpassing humans in most economically valuable work.

The DeepMind Institute released its first three articles, discussing AI safety, economic policy, and "new utopianism." One of them, written by Rohin Shah and Anca Dragan of DeepMind's safety team, criticizes the decline in information transparency around Astra's release and argues that intense competition will make AI companies increasingly reluctant to disclose model information.

Anthropic also established a similar Anthropic Institute in March this year, integrating its red team, social impact, and economic research teams.