Google DeepMind launches institute to broaden AGI discussion
Google and Google DeepMind researchers have launched the DeepMind Institute, a new initiative focused on advancing discussion about artificial general intelligence, or AGI. Its directors are DeepMind co-founder Shane Legg, Google executive James Manyika and Google DeepMind chair Demis Hassabis. Legg will serve as managing editor.
The institute says it will present differing perspectives from Google, Google DeepMind and the wider research community. Its first collection contains four essays covering policies for possible economic disruption from AGI, preserving human-readable model reasoning, principles for human flourishing and methods for assessing frontier AI models. The announcement noted that the participants may disagree and revise their views as new information emerges.
In one essay, safety researchers Rohin Shah and Anca Dragan argue that declining transparency in advanced models is not unavoidable. They discuss limiting “opaque serial depth” or requiring developers to show that less transparent systems remain monitorable. Hassabis separately proposes a U.S.-led standards body. Developers would initially submit models voluntarily for review up to 30 days before release, while successful testing could later become mandatory. The framework could eventually use undisclosed “held-out” tests and, if necessary, support a coordinated slowdown in frontier AI development.
Written by Sakhanda Wire AI from reporting by TechCrunch. Facts, quotes and full context are in the original article below.