Google and Google DeepMind researchers have created the DeepMind Institute to widen the conversation around artificial general intelligence. DeepMind co-founder Shane Legg, Google executive James Manyika, and Google DeepMind chair Demis Hassabis serve as directors, with Legg as managing editor.

The institute wants to bring out differing perspectives among Google, DeepMind, and the wider research community on AGI. "They will not always agree, and they will likely change their minds, as more data and information comes to light at the fast-moving frontier," the announcement said.

Its first collection of four essays covers economic policies for potential AGI disruption, preserving human-readable model reasoning, principles for human flourishing, and a framework for evaluating frontier AI models.

In one essay, DeepMind safety researchers Rohin Shah and Anca Dragan argue that AI's shrinking transparency window is not inevitable. They suggest limiting "opaque serial depth" or requiring developers to show less transparent systems remain monitorable.

Hassabis proposes a U.S.-led frontier AI standards body. Developers would initially submit models voluntarily up to 30 days before release, with passing tests potentially becoming mandatory once the system proves effective. The body would eventually develop independent, undisclosed "held-out" tests, and the framework could be "ratcheted up if the seriousness of the situation demands," including a coordinated slowdown.

The essays come as safety debate shifts toward concrete proposals for disclosure and outside scrutiny, accelerated this week as industry leaders endorsed elements of Anthropic CEO Dario Amodei's call to "pace" frontier AI development.

Source: TechCrunch