Google DeepMind Co-Founder Warns on AI Safety, Launches AGI Research Institute

Google DeepMind co-founder Shane Legg warned that the development of artificial intelligence capabilities must not outrun safety controls, coinciding with his launch of an institute to explore the deployment of artificial general intelligence (AGI), the Financial Times reported on Wednesday.

The comments address growing concerns within the technology sector about recursive self-improvement, where increasingly capable AI agents can write software and assist in their own development, potentially advancing faster than human-led safeguards can monitor or control them.

Balancing Progress and Precaution

According to the FT report, Legg described a call from Anthropic CEO Dario Amodei to slow, but not pause, the release of new frontier models as "interesting directionally" and "worth considering." He also stated it was premature to declare that AGI had been achieved, despite recent claims from executives at companies like Nvidia and OpenAI.

Legg reiterated his long-standing forecast on the arrival of advanced AI, stating he remains "comfortable" with his prediction of a 50% chance of achieving "minimal" AGI by 2028.

Ad

The DeepMind Institute

To foster a more informed public dialogue, Legg has launched the DeepMind Institute. The organization aims to broaden the discussion of AGI, which Google DeepMind defines as a system possessing the cognitive capabilities of the human brain.

The institute's initial focus will cover a wide range of topics and includes its first published essays:

• Key Areas: Science, education, society, policy, philosophy, and human flourishing.
• First Essays: The institute's first three publications address AI safety, economic policy, and principles for a "new utopianism."