The Model Evaluation and Threat Research (METR) organization has appointed Ryan Greenblatt to lead efforts to uncover verified data on AI development. Greenblatt will focus on investigations into AI capabilities, alignment, and control to address the lack of transparency regarding catastrophic risks. He warned that recursive self-improvement could accelerate capabilities to superhuman general levels within 6 months or a year, which would create a correspondingly large risk of worst-case outcomes.
Greenblatt, who previously worked at Redwood, stated that he once doubted the utility of public information but now views its acquisition as urgent. Chris Painter of METR described the independent publication of internal company alignment data as crucial for navigating a potential period of rapid capabilities progress.
Key sources
- SOURCE@ryangreenblatt“imminent recursive self-improvement could massively accelerate capabilities progress, which could then potentially yield extremely superhuman general capabilities within 6 months or a year”x.com
- SUPPORT@chrispainteryup“public information about the state of alignment inside of AI companies, investigated and published by independent parties, is crucial if intense recursive self-improvement begins”x.com
- SOURCEmarketbrief.now