New METR Report Highlights Alarm Over Anthropic's Automated AI Development

Artificial intelligence research organization METR has published a concerning new report detailing how AI firm Anthropic is utilizing artificial intelligence to build even more powerful AI systems. The comprehensive 11-week investigation reveals that the company’s models are currently improving at a staggering pace, roughly twice as fast as they were just a year ago, with no clear signs of deceleration.

According to the findings, the automation of AI development creates a compounding feedback loop where advanced models assist in writing code, optimizing architectures, and training subsequent generations of technology at superhuman speeds. While this technological acceleration demonstrates remarkable technical achievement, it has sparked considerable alarm among AI safety researchers and ethicists who monitor the societal risks of rapid, autonomous capability scaling.

The report underscores a growing anxiety within the scientific community regarding the ‘intelligence explosion’ hypothesis, wherein machine learning systems become capable of independently improving their own software without meaningful human oversight. As recursive self-improvement accelerates, researchers worry that humanity’s ability to maintain safety controls, align objectives, and understand complex emergent behaviors will quickly be outpaced by the technology itself.

Anthropic’s recent advancements, as highlighted by METR, serve as a stark reminder of the fast-evolving landscape of artificial intelligence development. The findings urge policymakers, developers, and safety advocates to reevaluate current governance frameworks and establish more rigorous guardrails before automated recursive improvement outstrips our capacity to manage its consequences.

Aggregated summary. Original: The Verge