Anthropic is using AI models to accelerate the development of more powerful AI systems, making in one year the kind of progress that would have taken the Claude-maker roughly two years, according to a new report.
The report was published on Thursday, November 5, by METR, an independent AI safety research organisation, following an 11-week investigation into Anthropic’s internal AI research and development work. Human researchers remain in control of the process, with much of the acceleration largely coming from their use of AI models as a research assistant.
However, “there are some signs that fully autonomous contributions will play a larger role soon,” METR said.
These findings have raised questions and concerns about a possibly near future in which AI systems could improve themselves with little or no human involvement, and whether existing safety measures can keep pace.
Technically known as recursive self-improvement (RSI), it means an AI system that can fully autonomously design and develop its own…