AI Model Claude Aids in Development of its Successor

Anthropic's artificial intelligence model, Claude, is actively contributing to the development of its next, more advanced version, the company announced Thursday. This marks a notable step in AI self-improvement, with Claude now leading 26% of Anthropic's research and development tasks.

While Claude is not yet operating autonomously, it can complete substantial portions of tasks from high-level prompts under human supervision. Approximately 90% of the company's research and development efforts are conducted in collaboration with Claude, enabling it to perform significant work with close human direction.

The announcement coincides with growing calls from AI leaders, including Anthropic CEO Dario Amodei, for a slowdown in AI development due to safety concerns. The company emphasized the importance of transparency, stating, "we should do everything possible to minimize the gap between what frontier labs know and what the public knows." This includes better measurement and public reporting of AI development to allow society to make informed decisions.

Anthropic highlighted that AI models accelerating their own development could pose challenges for human understanding and control. The company advocates for sharing development metrics to foster a better understanding of how close leading AI labs are to achieving recursive self-improvement—the ability of an AI to autonomously create its successor.

The pace of Claude's involvement in research and development has rapidly increased. In February, Claude led none of the research and development work. By August, it had taken the lead on 26% of these tasks. The company also shared that as of August, around 30,000 agents were engaged in research and engineering, with oversight measures in place to detect and address any misbehavior. Anthropic has also committed to embedding external third-party evaluators to monitor its safety efforts.

This development follows recent discussions on AI safety, intensified by the resignation of an Anthropic researcher who issued a stark warning about the technology's potential threats. While figures like Amodei, OpenAI CEO Sam Altman, and Elon Musk have supported slowing AI development, others, including some tech leaders and former President Donald Trump, have expressed differing views.