Claude AI Assists Anthropic in Developing Its Successor Model

Anthropic's artificial intelligence model, Claude, is actively participating in the development of its next-generation successor, according to an announcement made by the AI lab on Thursday. Claude is currently leading 26% of Anthropic's model research and development efforts. This significant involvement means that Claude can execute the majority of a given task "end-to-end from a high-level prompt," albeit still under direct human supervision, as the model has not yet achieved complete autonomy. Approximately 90% of Anthropic's research and development activities are conducted in "collaboration" with Claude, enabling the model to handle "large chunks of work under close human direction."

This development comes at a time when prominent figures in the AI field, including Anthropic CEO Dario Amodei, are advocating for a deliberate slowdown in the pace of AI advancement due to safety concerns. As leaders contemplate the trajectory of AI development, Anthropic emphasized the importance of minimizing the disparity between the knowledge held by frontier AI labs and that available to the public. In a blog post, the company stated, "This means better measuring the development of AI, reporting on it publicly, and giving society an opportunity to decide how to use this information." The company further elaborated that AI models accelerating their own development could present challenges for human comprehension and control over these sophisticated systems.

Anthropic proposed that sharing metrics related to AI development could foster a better understanding of how close leading AI laboratories are to achieving recursive self-improvement. Recursive self-improvement refers to a model's capacity to autonomously design and build its subsequent, more advanced version. The company also issued a call to action for other AI developers, urging them to regularly share comparable metrics. Anthropic encouraged the adoption of a public methodology to ensure that these figures can be consistently compared over time and potentially across different research institutions. While Anthropic's disclosure did not specify how imminent the company believes recursive self-improvement to be, the increasing contribution of Claude to research and development is a notable indicator of its evolving capabilities and role in its own advancement.