Explained: Anthropic is rapidly accelerating AI development amid slowdown calls, outside research group says

Anthropic is developing its AI systems around twice as quickly as it was a year ago, according to an independent analysis released Thursday.

METR

Model Evaluation and Threat Research, a Berkeley, California-based nonprofit that evaluates AI model capabilities for companies like Anthropic and OpenAI.

recursive self-improvement

AI's growing ability to build the next generation of AI, as defined by Dario Amodei.

compute

The amount of computing power used in training AI models, traditionally linked to how well a model performs.

well-specified tasks

Tasks that can be graded automatically or with a language model, as opposed to 'messy' tasks that are harder to define.

bottlenecks

Restrictions, such as limits on computing power or data, that could hold back AI progress; METR found none for Anthropic.