Anthropic is leveraging AI to enhance its own development, with its models improving at a rate double that of a year ago, according to a recent report by METR. The findings highlight a significant acceleration in AI research, suggesting the technology could soon improve itself with minimal human intervention. This comes as the company prepares for its public debut, despite its CEO advocating for a slowdown in AI development.
The report, based on an 11-week investigation, indicates that AI is playing a major role in this acceleration, with productivity gains reported by employees using internal models. One internal model reportedly increased productivity by 12 times compared to previous models, though the report warns that fully autonomous contributions may soon play a larger role.
Despite the insights, the report notes that Anthropic has redacted much of the information, limiting public understanding of the pace and risks of AI development. METR expressed concerns over the restrictions, stating that the redactions hindered their ability to fully communicate findings and allow for independent assessment.
The report also highlights that Anthropic has completed training on three new models, which remain unreleased. These models, codenamed Model 2, Model 3, and Model 4, were developed in rapid succession, suggesting the company's research is moving at an unprecedented pace. However, the lack of transparency raises questions about the broader implications of such rapid development.
Overall, the findings underscore the growing influence of AI in its own development, raising critical questions about control, safety, and the future trajectory of the technology.