23 Daily — According to theverge.com, AI research company Anthropic is leveraging AI automation to accelerate the development of its AI models, achieving approximately twice the progress compared to a year ago with no apparent slowdown. This insight comes from a report published by AI safety research group METR following an 11-week internal investigation into Anthropic's AI research and development processes.
Anthropic Doubles AI Model Development Speed
METR's findings reveal that Anthropic's AI models have made roughly two years’ worth of historical progress within the last year alone, marking a notable acceleration. Researchers emphasized that this improvement is not solely driven by increased computational power but largely attributed to AI itself, with its impact growing over time. Employees reported about a 12-fold increase in productivity when using Anthropic’s cutting-edge internal model, approximately double the boost observed a few months prior. Currently, AI assists human researchers significantly, though signs indicate fully autonomous AI contributions may rise soon.
Rapid Development of Three Unreleased AI Models
The report highlights that Anthropic has completed training three unreleased AI models in the past six months, codenamed Model 2, Model 3, and Model 4, with training completed in July, August–September, and September–October respectively. Additionally, a previously unreleased 2022 model named "Unreleased 2022" was noted. These new models have contributed substantially to the company’s swift progress and productivity gains.
Limited Transparency Raises Safety Concerns
Against this backdrop, Anthropic provided METR with extensive confidential information, including proprietary research data and employee interviews, surpassing initial expectations. However, Anthropic retained the right to redact content in METR’s public report and exercised it extensively, resulting in significant omissions of crucial evidence. This limitation restricts public and industry capacity to fully assess the pace and risks of Anthropic’s AI advancements, raising questions about transparency in frontier AI development.
Editor's take: Anthropic's AI-driven Acceleration Highlights Innovation and Oversight Challenges
Anthropic's AI automation doubling development speed demonstrates the transformative potential of AI in enhancing its own advancement efficiency, potentially reshaping R&D resource allocation and competitive dynamics. Simultaneously, extensive redactions constrain transparency and impede comprehensive safety oversight, intensifying debate over the speed and risks of autonomous AI progression. Entrepreneurs should note the productivity benefits of AI-assisted development while considering mechanisms to balance rapid innovation with robust transparency and risk management. Verification of genuine impact may come from analyzing project completion times, productivity metrics post-AI integration, new product launch intervals, and trends in safety disclosures.