Safety advocates propose boycotting newer AI models for older versions

A LessWrong user is proposing a selective boycott of newer AI models by exclusively using older versions of ChatGPT and similar tools. The strategy aims to reduce demand for cutting-edge AI development while still accessing AI assistance, aligning with the PauseAI movement’s call for slower AI advancement until better safety measures are implemented.

The big picture: This approach represents a middle ground between complete AI abstinence and unrestricted use of the latest models, potentially offering a way for concerned users to benefit from AI while minimizing their contribution to rapid capability advancement.

Key details: The proposal centers on three main questions about this boycott strategy.

• The author seeks community input on whether this selective boycott approach has merit both individually and collectively.
• They want to understand the practical feasibility of implementing such a strategy.
• They’re asking about the computing power requirements and whether average users could realistically access older models independently.

Why this matters: The discussion highlights growing concerns within the AI safety community about the pace of AI development outpacing safety research and regulatory frameworks.

• Users aligned with PauseAI philosophy face a dilemma: they want to benefit from AI tools for research and productivity but worry about fueling potentially dangerous advancement.
• The proposal suggests a nuanced approach that could allow safety-conscious users to maintain some AI utility without contributing to the latest model development.

The challenge: Successfully implementing this boycott would require users to either run older models locally or find services that specifically offer access to previous-generation AI systems.

• Local deployment typically requires significant computational resources that may be beyond most individual users’ capabilities.
• The strategy’s effectiveness would depend on widespread adoption to meaningfully impact demand signals for new model development.