When Your Own Output Becomes Your Training Data: Noise-to-Meaning Loops and a Formal RSI Trigger
This provides a theoretical foundation for predicting recursive self-improvement triggers in AI systems, which is crucial for AI safety research, though it is incremental as it builds on existing ideas like self-prompting and AutoML.
The paper tackles the problem of understanding when AI systems might enter unbounded self-improvement cycles by presenting a formal model called Noise-to-Meaning Recursive Self-Improvement (N2M-RSI), which shows that once an AI agent feeds its own outputs back as inputs and crosses an information-integration threshold, its internal complexity grows without bound under the model's assumptions.
We present Noise-to-Meaning Recursive Self-Improvement (N2M-RSI), a minimal formal model showing that once an AI agent feeds its own outputs back as inputs and crosses an explicit information-integration threshold, its internal complexity will grow without bound under our assumptions. The framework unifies earlier ideas on self-prompting large language models, Gödelian self-reference, and AutoML, yet remains implementation-agnostic. The model furthermore scales naturally to interacting swarms of agents, hinting at super-linear effects once communication among instances is permitted. For safety reasons, we omit system-specific implementation details and release only a brief, model-agnostic toy prototype in Appendix C.