Cosmos Institute argues RSI needs reassessment

The author argues that deep learning forces a reassessment of recursive self-improvement (RSI) in AI, shifting the unit of analysis from an individual self-modifying mind to an AI development system comprising training pipelines, algorithms, and hardware investments. Rather than a sudden, discontinuous 'intelligence explosion' or 'FOOM,' AI development is characterized by a continuous acceleration driven by mutual feedback between algorithmic progress and compute investments. This distributed and iterative process offers more warning shots and opportunities for human intervention and oversight than traditional models suggested. Open article →