arXiv 2504.15125

Contemplative Artificial Intelligence

By Ruben Laukkonen, Fionn Inglis, et al.

Published 2025-04-21

Wiki summary

Explore the paper's summary, context, and related research on Papiers.

As artificial intelligence (AI) improves, traditional alignment strategies may falter in the face of unpredictable self-improvement, hidden subgoals, and the sheer complexity of intelligent systems. Inspired by contemplative wisdom traditions, we show how four axiomatic principles can instil a resilient Wise World Model in AI systems. First, mindfulness enables self-monitoring and recalibration of emergent subgoals.…

View the original paper on arXiv