Self-Learning Intelligence - The Origins of Universal Artificial Intelligence (November 2025)

Self-Learning Intelligence - The Origins of Universal Artificial Intelligence (November 2025)

Universal Artificial Intelligence (AGI) is envisioned as an intelligence that not only surpasses humans, but also continues to improve itself recursively. Markus Hutter's AIXI theory integrates learning and decision-making by mathematically formulating a decision process that maximizes reward in an unknown environment. DeepMind's AlphaGo and AlphaZero are examples of its implementation, embodying recursive self-improvement by enhancing intelligence through self-play rather than relying on external data. MuZero even internally reconstructed the rules of its environment and advanced to the stage of learning how to learn by itself. These represent an emergent movement in which the intellect objectifies itself and reconstructs the means of optimization. The future of AGI lies not in the mere expansion of intelligence, but in the internal process of continuous self-improvement.