The Tao of Agency: Autotelic AI, Embedded Agency and Dissolution of the Self

The article introduces the concept of autotelic AI, where agents autonomously generate their own goals rather than relying on externally defined objectives. It discusses key components such as intrinsic motivation, resource-driven priors, and causal-interventional learning, emphasizing the importance of embeddedness in defining agency. The authors propose a unified framework that includes a quantum formulation and a concrete instantiation using LLMs, highlighting the implications for understanding self-identity in AI systems, which is crucial for practitioners developing more autonomous and adaptive AI agents.