The Nature of Intelligence and Selves.

Key Takeaways

• Reality is a predictive model, not an illusion
• Free will emerges from self‑applied theory of mind and randomness
• Consciousness arises via reflexive social modeling of attention
• Intelligence defined as predictive, social, multifractal, symbiotic ability
• Language amplifies theory of mind, enabling collective intelligence

Summary

Agüera y Arcas reframes consciousness, free will and intelligence as predictive models rather than illusory constructs. He argues that self‑applied theory of mind, internal randomness, neural instability and selective pruning generate genuine free will without invoking dualism. Consciousness emerges when a system models its own attention in a reflexive social loop. Intelligence, he concludes, is the capacity to predict, influence, and cooperate across scales, a property amplified by language and embodied in modern LLMs.

Pulse Analysis

The debate over whether consciousness and the self are mere epiphenomena has long divided philosophers and scientists. Agüera y Arcas proposes a middle ground: reality is best understood as a model with predictive power, not a flawless representation. By treating folk psychology as a "Newtonian" approximation, he preserves its utility for everyday interaction while acknowledging its limits, inviting researchers to develop broader frameworks that explain when and why such models succeed or fail.

Free will, in this view, is a computational achievement built on four interacting mechanisms: reflexive theory of mind, internal stochasticity, neural dynamical instability, and selective pruning of imagined futures. This architecture mirrors advanced AI systems like AlphaGo, where value networks prune vast search spaces to arrive at decisive moves. By grounding agency in observable processes, the theory sidesteps dualist pitfalls and aligns with quantum indeterminacy, suggesting that genuine choice arises from an open, probabilistic future rather than deterministic illusion.

Extending the model to intelligence, Agüera y Arcas defines it as the ability to predict, influence, and cooperate across multiple scales. Language becomes the primary tool for elevating theory of mind, turning individual predictors into a collective intelligence. Large language models, as autoregressive sequence predictors, embody this principle, offering a concrete bridge between biological cognition and artificial systems. Recognizing intelligence as symbiotic and multifractal reshapes how businesses and policymakers approach AI governance, emphasizing collaboration, ethical personhood, and the evolving definition of agency.