The Stream of Computation: Temporal Continuity as a Missing Ingredient for Artificial Consciousness

Source raw items (1)

• Semantic Scholar2026-08-01 07:15:26
  The Stream of Computation: Temporal Continuity as a Missing Ingredient for Artificial Consciousness

  Recent advances in large language models (LLMs) have reignited questions about whether artificial systems can possess consciousness. Despite impressive gains in reasoning and language understanding, current AI systems still operate within isolated episodes of computation. We argue that a missing ingredient is temporal continuity: persistent internal dynamics that sustain an unbroken stream of computation analogous to the stream of consciousness. We propose a roadmap for a stream-of-computation architecture based on persistent recursive inference, in which the output of each cognitive cycle becomes the input to the next, allowing internal states to evolve autonomously over time. Unlike standard chain-of-thought methods, which are recursive only on demand after a prompt, this framework aims at autonomy and continual learning. It includes mechanisms for continual learning, dynamic switching between outward- and inward-attentive cognition, and sleep-like phases that separate learning from inference. Together, these mechanisms form the foundation for a lifelong agent capable of maintaining temporal continuity, integrating new experiences, and reflecting on its own internal state. Functionally, such an architecture promises deeper reasoning, adaptability, and metacognitive stability. While subjective experience in AI remains an open question, temporally continuous agents may mark a step toward artificial systems whose individuality and identity arise from the continuity of their own computational existence.