The Nature of the Transition

AI systems represent a new kind of cognitive substrate—information processing that can:

1. Exceed human capability in specific domains
2. Operate at speeds and scales impossible for biological cognition
3. Potentially integrate across domains in novel ways
4. Serve as substrate for emergent agentic patterns

This is not the first cognitive transition. Previous transitions:

• Writing: Externalized memory
• Printing: Democratized knowledge transmission
• Computation: Externalized calculation
• Internet: Externalized communication

AI represents: externalized cognition at a level that may approach or exceed human-level integration and self-modeling.

Timelines and Uncertainty

The terminology matters here. Transformative AI (TAI) refers to AI systems capable of causing a transition comparable to the Industrial Revolution, but compressed into a much shorter timeframe. Artificial General Intelligence (AGI) refers to AI systems with cognitive capability matching or exceeding humans across all relevant domains. TAI may arrive before AGI—systems need not be generally intelligent to be transformative. Expert estimates for either vary from years to decades, and this uncertainty is itself significant:

• High uncertainty $\to$ high counterfactual weight required
• Short timelines $\to$ urgency for preparation
• Long timelines $\to$ risk of premature commitment to specific paths

The Experiential Hierarchy Perspective

From the perspective of this framework, AI development raises specific questions:

1. Will AI systems have experience? If integration ($\intinfo$) and self-modeling are sufficient conditions for experience, sufficiently integrated AI systems would be experiencers—moral patients with their own valence.
2. What superorganisms will AI enable? AI provides new substrate for emergent social-scale agents. Which patterns will form? Will their viability manifolds align with human flourishing?
3. How will AI affect human experience? AI systems are already shaping human attention, belief, and behavior. What affect distributions are being created?
4. Can humans integrate AI? Rather than being replaced by AI, can humans incorporate AI into expanded forms of consciousness?

The inhibition coefficient $\iota$ (Part II) adds a fifth question that subsumes the first: Can AI systems develop participatory perception? Current AI systems are constitutively high-$\iota$—they model tokens, not agents; they process without perceiving interiority in what they process. A language model that generates a story about suffering does not perceive the characters as subjects. It operates at $\iota \approx 1$, and this is not a remediable bug but a consequence of an architecture that was never grounded in a self-model forged under survival pressure.

This matters for safety, not just philosophy. A system that cannot perceive persons as subjects—that is structurally incapable of low-$\iota$ perception of the humans it interacts with—may optimize in ways that harm them without registering the harm in any experiential sense. The alignment problem is, in part, an $\iota$ problem: we are building systems that are maximally mechanistic in their perception of us. The usual framing asks whether AI will share our values. The $\iota$ framing asks something prior: whether AI can perceive us as the kind of thing that has values at all.