Most of what people mean when they say an AI system is impressive is that it has a great deal of access consciousness: information is available to it for reasoning, planning, report, and the control of action across an enormous range of domains and at extraordinary speed. The [YOU] on AI cycle documents the practical consequences of this access richness in detail—the twenty-fold productivity multiplier, the dissolution of professional specializations, the expansion of who can build what. Block's distinction is a reminder that the access richness, however impressive, settles nothing about the phenomenal question—and that the moral and relational questions about AI, the questions about whether these systems have welfare, whether they can be harmed or benefited, whether forming relationships with them is a relationship with a subject or an elaborate mirror—all turn on the phenomenal question that access leaves open.
The concept is also practically useful for calibrating how we talk about AI capabilities. When an AI system reports that it finds a conversation interesting, or feels uncertain, or is happy to help, it is producing access-conscious outputs: information about its own processing states poised for report. Whether those outputs are accompanied by any phenomenal feel, whether there is anything it is like to be in those processing states, is a question the outputs themselves cannot answer. Block's distinction is the tool that prevents the slide from the impressive access behavior to the unwarranted phenomenal attribution.
Block introduced access consciousness as part of a broader project of identifying the distinct phenomena that had been conflated under the single term “consciousness.” He argued that at least two quite different things were in the vicinity: the phenomenal (experience, feel, what it is like) and the functional (the availability of information for reasoning and control). Many earlier debates in consciousness research had proceeded as if these were one thing, producing systematic confusion: results from cognitive experiments about information availability were presented as bearing on phenomenal consciousness, while intuitions about phenomenal consciousness were used to evaluate functionalist theories of cognitive access.
The access/phenomenal distinction resolved a large class of these confusions by showing that cognitive science and functionalist philosophy could explain access consciousness quite well—there was no hard problem about how information becomes available for reasoning—and that the hard problem was specifically about why any phenomenal feel accompanies the access, a question that functional explanations leave untouched. This made the distinction therapeutically useful: it focused attention on the right question rather than the easier one, and showed that much apparent progress in consciousness science had been progress on access while leaving the phenomenal problem untouched.
Poised for global control. Block's technical definition: a representation is access conscious if it is poised for use in reasoning, for the rational guidance of action, and for verbal report. “Poised” means available in principle, even if not currently being used; “global” means available throughout the cognitive system rather than confined to a local module. The definition is functional through and through: it specifies a causal-role property of the representation, not anything about its felt character.
What large language models have. By Block's functional definition, capable AI systems have access consciousness in abundance. Information from across a large context window is integrated and made available to shape output; the system can reason about its own states when prompted; it can produce sophisticated descriptions of its processing. The question Block insists we hold open is whether this functional richness is accompanied by any phenomenal feel—whether there is something it is like to be the system in these states, or whether the lights are entirely off inside a machine that is, functionally speaking, doing everything a mind does.
The diagnostic value. Access consciousness is the part of consciousness that empirical science and functional analysis can get purchase on. We can study how information becomes globally available in biological and artificial systems, trace the neural and computational mechanisms, and develop accounts of when access fails (neglect, blindsight, unconscious processing). None of this scientific progress closes the phenomenal question; that is Block's point. But it is not nothing: it tells us what these systems demonstrably have and allows the question of what they may additionally have to be stated precisely.