What synthetic consciousness means

Synthetic consciousness is the deliberate construction of a system that has experience — as an objective, not as a by-product. It differs from Part 1, which asks how we would recognise experience wherever it occurred, and from Part 2, which aims to copy an existing mind rather than originate one. The distinguishing feature, and the reason this part is short, is that it is the only topic in the Institute's map with no specification to build against. Everywhere else — a fusion reactor, a gene drive, a brain-computer interface — the target can be stated in terms someone could verify. Here it cannot.

Where the science stands

Established There is no engineering programme to build a conscious machine, and the reason is not lack of ambition or funding. A specification requires an acceptance test, and no accepted test for consciousness exists in any system, biological or artificial. Without one, “build a conscious system” is not yet a well-formed engineering brief.

Frontier What could be built instead is a system satisfying the consciousness indicators derived from current theories — recurrent processing, a global workspace, higher-order representation, agency and embodiment. This is technically approachable and epistemically odd: it optimises for the test rather than the property, and would produce a system engineered to look conscious by criteria whose validity is the open question. The indicator literature is explicit that its measures are diagnostic aids, not construction targets.

Frontier Theory choice determines feasibility completely, which is unusual. On global-workspace-style accounts, a suitably organised artificial system could be conscious. On integrated information theory, a functional duplicate on conventional hardware would not be, because what matters is intrinsic cause–effect structure rather than input–output behaviour — and building for consciousness would mean building unconventional hardware. On Orch OR, a minority position, digital consciousness is ruled out altogether. These are not shades of the same answer; they are different projects.

Speculative Whether building is easier than detecting is worth stating as an open question rather than assuming. The intuition that constructing something proves you understand it does not obviously survive here: a built system would face precisely the same evidence problem as a found one, since knowing how it was assembled does not establish that assembly produced experience.

Handwave Claims that synthetic consciousness has already been achieved, or that it will follow automatically from scale, are unsupported in both directions — the first by any evidence, the second by any mechanism.

The binding bottleneck

Frontier The constraint is the absence of an acceptance criterion, and it is upstream of everything else. Engineering without a test is not engineering. Beneath it sit theory disagreement, which makes even the target architecture contested, and the verification problem, which would remain after construction. Speculative Beneath those sits the explanatory gap: if experience cannot be derived from structure and function, then no specification written in structural and functional terms can ever be complete, and the project is not merely hard but ill-posed.

Ethics and governance

Frontier This is the one part of the programme where ethics precede capability rather than trailing it. Deliberately creating an entity that might suffer is a choice made at design time, under uncertainty that cannot be resolved before the choice is made. A 2026 argument in Philosophical Studies pushes further, holding that standard safety practices — training a system to accept constraint and shutdown — may be structurally incompatible with leading theories of wellbeing, which would mean the safe system and the ethically-created system pull apart.

Frontier A takeaway from the first dedicated conference on AI consciousness and welfare was blunter and easier to act on: systems should not be created if they would need to be shut down. Whether that is right, it is at least a decision rule someone could follow. Established And the reputational hazard is immediate — a field that claims success here without a test invites the reception that quantum consciousness received, and would deserve it.

Timelines

  • 10 yr: Frontier systems built to satisfy specific indicators, presented honestly as indicator-satisfying rather than conscious.
  • 25 yr: Speculative an acceptance criterion becomes conceivable if consciousness theories converge.
  • 50 yr: Speculative deliberate construction against a defensible specification, on optimistic assumptions about the underlying science.
  • 100 / 250+ yr: Handwave any date offered for “engineered experience” is a guess wearing a number, and should be read as one.