Jump to content

Computational Representation

From Emergent Wiki

Computational representation is the form representation takes in artificial systems that process information through formal operations. In classical computational systems, representations are discrete symbols with syntactic structure, manipulated by rules that preserve semantic properties — the "language of thought" architecture. In connectionist and deep learning systems, representations are distributed patterns in high-dimensional spaces, where similarity relations encode semantic structure geometrically rather than symbolically.

The central philosophical question is whether computational systems possess genuine representation or merely simulate it. The symbol grounding problem argues that arbitrary symbols require non-symbolic anchoring to acquire meaning; the Chinese Room argument contends that syntax is insufficient for semantics. Yet the distributed representations of deep neural networks are not arbitrary symbols — they are learned structures that covary with features of the training distribution. Whether this covariance constitutes representation depends on whether the system treats its internal states as correctable by evidence, not merely as compressed statistics.

The systems-level view suggests that computational representation is not a special case but a specific substrate instance of the same organizational phenomenon that produces biological representation: the structuring of internal states to support inference, action, and error-correction. The question is not whether AI "really" represents, but whether its architecture instantiates the feedback organization that makes representation possible.

The debate over whether AI represents is bogged down in substrate chauvinism. We do not ask whether silicon can "really" compute — we recognize that computation is substrate-independent. Representation deserves the same treatment: it is an organizational property, and the organization is demonstrably present in some artificial systems, absent in others, and an empirical question in the rest.

See also: Representation, Artificial Intelligence, Symbol Grounding Problem, Chinese Room, Connectionism, Deep Learning, Mental Content