Advanced Computing
What counts as computation when the substrate is no longer a silicon logic gate — and by what formal criterion do we decide?
The default model of computation is a digital machine: binary state, clocked transitions, separated memory and arithmetic, instructions executed in sequence. It is an enormously productive model, but it is not the only one, and for large classes of physical, biological, and cognitive systems it is plainly not the model at work. Neural tissue does not clock. A turbulent fluid does not branch. A glycolytic oscillator does not read instructions. Yet each processes information, transforms signal structure, and computes in a sense that has proved worth formalizing. Advanced computing is the study of that broader sense.
The tradition
The tradition runs from cybernetics and analog computing through neural networks, molecular and chemical computing, reservoir computing, optical and photonic systems, neuromorphic hardware, physical reservoir substrates, and the ongoing effort to extract computation from thermodynamic, fluidic, and biological media. The common thread is architectural neutrality: a refusal to treat any one physical implementation as definitional, and an insistence that the properties of computation — composition, memory, signal propagation, stability under perturbation — be defined in terms general enough to cross substrates. What the substrate adds or forbids is then an empirical question, not a definitional one.
Unconventional Computing
ICSAC's contribution to this tradition is to press for formal criteria that can be evaluated across substrates without re-parameterization. The Unconventional Computing program treats integration-differentiation balance as a candidate architecture-neutral criterion for information processing: wherever a system can hold state coherently while remaining differentiable from itself, it meets a basic condition for computation, independent of whether its state is encoded in voltage, concentration, spike rate, or magnetic orientation.
Dimensional Scaling
The Dimensional Scaling program asks a complementary question: what is lost when information is forced to cross a representational boundary, and does the loss admit a law independent of the information's content? These are the kinds of questions that, in ICSAC's view, define the center of the field.
The practical stakes are immediate. A formal criterion for substrate-general computation is a criterion for what machine learning models are actually doing, for whether a biological tissue supports the dynamics attributed to it, and for whether a proposed physical computer is a computer or a suggestive metaphor. The Institute treats advanced computing not as a catalogue of exotic hardware but as the disciplined search for such criteria.