The Dimensionality Gap and the Case Against Von Neumann Minds
Jyotiprakash Mishra posted The Dimensionality Gap: Three Scaling Barriers to AGI on Von Neumann Architectures, and a Path Forward Through Neuromorphic-Photonic Substrates on MDPI Preprints in March 2026 (DOI 10.20944/preprints202603.2196.v1). The title states the argument’s shape, and the shape is what counts for this site. Three barriers, all belonging to the von Neumann architecture, stand between current artificial intelligence and artificial general intelligence, and the way past them runs through substrates that compute differently, neuromorphic and photonic.
| Claim component | Content |
|---|---|
| The diagnosis | Scaling to AGI on von Neumann architectures hits three distinct barriers |
| The gap | Dimensionality, the mismatch between the problem’s structure and the machine’s memory-compute split |
| The path | Neuromorphic and photonic substrates that compute in different physical media |
The architecture argument, and its lineage
Mishra’s preprint belongs to a line this site has tracked from its hardware end. The von Neumann bottleneck was the stated starting point of the memristive neuron analysis, where synaptic storage and access dominated the cost of running a cortical learning rule digitally. The optical computing literature reached the same verdict from the photonics side, examined in the optical neural networks analysis, and the neuromorphic chip record, examined in the Loihi 2 analysis, shows the industry building around the bottleneck in silicon. The preprint’s contribution is to bundle the bottleneck argument into a scaling claim about AGI specifically, and to name the alternative substrates as a path rather than a curiosity.
Two cautions belong next to the argument, and honesty requires them. The work is a preprint rather than a peer-reviewed publication, and this site reads it as a position paper whose three barriers are a research program rather than a demonstrated result. The barriers are named in the title; their formalization lives in the preprint’s body and has not been independently assessed.
Why the substrate argument reaches the uploading debate
The preprint’s structure carries weight for mind uploading because of what it denies. If AGI-scale cognition runs into walls that belong to the von Neumann architecture itself, then computation is not an abstract quantity that any substrate can host in any volume. The physical organization of the computing substrate becomes a first-order variable in what the system can do. That is the hardware-side version of the substrate question this site tracks from the philosophy side, where biological naturalists argue that the medium carries what the program abstracts away. Mishra’s argument and the biological naturalist’s arrive at adjacent positions from opposite motives, one engineering, one metaphysical.
The connection runs through the emulation program directly. A whole brain emulation assumes that the brain’s dynamics can be reinstantiated on different hardware. The dimensionality gap argument says some computations resist reinstantiation on the wrong substrate. If both are right in their domains, the emulation question becomes a matching problem, which substrates can host which dynamics at what fidelity, examined on the brain emulation page.
Comparison to The Consciousness AI
This project treats consciousness as an emergent property of organized physical activity, substrate independent in principle, and a preprint like this one sharpens the price of that principle. Substrate independence cannot mean that organization floats free of physics. It means the right organization, on a substrate capable of hosting it, carries the properties. The project’s research code, maintained in the tlcdv/the_consciousness_ai repository, runs its dynamics on ordinary digital hardware while stating openly that the hardware question is open, and the substrate console page exists to make those dynamics inspectable rather than assumed.
What the Preprint Leaves Open
The preprint names a path and leaves its demonstration open. No neuromorphic-photonic system has shown the AGI-relevant scaling the argument predicts, and the three barriers await formal treatment that survives review. What the argument adds to this site’s record is directional. The hardware literature, the device work and the architecture critiques are converging on the same claim from separate starting points, that the substrate is a variable that determines what computation can become. The field’s state of the argument is reviewed in the flagship overview of AI consciousness research.