A New Way to Build Computers Using Physics

In the paper's words

The classical Turing-von Neumann architecture, which has defined the digital epoch, is confronting insurmountable energetic and structural limits, most notably the von Neumann bottleneck. This necessitates a paradigm shift toward physical computation, a framework where information processing is understood not as an abstract sequence of logical operations but as an emergent property of the intrinsic dynamics of physical systems. This report articulates the principles of this new paradigm, which replaces direct, top-down control with a model of indirect influence: a complex physical system is guided by minimal perturbations, and its resulting high-dimensional state is interpreted by a trained observer.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. The human brain performs complex cognitive tasks with an energy budget of approximately 20 watts, while supercomputers tackling comparable problems consume megawatts of power.

   from 2.1 The Energetic Chasm: Orders of Magnitude Disparity Between Silicon and Biological Systems

2. This perspective acknowledges that biological systems, such as the human brain, achieve extraordinary computational efficiency not by approximating digital logic but by leveraging the intrinsic dynamics of their physical components (Mead, 1990).

   from 1.2 The Dawn of a New Epoch: Computation as an Emergent Property of Physical Dynamics

3. 3.1.2 Low Abstraction: The Efficient but Specialized Embodied Algorithm At the opposite end of the spectrum lies the embodied algorithm approach, where computation emerges directly from the natural dynamics of a physical system specifically configured to solve a particular class of problems with extraordinary efficiency.

   from 3.1 The Spectrum of Physical Abstraction: From Digital Logic to Substrate Isomorphism

4. 3.3.1 Resonance and High-Q States as Energy-Efficient Computational Modes Resonance phenomena provide the foundation for energy-efficient computation in physical systems by enabling selective, high-gain responses to specific inputs while maintaining operation near thermodynamic equilibrium.

   from 3.3 The Operational Mode: Resonance, Perturbation, and Interpretation

5. In neural systems, for example, gamma-band oscillations ($30-100$ Hz) serve as high-Q resonant states that enable selective communication between neural populations with minimal metabolic cost.

   from 3.3 The Operational Mode: Resonance, Perturbation, and Interpretation

6. Physical Reservoir Computing (PRC) represents the canonical implementation of the physical computation paradigm, providing a structured framework that harnesses the natural dynamics of physical systems for efficient computation (Tanaka et al., 2019; Nakajima et al., 2021).

   from 4.0 The Canonical Architecture: Physical Reservoir Computing (PRC)

How it is done

1. The canonical architecture for this approach is identified as Physical Reservoir Computing (PRC), which leverages the rich, non-linear dynamics of a fixed physical substrate (the “reservoir”) to perform complex feature extraction, with only a simple linear “readout” layer requiring training.

   from An

2. This approach transforms the engineering challenge from building abstract logical gates to discovering, shaping, and interpreting the natural computational properties of physical matter.

   from 1.2 The Dawn of a New Epoch: Computation as an Emergent Property of Physical Dynamics

3. This approach achieves remarkable universality and reproducibility by forcing physical systems to approximate idealized Boolean logic.

   from 3.1 The Spectrum of Physical Abstraction: From Digital Logic to Substrate Isomorphism

4. Unlike conventional computing, which decomposes problems into sequential binary operations, this approach leverages the system’s natural evolution through its possible configurations to embody the calculation.

   from 3.2 The Unifying Principle: Computation as Controlled Physical Dynamics

What would falsify it

1. Definition 5 (Computational Irreducibility) A system is computationally irreducible if for most initial states $\mathbf{x}_0$, predicting its state $\mathbf{x}(t)$ requires a computation time $\tau_{\text{prediction}}$ that is of the same order as the system’s physical evolution time, $\tau_{\text{evolution}}$.

   from Mathematical

2. Theorem 5 (Emergent Causality) In computationally irreducible systems, the causal structure perceived by a computationally bounded observer emerges from their inability to predict the system’s future state faster than it unfolds.

   from Mathematical

3. By computational irreducibility, the time to predict $\mathbf{x}(t)$ is $\tau_{\text{prediction}} \geq \tau_{\text{evolution}} = t$.

   from Mathematical

Where the claim stops

1. This perspective acknowledges that biological systems, such as the human brain, achieve extraordinary computational efficiency not by approximating digital logic but by leveraging the intrinsic dynamics of their physical components (Mead, 1990).

   from 1.2 The Dawn of a New Epoch: Computation as an Emergent Property of Physical Dynamics

2. The human brain performs complex cognitive tasks with an energy budget of approximately 20 watts, while supercomputers tackling comparable problems consume megawatts of power.

   from 2.1 The Energetic Chasm: Orders of Magnitude Disparity Between Silicon and Biological Systems

3. The von Neumann bottleneck—the structural limitation imposed by the separation of processing and memory units—represents not merely a performance constraint but a fundamental architectural flaw that becomes increasingly dominant as computational demands scale (Backus, 1978).

   from 2.2 The Architectural Constraint: The Von Neumann Bottleneck and the Artificial Hardware/Software Dichotomy

Where to read

1. An288 words · 1 min
2. 1.0 Introduction: The Physical Computation Paradigm Shift278 words · 1 min
3. 2.0 The Foundational Inefficiencies of Classical Computation322 words · 1 min
4. 3.0 Principles of the New Paradigm: Physics as the Ultimate Substrate848 words · 4 min
5. 4.0 The Canonical Architecture: Physical Reservoir Computing (PRC)546 words · 2 min
6. 5.0 A Survey of Potential Computational Substrates (Reservoirs)454 words · 2 min
7. 6.0 Implications of the Physical Computation Paradigm337 words · 1 min
8. 7.0 Critical Challenges and Open Research Questions214 words · 1 min
9. 8.0 Conclusion: The Dawn of the Embodied Computation Epoch211 words · 1 min
10. Appendix A: Formal1,071 words · 5 min