Interview
Simon Benjamin on Architectures for Quantum Computing
- The gap between theoretical requirements and laboratory capabilities is considered fully closed, shifting industry interest from a long-term horizon to an immediate priority for joint research.
- Error correction thresholds have improved from approximately 99.9999% in the 1990s to roughly 99% today, establishing a turning point where the precision gap is no longer physically prohibitive.
- Topological surface codes are expected to enable qubits to operate on a simple grid communicating only with immediate neighbors, removing the need for complex long-range swaps while maintaining scalability as machine size doubles.
- Quantum devices reaching the 50 to 64 qubit threshold for "quantum inimitability" or "quantum supremacy" are anticipated within the coming year, though the first useful algorithms may also emerge in this timeframe using devices without full error correction.
- Embryonic quantum computers are projected to perform specific chemistry and material simulations in the next two to three years, potentially operating without a full error-correction layer.
- Code-breaking applications requiring fault tolerance for long-running algorithms are expected to need thousands, eventually millions, of physical qubits, far exceeding the capacity of first-generation machines.
- First-generation modular machines are predicted to be slower with clock speeds limited by optical link rates due to photon loss, yet they will offer higher fidelity and connectivity compared to 50-qubit grid approaches.
- Version one of the modular architecture is designed for small problems fitting within 50 to 200 qubits where calculation duration is less critical, with a demonstration of linking two modules into a single unit targeted for the next year.
- Once the physics of linking modules is demonstrated, the architecture allows for immediate scaling to millions of qubits, provided the financial cost of replicating the module is affordable, avoiding the control challenges associated with grid-based scaling.
- Building a fully realized modular quantum computer is technically feasible contingent on funding, with the capacity to proceed immediately if approximately one billion dollars is provided.
- The speaker anticipates that a billion-dollar investment could enable an immediate attempt at building the technology, though the final modular system will likely reside in a facility with fiber optics rather than in consumer devices.
- Future solutions involving optical cavities are expected to eliminate light waste and increase link speeds to match other approaches while preserving high connectivity as the technology matures.
- Machine learning is identified as a specific application area being explored, though its benefits on small-scale quantum machines remain unproven.
- Industry diversification is expected to bring in software engineers, systems engineers, and programmers without deep quantum physics backgrounds to support machine building and simulation software development.
- Significant risks include over-investment, inflated expectations similar to "AI winters," and the potential for the field to become toxic or struggle to secure future funding if hype outpaces deliverables.
- There is a risk that companies and startups may pitch optimistic timelines to investors, leading the public to believe that breakthroughs like new drug discoveries will occur within the next one to two years.