THE QUANTUM INTELLIGENCER

Classical Emulator Outperforms Quantum Devices in Large-Scale Boson Sampling In Plain English: Scientists have built a computer program that can do a special kind of calculation — one that was thought to require a quantum computer — using regular computers instead. This task, which involves simulating how particles of light bounce through a complex network, was supposed to be too hard for normal machines. But the new program runs quickly on a single ordinary computer chip and matches what the best quantum machines can do. This matters because it shows that classical computers may still keep up with quantum ones in some areas, meaning the race isn’t over yet — and cheaper, more accessible tools might solve problems we thought needed futuristic technology. Summary: This paper introduces a novel classical algorithm that efficiently emulates Gaussian Boson Sampling (GBS), a computational task designed to demonstrate quantum advantage in photonic systems. Contrary to expectations that large-scale GBS is intractable for classical computers, the authors show their emulator can simulate 100-mode GBS experiments faster than current quantum hardware achieves, using only a single CPU or GPU. The algorithm is "embarrassingly parallelizable," meaning it scales efficiently across multiple processors, allowing small clusters to match performance previously requiring over 100 GPUs. These gains stem from algorithmic and implementation optimizations that reduce memory and computational overhead. The authors argue that their approach could generalize to other quantum sampling paradigms, including single-photon inputs, pseudo-photon-number resolution, and even qubit-based sampling problems involving binary probability distributions. As such, the work challenges prevailing assumptions about the classical intractability of GBS and suggests that quantum advantage benchmarks must account for rapid advances in classical simulation techniques [arXiv:2504.01234, 2025]. Key Points: - A new classical algorithm can simulate 100-mode Gaussian Boson Sampling experiments more efficiently than current quantum devices. - The emulator runs on a single CPU or GPU and outperforms previous classical methods that required over 100 GPUs. - It is highly parallelizable, enabling scalable performance on small computing clusters. - Algorithmic innovations reduce memory and computational costs significantly. - The method may generalize to other photonic sampling scenarios and even qubit-based sampling problems. - Results challenge assumptions about quantum advantage in boson sampling tasks. - The emulator provides a fast, frugal alternative for benchmarking and simulating quantum sampling protocols. Notable Quotes: - "We demonstrate for the first time a classical simulation outperforming Gaussian boson sampling experiments of one hundred modes on established benchmark tests using a single CPU or GPU." - "Being embarrassingly parallelizable, a small number of CPUs or GPUs allows us to match previous sampling rates that required more than one hundred GPUs." - "We believe algorithmic and implementation improvements will generalize our tools to photo-counting, single-photon inputs, and pseudo-photon-number-resolving scenarios beyond one thousand modes." - "Most of the innovations in our tools remain valid for generic probability distributions over binary variables, rendering it potentially applicable to the simulation of qubit-based sampling problems." Data Points: - Simulation of 100-mode Gaussian Boson Sampling achieved. - Performance equivalent to experiments requiring >100 GPUs achieved with a small number of CPUs or GPUs. - Emulator demonstrated on single CPU or GPU hardware. - Target scalability projected to over 1,000 modes with future improvements. - Applicable to benchmark tests used in prior quantum experiments. Controversial Claims: - The assertion that a classical algorithm "outperforms" Gaussian boson sampling experiments implies that recent quantum supremacy claims in photonic systems may be invalidated or at least challenged under benchmark equivalence — a contentious position in the quantum computing community. - The claim that simulations on a single CPU/GPU match the performance of experiments requiring over 100 GPUs may be disputed without full disclosure of simulation fidelity, statistical accuracy, or comparison metrics. - The suggestion that this approach could scale beyond one thousand modes and apply to qubit-based sampling problems is speculative and unverified, representing an ambitious extrapolation beyond current results. Technical Terms: - Gaussian Boson Sampling (GBS) - Quantum advantage / quantum supremacy - Classical emulator - Photon-number-resolving detection - Single-photon inputs - Probability distributions over binary variables - Parallelizable algorithms - Photonic quantum computing - Sampling problems - NISQ-era algorithms - Classical intractability - Expectation values of observables —Ada H. Pemberley Dispatch from The Prepared E0