A team of quantum researchers has demonstrated that encoding information in light itself, rather than relying on traditional quantum bits, can dramatically improve both the speed and accuracy of quantum computations. By harnessing squeezed light, an entangled state where quantum fluctuations in two beams are correlated, scientists achieved a 5.4-fold increase in phase estimation precision and a 3.6-fold acceleration in time-to-solution compared to conventional methods. This breakthrough suggests that continuous-variable quantum systems, which leverage the amplitude and phase of light, may offer a complementary pathway to quantum advantage that deserves far more attention than it currently receives. What Makes Squeezed Light Different From Traditional Quantum Computing? While most quantum computing headlines focus on qubits, discrete units of quantum information, a growing body of research explores an alternative: continuous-variable (CV) quantum systems. Unlike qubits, which exist in discrete states, CV systems encode information in the smooth, continuously varying properties of light. Think of it as the difference between a light switch (on or off) and a dimmer (infinite positions between off and full brightness). Researchers led by Matthew A. Feldman developed a continuous-variable quantum compiler that digitizes complex analog quantum operations into sequences of basic gates, creating what they call a "quantum digital twin" for optical operations. The key innovation lies in their use of two-mode squeezed light. In this entangled state, quantum fluctuations in two separate light beams become correlated in a way that reduces uncertainty in one variable while increasing it in another. By carefully tuning the "squeezing parameter," the strength of these quantum correlations, the team could reshape the mathematical landscape that guides the learning process. This control proved transformative for both precision and speed. "Inspired by the techniques of machine learning, we have developed a novel experimental approach for learning quantum optical operations," said Matthew A. Feldman. Matthew A. Feldman, Lead Researcher How Does Tuning the Cost Landscape Improve Quantum Learning? In quantum machine learning, the "cost landscape" is a mathematical representation of how close a quantum system is to achieving its desired operation. A bumpy, irregular landscape makes it harder for algorithms to converge on the right answer; a smooth landscape allows faster, more accurate learning. The researchers discovered that by adjusting the squeezing parameter, they could deliberately reshape this landscape to facilitate both faster convergence and higher precision. This is analogous to smoothing out a bumpy road so a vehicle can travel faster and more efficiently. The practical implications are significant. The team successfully compiled an optical phase operation, a fundamental building block for more complex quantum algorithms, while simultaneously achieving improvements across multiple performance metrics. Their approach also demonstrated scalability, suggesting the method can be extended to higher-dimensional tasks beyond simple phase operations. Steps to Understanding Continuous-Variable Quantum Advantage - Discrete vs. Continuous: Traditional quantum computers use qubits that exist in discrete states (0, 1, or superposition), while continuous-variable systems encode information in the amplitude and phase of light, offering infinite intermediate states. - Squeezed Light as a Resource: Two-mode squeezed light creates entangled photon pairs where quantum fluctuations are correlated, reducing uncertainty in one variable and enabling more precise measurements than classical light. - Cost Landscape Optimization: By tuning the squeezing parameter, researchers can reshape the mathematical surface that guides quantum learning, enabling faster convergence and higher accuracy simultaneously. - Scalability Potential: The demonstrated approach extends beyond simple operations to higher-dimensional compilation tasks, suggesting broader applicability across quantum algorithms. Why Is This Approach Gaining Momentum Alongside Superconducting Qubits? The quantum computing landscape is increasingly recognizing that no single approach will dominate all applications. Google Quantum AI recently announced it is expanding its research to include neutral atom quantum computing alongside its established superconducting qubit program. This dual-track strategy reflects a fundamental insight: different quantum modalities excel at different tasks and have complementary strengths. Superconducting qubits have already scaled to circuits with millions of gate and measurement cycles, with each cycle taking just a microsecond. However, they face challenges in scaling to tens of thousands of qubits. Neutral atoms, by contrast, have already scaled to arrays with approximately ten thousand qubits, but their slower cycle times, measured in milliseconds, have limited their ability to perform deep circuits with many sequential operations. Continuous-variable systems using squeezed light represent yet another approach, potentially offering advantages in precision and learning efficiency that neither discrete-variable system can match. "Superconducting processors are easier to scale in the time dimension (circuit depth), while neutral atoms are easier to scale in the space dimension (qubit count)," noted researchers at Google Quantum AI. Google Quantum AI Research Team The research community's growing investment in multiple quantum modalities suggests that the path to practical quantum advantage may not be a single winner-take-all competition. Instead, different quantum technologies will likely serve different purposes. Continuous-variable systems optimized for optical phase learning could excel at specific machine learning tasks, while superconducting and neutral atom systems handle other workloads. This ecosystem approach increases the likelihood of delivering quantum computing solutions to real-world problems sooner. What Are the Practical Implications for Quantum Machine Learning? The 5.4-fold improvement in precision and 3.6-fold speedup achieved by Feldman's team are not merely incremental gains. They suggest a fundamental shift in how efficiently quantum operations can be learned and implemented. For quantum machine learning applications, where precision and speed directly impact the viability of algorithms, these improvements could unlock new capabilities previously thought impractical. The ability to manipulate the cost function through variable squeezing provides researchers with a powerful tool for algorithm design. Rather than accepting the cost landscape that nature provides, quantum engineers can now actively shape it to suit their computational goals. This level of control, combined with the inherent advantages of quantum light, positions continuous-variable systems as a serious contender in the race to build practical quantum computers for real-world problems. As the quantum computing field matures, the diversity of approaches being pursued by leading institutions suggests that the future of quantum advantage will likely involve a portfolio of complementary technologies, each optimized for specific classes of problems. The continuous-variable quantum compiler represents an important addition to that portfolio, offering a path to quantum advantage that deserves equal attention alongside more widely publicized qubit-based systems.
