Google 发布 Willow 量子芯片

2024 年 12 月，谷歌的量子团队公布了一款名为 Willow 的芯片。它的核心参数是 105 个量子比特，但真正让物理学界和科技圈沸腾的，不是比特数量，而是一个反直觉的结果：当芯片规模增大时，错误率竟然呈指数级下降。这正是量子计算几十年来梦寐以求的东西——有效的量子纠错。

要理解这个结果的份量，得先明白量子计算的困境。量子比特极其脆弱，外界一丁点噪声就会让计算出错；更麻烦的是，纠错本身也要消耗额外的量子比特，规模越大，噪声越多，纠错成本越高。很长一段时间，研究者只能在「小规模、易出错」和「大规模、难纠错」之间挣扎。Willow 的实验第一次证明：把量子比特组织得更大，可以让错误率真正降下来——「越大越稳」从理论变成了可验证的事实。

Willow 还跑了一个经典基准测试：随机电路采样。谷歌声称这个任务用经典超算需要极长时间模拟，而 Willow 在几分钟内完成。这个对比充满争议——反对者认为问题本身对经典计算机不友好——但它确实展示了量子芯片在某些特定任务上的潜力。

对 AI 行业来说，Willow 的意义在于它和「算力」的叙事紧密相连。AI 的每次飞跃都伴随着算力的指数增长，而当摩尔定律逼近物理极限，量子计算被普遍视为后摩尔时代算力的潜在拐点。Willow 虽然离「容错量子计算机」还有距离，但它把这条路的时间表，向前推进了一步。

回看 Willow，它的价值不在于立刻改变 AI，而在于让「量子纠错可行」从论文里的推演变成了芯片上的证据。这是量子计算发展史上的一个关键坐标，也是人类探索「下一代算力」过程中的重要一步。下一次量子计算真正改变世界时，Willow 会是被反复提及的起点之一。

In December 2024 Google's quantum team unveiled a chip named Willow. Its headline spec was 105 qubits, but what excited physicists and the tech world was a counterintuitive result: as the chip scaled up, error rates fell exponentially. That is what quantum computing had dreamed of for decades—effective quantum error correction.

To understand the weight, consider quantum computing's dilemma. Qubits are fragile; the slightest noise corrupts a computation, and correcting errors itself consumes extra qubits. Scale up and you get more noise and higher correction costs. For a long time researchers were stuck between "small and error-prone" and "large and hard to correct". Willow's experiment was the first to show that organizing qubits on a larger scale can genuinely bring errors down—"bigger means more stable" went from theory to verified fact.

Willow also ran a classic benchmark: random circuit sampling. Google claimed the task would take classical supercomputers an extreme amount of time, while Willow finished in minutes. The comparison was contested—critics argued the problem is inherently unfriendly to classical machines—but it did showcase quantum's potential on specific tasks.

For the AI industry, Willow matters because of its tie to the compute narrative. Every AI leap has come with exponential compute growth, and as Moore's law nears physical limits, quantum is widely seen as the potential inflection point for post-Moore compute. Willow is far from a fault-tolerant quantum computer, but it moved that timeline forward a step.

Looking back, Willow's value lies not in immediately changing AI, but in turning "quantum error correction is feasible" from paper reasoning into on-chip evidence. It is a key coordinate in quantum computing history and an important step in the human quest for next-generation compute. When quantum computing truly changes the world, Willow will be among the starting points people keep returning to.