Decades have passed since the great minds of physics, including Richard Feynman and David Deutsch, predicted that the laws of quantum mechanics could give rise to a computing paradigm that — for certain tasks — is superior to classical computing. But controlling fragile quantum systems well enough to construct even the most primitive quantum computing hardware has proved taxing. Experimental advances in the past few years have hushed the sceptics of quantum computing. However, the point that it is not entirely clear which application of quantum computers will redeem the hard work remains valid. This Insight discusses the applications in which quantum computers may excel and how software will actually run on such machines. Just as programming languages and compilers facilitate interaction with the semiconductor transistors in a classical computer, many layers of software tools will sit between quantum algorithms and hardware. An important component is quantum error-correcting code. The fragility of quantum bits leads to errors during computation, and choices about how to make quantumcomputing architectures fault tolerant have a knock-on effect on higher layers of the quantum tool chain. With quantum programming languages and compilers to hand, the quantum software engineer can implement ‘killer’ software applications, in which the speed afforded by quantum computers will have real-world impact. Factoring using Shor’s algorithm is one potential application because it could break current methods of encryption. Yet cryptographers are already devising classical cryptosystems that would guarantee security even if quantum computers achieve factoring at mesmerising speeds. Perhaps quantum machine learning will also turn out to be a killer application — it has, at least, been an important motivator for big technology companies to invest in quantum computing. As the practical relevance of quantum computing becomes clearer, we should not forget the part that foundational thinking played in its inception. Research on the fundamental limits of classical versus quantum computing remains fascinating — and may even help to surmount quantum engineering hurdles.

中文翻译：

---

量子软件

自从理查德·费曼 (Richard Feynman) 和大卫·多伊奇 (David Deutsch) 等伟大的物理学思想家预测量子力学定律可能会产生一种计算范式，对于某些任务而言，这种范式优于经典计算已经过去了几十年。但控制脆弱的量子系统足以构建最原始的量子计算硬件已被证明是繁重的。过去几年的实验进展使量子计算的怀疑论者平息。然而，目前尚不完全清楚量子计算机的哪些应用将弥补艰苦的工作仍然有效。此见解讨论了量子计算机可能擅长的应用程序以及软件将如何在此类机器上实际运行。正如编程语言和编译器促进与经典计算机中的半导体晶体管的交互一样，许多软件工具层将位于量子算法和硬件之间。一个重要的组成部分是量子纠错码。量子比特的脆弱性会导致计算过程中出现错误，而关于如何使量子计算架构容错的选择会对量子工具链的更高层产生连锁反应。有了量子编程语言和编译器，量子软件工程师就可以实现“杀手级”软件应用程序，其中量子计算机提供的速度将对现实世界产生影响。使用 Shor 算法的因式分解是一种潜在的应用，因为它可能会破坏当前的加密方法。然而，密码学家已经在设计经典密码系统，即使量子计算机以令人着迷的速度实现分解，也能保证安全性。也许量子机器学习也将成为一个杀手级应用——至少，它已经成为大型科技公司投资量子计算的重要动力。随着量子计算的实际相关性变得越来越清晰，我们不应该忘记基础思维在其诞生之初所起的作用。对经典计算与量子计算的基本限制的研究仍然令人着迷——甚至可能有助于克服量子工程障碍。