立即打开
霍尼韦尔宣布研发出最新量子计算技术

霍尼韦尔宣布研发出最新量子计算技术

Jeremy Kahn 2021年07月22日
这种情况表明,霍尼韦尔已经在一定程度上成为新兴量子计算领域的领导者。

工业巨头霍尼韦尔和其最近收购的一家英国软件公司——剑桥量子计算(Cambridge Quantum Computing)宣布在量子计算性能方面实现了三次大飞跃。

这些声明旨在表明,霍尼韦尔已经在一定程度上成为新兴量子计算领域的领导者,而该领域此前一直由谷歌和IBM等消费者更为熟悉的大型科技公司主导。霍尼韦尔的量子计算机能够通过微软的Azure云计算平台访问,不过微软也在研发自家量子计算硬件,只是还没有完善。据报道,亚马逊也正在组建团队,以建造自研量子硬件。

此外,7月21日的这份声明还意在展示霍尼韦尔的量子部门与剑桥量子计算的合并前景。前者专注于量子硬件,后者则专注于量子计算机的软件。今年6月,两家公司宣布合并,合并后的公司将从霍尼韦尔剥离出来,成为一家独立的公司。

霍尼韦尔将仍然持有新公司55%的股权并将有权使用其技术,但新公司将有能力从外部寻求更多的资金。

实时校正

霍尼韦尔的量子计算部门——霍尼韦尔量子解决方案的研究人员证明,计算机可以实时纠正目前量子计算机在计算中可能出现的错误,并且在错误出现时发现和纠正。这是对先前技术的一个改进,在以前,这些错误只能够在计算完成后才被发现和校正,而这可能会导致处理时间更长。

计算中出现的错误是目前阻碍量子计算在现实世界应用的因素之一。霍尼韦尔将10个物理量子位(量子计算机中执行计算的部分)中的7个组合在一起,形成了一个计算单元,即所谓的“逻辑量子位”,从而实现了实时误差校正。

“大型企业级问题需要精确以及错误修正的逻辑量子位元来校正和缩小。”霍尼韦尔量子解决方案的负责人托尼•阿特利在一份声明中说。

霍尼韦尔还表示,公司目前已经实现了1024个量子体积,这是今年3月公布的纪录的两倍。量子体积是一种考虑了几个不同变量的性能刻度。去年10月,初创公司IonQ推出了一款量子计算机,采用了与霍尼韦尔类似的底层捕获离子技术,据称其“预计量子体积”达到了400万。IonQ还表示,其认为IBM在2016年首次发布的这个基准指标可能无法持续很久,因为底层硬件类型之间存在差异。

以少抵多

这份声明的最后,剑桥量子计算宣布,他们在算法上取得了进展,可以用较少的量子位(量子计算机的逻辑处理部分)来解决复杂的优化问题。今天的量子计算机通常只有几十个这样的量子位,而与之相比,在今天的标准计算机芯片上有数十亿个开关——所以能够用更少的量子位做更多的事情,对现实世界的应用来说相当重要。同时,这些优化问题也往往在商业世界的“数学题”中涌现,比如快递员送货的最优路线、金融投资组合中平衡风险和收益的最优方式、工厂设备的最高效运行方式,以及减少维护性停工的最好办法等等。

这家初创公司认为,新的办法可以将解决部分复杂优化问题的时间缩短到原来的100倍。“更快的量子算法能够对面临复杂优化问题的各行各业产生深远的影响。”剑桥量子计算的创始人及首席执行官伊利亚斯•汗在一份新闻稿中说。他还提及了钢铁制造这一领域,量子算法可以让制造工艺更高效、更节省成本。

同时,量子计算正在稳步进军商业应用领域。从高盛到博世,数十家公司已经建立了依赖通过云计算平台访问量子计算机的试点项目。还有更多的公司在使用受量子计算技术启发的算法,但这些算法仍然在标准硬件上运行。

量子计算机利用量子物理中的现象来进行计算。例如,在标准计算机中,信息是由一个名为“位”的二进制单位表示的,这个二进制单位只能够是0或者1;而在量子计算机中,信息由一个量子位表示,这个量子位可以同时是0和1。在标准计算机中,每一位是独立于其他每一位的;而在量子计算机中,每个量子位的状态可能会影响其他量子位的状态。这两种特性被称为叠加和纠缠,因此从理论上讲,量子计算机比现有的数字计算机(计算机科学家通常称其为“经典计算机”)具有指数级的处理能力。

此外还有其他不同之处:今天的数字计算机几乎都在由半导体构成的类似逻辑电路上运行,而量子计算机中的量子位能够通过多种不同的方式工作。比如,IBM和谷歌的量子计算机通过在极低温度下使用超导材料创建电路来工作;微软一直在尝试用超导和半导体材料相结合的量子位创造量子计算机;霍尼韦尔的计算机则采用了一种完全不同的方法,即依靠向材料发射激光来捕获离子。

目前的问题是,很难让量子位长时间处于量子状态,对捕获的离子量子位来说,最多只可以持续一分钟,而对于大多数其他超导电路来说,只有几分之一秒。当这些量子位脱离量子状态时,错误就会在计算中堆积起来。这些错误的存在,以及研究人员还没有弄清楚如何制造出量子位数量接近标准数字计算机逻辑门数量的芯片这一问题,意味着在大多数情况下,今天的量子计算机仍然不如大多数笔记本电脑高效、有用。这是正确的,除了一小部分非常困难的问题(许多问题与量子物理本身有关)之外,量子计算机是得到答案的唯一可证明的方法。(财富中文网)

编译:杨二一

工业巨头霍尼韦尔和其最近收购的一家英国软件公司——剑桥量子计算(Cambridge Quantum Computing)宣布在量子计算性能方面实现了三次大飞跃。

这些声明旨在表明,霍尼韦尔已经在一定程度上成为新兴量子计算领域的领导者,而该领域此前一直由谷歌和IBM等消费者更为熟悉的大型科技公司主导。霍尼韦尔的量子计算机能够通过微软的Azure云计算平台访问,不过微软也在研发自家量子计算硬件,只是还没有完善。据报道,亚马逊也正在组建团队,以建造自研量子硬件。

此外,7月21日的这份声明还意在展示霍尼韦尔的量子部门与剑桥量子计算的合并前景。前者专注于量子硬件,后者则专注于量子计算机的软件。今年6月,两家公司宣布合并,合并后的公司将从霍尼韦尔剥离出来,成为一家独立的公司。

霍尼韦尔将仍然持有新公司55%的股权并将有权使用其技术,但新公司将有能力从外部寻求更多的资金。

实时校正

霍尼韦尔的量子计算部门——霍尼韦尔量子解决方案的研究人员证明,计算机可以实时纠正目前量子计算机在计算中可能出现的错误,并且在错误出现时发现和纠正。这是对先前技术的一个改进,在以前,这些错误只能够在计算完成后才被发现和校正,而这可能会导致处理时间更长。

计算中出现的错误是目前阻碍量子计算在现实世界应用的因素之一。霍尼韦尔将10个物理量子位(量子计算机中执行计算的部分)中的7个组合在一起,形成了一个计算单元,即所谓的“逻辑量子位”,从而实现了实时误差校正。

“大型企业级问题需要精确以及错误修正的逻辑量子位元来校正和缩小。”霍尼韦尔量子解决方案的负责人托尼•阿特利在一份声明中说。

霍尼韦尔还表示,公司目前已经实现了1024个量子体积,这是今年3月公布的纪录的两倍。量子体积是一种考虑了几个不同变量的性能刻度。去年10月,初创公司IonQ推出了一款量子计算机,采用了与霍尼韦尔类似的底层捕获离子技术,据称其“预计量子体积”达到了400万。IonQ还表示,其认为IBM在2016年首次发布的这个基准指标可能无法持续很久,因为底层硬件类型之间存在差异。

以少抵多

这份声明的最后,剑桥量子计算宣布,他们在算法上取得了进展,可以用较少的量子位(量子计算机的逻辑处理部分)来解决复杂的优化问题。今天的量子计算机通常只有几十个这样的量子位,而与之相比,在今天的标准计算机芯片上有数十亿个开关——所以能够用更少的量子位做更多的事情,对现实世界的应用来说相当重要。同时,这些优化问题也往往在商业世界的“数学题”中涌现,比如快递员送货的最优路线、金融投资组合中平衡风险和收益的最优方式、工厂设备的最高效运行方式,以及减少维护性停工的最好办法等等。

这家初创公司认为,新的办法可以将解决部分复杂优化问题的时间缩短到原来的100倍。“更快的量子算法能够对面临复杂优化问题的各行各业产生深远的影响。”剑桥量子计算的创始人及首席执行官伊利亚斯•汗在一份新闻稿中说。他还提及了钢铁制造这一领域,量子算法可以让制造工艺更高效、更节省成本。

同时,量子计算正在稳步进军商业应用领域。从高盛到博世,数十家公司已经建立了依赖通过云计算平台访问量子计算机的试点项目。还有更多的公司在使用受量子计算技术启发的算法,但这些算法仍然在标准硬件上运行。

量子计算机利用量子物理中的现象来进行计算。例如,在标准计算机中,信息是由一个名为“位”的二进制单位表示的,这个二进制单位只能够是0或者1;而在量子计算机中,信息由一个量子位表示,这个量子位可以同时是0和1。在标准计算机中,每一位是独立于其他每一位的;而在量子计算机中,每个量子位的状态可能会影响其他量子位的状态。这两种特性被称为叠加和纠缠,因此从理论上讲,量子计算机比现有的数字计算机(计算机科学家通常称其为“经典计算机”)具有指数级的处理能力。

此外还有其他不同之处:今天的数字计算机几乎都在由半导体构成的类似逻辑电路上运行,而量子计算机中的量子位能够通过多种不同的方式工作。比如,IBM和谷歌的量子计算机通过在极低温度下使用超导材料创建电路来工作;微软一直在尝试用超导和半导体材料相结合的量子位创造量子计算机;霍尼韦尔的计算机则采用了一种完全不同的方法,即依靠向材料发射激光来捕获离子。

目前的问题是,很难让量子位长时间处于量子状态,对捕获的离子量子位来说,最多只可以持续一分钟,而对于大多数其他超导电路来说,只有几分之一秒。当这些量子位脱离量子状态时,错误就会在计算中堆积起来。这些错误的存在,以及研究人员还没有弄清楚如何制造出量子位数量接近标准数字计算机逻辑门数量的芯片这一问题,意味着在大多数情况下,今天的量子计算机仍然不如大多数笔记本电脑高效、有用。这是正确的,除了一小部分非常困难的问题(许多问题与量子物理本身有关)之外,量子计算机是得到答案的唯一可证明的方法。(财富中文网)

编译:杨二一

Industrial giant Honeywell and Cambridge Quantum Computing, a British software company Honeywell recently acquired, have announced a trio of big leaps forward in quantum computing performance.

The announcements were intended to show the extent to which Honeywell has emerged as a leader in the nascent quantum computing field, which has otherwise been dominated by big technology companies that are better known to consumers, such as Google and IBM. Honeywell’s quantum computer can be accessed through Microsoft’s Azure cloud-computing platform, although Microsoft has also been working on its own quantum computing hardware, which it has yet to perfect. Amazon is also reportedly assembling a team to build its own quantum hardware too.

July 21's announcement was also meant to showcase the promise of the merger of Honeywell’s quantum division, which has focused on quantum hardware, with Cambridge Quantum Computing, which has focused on software for quantum computers. In June, the two companies announced the combination and that the merged company would be spun out from Honeywell as a stand-alone corporation.

Honeywell will still own 55% of the new company and have the rights to use its technology, but the corporation will have the ability to raise additional outside financing.

Real-time corrections

Researchers at Honeywell Quantum Solutions, the company’s quantum computing division, demonstrated that it could correct in real time the errors that tend to creep into the calculations of today’s quantum computers, finding and correcting the mistakes as they occur. This is an advance on previous methods, in which such errors could only be spotted and corrected once a calculation had finished running, which made for potentially much slower processing times.

Errors are one of the things that is currently holding back many real-world applications of quantum computing. Honeywell managed to achieve its real-time error correction by yoking together seven of the 10 physical qubits—the parts of a quantum computer that perform calculations—to form a single calculating unit, known as a “logical qubit.”

“Big enterprise-level problems require precision and error-corrected logical qubits to scale successfully,” Tony Uttley, president of Honeywell Quantum Solutions, said in a statement.

Honeywell also said it had achieved a quantum volume of 1,024, doubling its previous record announced just in March. Quantum volume is a measure of performance that takes into account several different variables. Startup IonQ unveiled a quantum computer using an underlying trapped-ion technology similar to Honeywell’s in October that it said had an “expected quantum volume” of 4 million. IonQ also said it thought the benchmark metric, which was first promulgated by IBM in 2016, might not be useful for much longer because of differences between underlying hardware types.

More with less

Finally, Cambridge Quantum Computing announced that it had made an algorithmic advance that makes it possible to solve complex optimization problems using few qubits, which are the logical processing parts of a quantum computer. Today’s quantum computers often have just a few dozen of these qubits, compared to the billions of switches in today’s standard computer chips, so being able to do more with a fewer number of qubits is important for real-world applications. Optimization problems are also the sort of mathematical problem that pop up frequently in business, whether it is trying to find the most efficient route for a delivery courier, the best way to balance risk and reward in a financial portfolio, or the best way to run factory equipment quickly and reduce maintenance stoppages.

The startup said its methods could speed up the time it would take to solve some complex optimization problems by up to a factor of 100. “Faster quantum algorithms can have a profound impact on a variety of industries that face complicated optimization problems,” Ilyas Khan, the founder and chief executive officer of Cambridge Quantum Computing, said in a press release. He mentioned steel manufacturing as an area where quantum algorithms could help make manufacturing processes more efficient and cost-effective.

Quantum computing is making steady inroads into commercial applications. Dozens of companies, from Goldman Sachs to Bosch, have set up pilot projects that rely on access to quantum computers accessed through cloud-computing platforms. More still are using algorithms inspired by quantum computing techniques, but which run on standard hardware.

Quantum computers use phenomenon from quantum physics to make their calculations. For instance, while in a standard computer, information is represented by a binary unit called a bit, that can be either a 0 or 1. In a quantum computer, information is represented by a qubit, which can be both 0 and 1 at the same time. In a standard computer, each bit is also independent from every other bit. In a quantum computer, however, the state of every qubit can potentially influence the state of every other qubit. These two properties, which are called superposition and entanglement, theoretically give quantum computers exponentially more processing power than existing digital computers, which computer scientists often call “classical computers.”

There are other differences as well: Today’s digital computers pretty much all run on similar logic circuits constructed from semiconductors. The qubits in quantum computers can be made in a wide variety of different ways: IBM’s and Google’s quantum machines work by creating circuits using superconducting materials at extremely low temperatures. Microsoft has been trying to create a quantum computer using qubits that marry superconducting and semiconducting materials. Honeywell’s machine uses an entirely different method that relies on firing lasers into a material to trap ions.

The problem is that it is difficult to keep the qubits in a quantum state for long—one minute at most for a trapped ion qubit, just fractions of a second for most other superconducting circuits. And when these qubits fall out of a quantum state, errors pile up in their calculations. The presence of these errors and the fact that researchers have not yet figured out how to create chips with anywhere near the number of qubits as there are logical gates in a standard digital computer, mean that in most cases today’s quantum computers are still less powerful and less useful than even most laptops. This is true except for a small subset of very difficult problems—many related to quantum physics itself—where a quantum computer is the only provable way to arrive at a valid answer.

最新:
  • 热读文章
  • 热门视频
活动
扫码打开财富Plus App