一家小型英國初創(chuàng)公司開發(fā)了一種算法,該算法可顯著降低運行計算所需的量子計算能力水平,從而有望為開發(fā)特殊新材料鋪平道路。
Phasecraft公司的科學(xué)家和布里斯托大學(xué)(University of Bristol)的研究人員共同研究發(fā)現(xiàn),按照現(xiàn)在科技發(fā)展的速度,量子物理學(xué)中的一個重大難題(用當(dāng)今超級計算機難以解決)有望會在三年內(nèi)用量子計算機解決。這比大多數(shù)專家先前的預(yù)測要快得多。
量子計算機是利用量子物理的特殊屬性來運行計算的機器。因此,量子計算機有可能比當(dāng)今最快的超級計算機強大得多。
許多公司開始試用量子計算機,量子計算機由IBM、霍尼韋爾和谷歌等公司以及一些初創(chuàng)公司通過云計算接口提供。但到目前為止,大多數(shù)企業(yè)只在這些機器上運行概念驗證項目,這些機器還無法模擬許多復(fù)雜的系統(tǒng),例如模擬亞原子或分子級相互作用。
去年,谷歌宣布已經(jīng)實現(xiàn)了“量子霸權(quán)(quantum supremacy)”,這是量子計算領(lǐng)域的一個里程碑,即使用量子計算機運行傳統(tǒng)計算機無法在合理時間范圍內(nèi)完成的計算。上周,一個中國研究小組宣布,他們已經(jīng)用另一種量子計算機取得了類似的突破。
“玩具問題”
但在這兩種情況下,量子計算機所解決的特殊難題是Phasecraft聯(lián)合創(chuàng)始人兼?zhèn)惗卮髮W(xué)學(xué)院(University College London)量子物理學(xué)教授約翰?莫頓所述的“玩具問題”,即這些計算僅僅是為了證明量子計算機可以完成普通超級計算機做不到的事情。而這些問題對實際應(yīng)用并無太大的意義,例如弄清楚如何開發(fā)更高效的肥料制造工藝或更好的電池。
Phasecraft在美國物理學(xué)會(American Physical Society)出版的學(xué)術(shù)期刊《物理評論B》(Physical Review B)上今天發(fā)表的一篇同行評議論文指出了另一種情況。文中提及一個稱為費米-哈伯德模型(Fermi-Hubbard model)的難題,描述了一類稱為費米子(包含電子的一組費米子)的亞原子粒子在固體中跳躍的行為。能夠計算出這個模型是向制造出無需保持超低溫也能具備超導(dǎo)特性的材料邁出的重要一步。但是,對于一個有超過幾十個粒子位置的系統(tǒng)而言,這一性能是當(dāng)今傳統(tǒng)超級計算機無法企及的。
Phasecraft研究人員證實,一種結(jié)合量子元素和經(jīng)典元素的算法可以使用一臺約有8,000個量子門(即量子計算機可以執(zhí)行的邏輯操作次數(shù))的量子計算機來求解大型固體的費米-哈伯德模型。這是以前認為求解該模型所需的量子門數(shù)量的十分之一。
馬里蘭大學(xué)(University of Maryland)計算機科學(xué)家安德魯?柴爾德斯表示:“他們的工作表明,較淺電路可以提供這個模型的有用信息,使之更適合實際量子硬件使用,這太令人驚訝了?!?/p>
現(xiàn)有的量子計算機已經(jīng)有足夠的量子處理單元(即量子比特),理論上可以執(zhí)行如此多次邏輯運算,但到目前為止,科學(xué)家們尚未弄清楚如何構(gòu)建這種規(guī)模的電路。谷歌的量子優(yōu)越性實驗是在54個量子比特的Sycamore量子處理器上進行的,使用了由430個雙量子比特門和1,113個單量子比特門組成的電路。
IBM已經(jīng)宣布計劃在2023年前推出有1,000個量子比特的量子計算機。有了這種規(guī)模的量子計算機,科學(xué)家們就有可能構(gòu)建具有足夠量子門的電路,就可以使用Phasecraft演示的算法來求解費米-哈伯德模型。Phasecraft聯(lián)合創(chuàng)始人兼董事及布里斯托大學(xué)量子計算機研究人員阿什利?蒙塔納羅說:“我們有可能在未來兩三年內(nèi)做一些激動人心的事情?!?/p>
融資協(xié)議
Phasecraft已與制造量子計算機的初創(chuàng)公司Rigetti(總部位于加州)及谷歌建立了合作關(guān)系,該公司致力于與材料科學(xué)和化學(xué)公司合作設(shè)計量子算法,使其能夠使用量子計算機解決難題。
這家初創(chuàng)公司目前約有10名員工,公司周四宣布,在倫敦風(fēng)投公司LocalGlobe牽頭的一輪融資中,公司獲得了500萬美元種子資金,另一家專門從事早期投資的倫敦風(fēng)投公司Episode 1也參與了此輪融資。演唱會信息和票務(wù)服務(wù)應(yīng)用程序Songkick的前聯(lián)合創(chuàng)始人伊恩?霍加斯現(xiàn)在是著名的天使投資人和種子投資人,將出任Phasecraft的董事會主席。
新的投資使Phasecraft自2018年成立以來通過風(fēng)險投資和研究資助籌集的資金總額達到740萬美元。該公司已經(jīng)收到了UCL技術(shù)基金(UCL Technology Fund)和Parkwalk Advisors的前期融資以及創(chuàng)新英國(Innovate UK)基金的資助。(財富中文網(wǎng))
翻譯:郝秀
審校:汪皓
一家小型英國初創(chuàng)公司開發(fā)了一種算法,該算法可顯著降低運行計算所需的量子計算能力水平,從而有望為開發(fā)特殊新材料鋪平道路。
Phasecraft公司的科學(xué)家和布里斯托大學(xué)(University of Bristol)的研究人員共同研究發(fā)現(xiàn),按照現(xiàn)在科技發(fā)展的速度,量子物理學(xué)中的一個重大難題(用當(dāng)今超級計算機難以解決)有望會在三年內(nèi)用量子計算機解決。這比大多數(shù)專家先前的預(yù)測要快得多。
量子計算機是利用量子物理的特殊屬性來運行計算的機器。因此,量子計算機有可能比當(dāng)今最快的超級計算機強大得多。
許多公司開始試用量子計算機,量子計算機由IBM、霍尼韋爾和谷歌等公司以及一些初創(chuàng)公司通過云計算接口提供。但到目前為止,大多數(shù)企業(yè)只在這些機器上運行概念驗證項目,這些機器還無法模擬許多復(fù)雜的系統(tǒng),例如模擬亞原子或分子級相互作用。
去年,谷歌宣布已經(jīng)實現(xiàn)了“量子霸權(quán)(quantum supremacy)”,這是量子計算領(lǐng)域的一個里程碑,即使用量子計算機運行傳統(tǒng)計算機無法在合理時間范圍內(nèi)完成的計算。上周,一個中國研究小組宣布,他們已經(jīng)用另一種量子計算機取得了類似的突破。
“玩具問題”
但在這兩種情況下,量子計算機所解決的特殊難題是Phasecraft聯(lián)合創(chuàng)始人兼?zhèn)惗卮髮W(xué)學(xué)院(University College London)量子物理學(xué)教授約翰?莫頓所述的“玩具問題”,即這些計算僅僅是為了證明量子計算機可以完成普通超級計算機做不到的事情。而這些問題對實際應(yīng)用并無太大的意義,例如弄清楚如何開發(fā)更高效的肥料制造工藝或更好的電池。
Phasecraft在美國物理學(xué)會(American Physical Society)出版的學(xué)術(shù)期刊《物理評論B》(Physical Review B)上今天發(fā)表的一篇同行評議論文指出了另一種情況。文中提及一個稱為費米-哈伯德模型(Fermi-Hubbard model)的難題,描述了一類稱為費米子(包含電子的一組費米子)的亞原子粒子在固體中跳躍的行為。能夠計算出這個模型是向制造出無需保持超低溫也能具備超導(dǎo)特性的材料邁出的重要一步。但是,對于一個有超過幾十個粒子位置的系統(tǒng)而言,這一性能是當(dāng)今傳統(tǒng)超級計算機無法企及的。
Phasecraft研究人員證實,一種結(jié)合量子元素和經(jīng)典元素的算法可以使用一臺約有8,000個量子門(即量子計算機可以執(zhí)行的邏輯操作次數(shù))的量子計算機來求解大型固體的費米-哈伯德模型。這是以前認為求解該模型所需的量子門數(shù)量的十分之一。
馬里蘭大學(xué)(University of Maryland)計算機科學(xué)家安德魯?柴爾德斯表示:“他們的工作表明,較淺電路可以提供這個模型的有用信息,使之更適合實際量子硬件使用,這太令人驚訝了?!?/p>
現(xiàn)有的量子計算機已經(jīng)有足夠的量子處理單元(即量子比特),理論上可以執(zhí)行如此多次邏輯運算,但到目前為止,科學(xué)家們尚未弄清楚如何構(gòu)建這種規(guī)模的電路。谷歌的量子優(yōu)越性實驗是在54個量子比特的Sycamore量子處理器上進行的,使用了由430個雙量子比特門和1,113個單量子比特門組成的電路。
IBM已經(jīng)宣布計劃在2023年前推出有1,000個量子比特的量子計算機。有了這種規(guī)模的量子計算機,科學(xué)家們就有可能構(gòu)建具有足夠量子門的電路,就可以使用Phasecraft演示的算法來求解費米-哈伯德模型。Phasecraft聯(lián)合創(chuàng)始人兼董事及布里斯托大學(xué)量子計算機研究人員阿什利?蒙塔納羅說:“我們有可能在未來兩三年內(nèi)做一些激動人心的事情?!?/p>
融資協(xié)議
Phasecraft已與制造量子計算機的初創(chuàng)公司Rigetti(總部位于加州)及谷歌建立了合作關(guān)系,該公司致力于與材料科學(xué)和化學(xué)公司合作設(shè)計量子算法,使其能夠使用量子計算機解決難題。
這家初創(chuàng)公司目前約有10名員工,公司周四宣布,在倫敦風(fēng)投公司LocalGlobe牽頭的一輪融資中,公司獲得了500萬美元種子資金,另一家專門從事早期投資的倫敦風(fēng)投公司Episode 1也參與了此輪融資。演唱會信息和票務(wù)服務(wù)應(yīng)用程序Songkick的前聯(lián)合創(chuàng)始人伊恩?霍加斯現(xiàn)在是著名的天使投資人和種子投資人,將出任Phasecraft的董事會主席。
新的投資使Phasecraft自2018年成立以來通過風(fēng)險投資和研究資助籌集的資金總額達到740萬美元。該公司已經(jīng)收到了UCL技術(shù)基金(UCL Technology Fund)和Parkwalk Advisors的前期融資以及創(chuàng)新英國(Innovate UK)基金的資助。(財富中文網(wǎng))
翻譯:郝秀
審校:汪皓
A small U.K. startup has developed an algorithm that significantly reduces the level of quantum computing power needed to run a calculation that could pave the way for the development of exotic new materials.
Scientists from the company, Phasecraft, along with researchers from the University of Bristol, showed that an important problem from quantum physics—one that is too difficult to solve on today’s supercomputers—could be within the reach of quantum computers within three years, given current rates at which the technology is developing. That is much sooner than most experts had previously forecast.
Quantum computers are machines that harness the peculiar properties of quantum physics to run their calculations. This makes them potentially much more powerful than today’s fastest supercomputers.
Many companies are beginning to experiment with quantum computers, which are being offered through cloud computing interfaces from companies such as IBM, Honeywell, and Google, as well as a number of startups. But so far, most businesses have only run proof-of-concept projects on these machines, which are not yet powerful enough to simulate many complex systems, such as the modeling of subatomic or molecular level interactions.
Last year, Google claimed it had achieved a milestone called “quantum supremacy,” using a quantum computer to run a calculation that a conventional computer could not crunch in a reasonable time span. Last week, a group of Chinese researchers said they had achieved a similar breakthrough using a different kind of quantum computer.
“Toy problems”
But in both cases, the particular problems that the quantum computers solved were what John Morton, a Phasecraft cofounder and professor of quantum physics at University College London, calls “toy problems”—calculations formulated solely to show that the quantum computer could do something a regular supercomputer can’t. They weren’t problems with clear implications for real world applications, such as figuring out how to create more efficient fertilizer manufacturing processes or better batteries.
What Phasecraft shows in a peer-reviewed paper published today in the academic journal Physical Review B, published by the American Physical Society, is different. It involves a problem called the Fermi-Hubbard model, which describes the behavior of a class of subatomic particles known as fermions—a group that includes electrons—as they hop around within a solid. Being able to compute this model is an important step toward creating materials that will exhibit superconducting properties without the need to keep them at ultra-freezing temperatures. But doing so for a system with more than a few tens of particle positions is beyond the reach of today’s conventional supercomputers.
The Phasecraft researchers proved that a kind of algorithm that combines quantum and classical elements could be used to solve the Fermi-Hubbard model for a large solid using a quantum computer with about 8,000 gates, a term that refers to the number of logical operations a quantum computer can perform. That is a tenth of the number of quantum gates that were previously thought necessary to solve the model.
“Their work suggests that surprisingly low-depth circuits could provide useful information about this model, making it more accessible to realistic quantum hardware,” Andrew Childs, a computer scientist at the University of Maryland, said of the research.
Existing quantum computers already have enough quantum processing units—known as qubits—to in theory perform this number of logical operations, but so far scientists have not figured out how to build circuits of that size. Google’s quantum supremacy experiment, which it performed on its 54-qubit Sycamore quantum processor, used a circuit consisting of 430 two-qubit gates and 1,113 single qubit gates.
IBM has already announced its plans to have a 1,000-qubit quantum computer available by 2023. With a quantum computer of that size, it is possible scientists will be able to build a circuit with enough gates to solve the Fermi-Hubbard model using the algorithm Phasecraft demonstrated. “We think it is plausible to do exciting things in the next two- to three-year period,” Ashley Montanaro, one of Phasecraft’s cofounders and directors, as well as a quantum computer researcher at the University of Bristol, said.
Funding deal
Phasecraft, which has partnerships with Rigetti, a California-based startup building quantum computers, and Google, is focused on working with materials science and chemical companies to design quantum algorithms that will allow them to start solving complex problems using quantum computers.
The startup, which currently employs about 10 people, also announced on Thursday that it has received $5 million in seed funding in a financing round led by LocalGlobe, a London venture capital firm, with participation from Episode 1, another London venture firm specializing in early stage investment. Ian Hogarth, the former cofounder of concert discovery app Songkick and now a prominent angel and seed investor, is joining Phasecraft’s board as chairperson.
The new investment brings the total Phasecraft has raised, in both venture funding and research grants, since its founding in 2018, to $7.4 million. The company had received prior funding from the UCL Technology Fund and Parkwalk Advisors and grants from Innovate UK.