【视频】实现量子计算有哪些可能方案 | 姚期智

姚期智

中国科学院院士，美国科学院外籍院士，美国科学与艺术学院外籍院士，国际密码协会会士，清华大学交叉信息研究院院长

2017年初，姚期智与物理学家杨振宁放弃美国国籍转为中国国籍。

自从2004年辞去普林斯顿大学终身教职回国后，姚期智始终活跃在教育第一线，他主导并与微软研究院共同合作，在清华创办了如今大名鼎鼎的「姚班」，培养出了一大批中国计算机科学的顶尖人才，其门生遍布国内外 AI 产业和计算机科学研究的各个关键领域。

2010年，姚期智还创立了清华交叉信息研究院，信息科学与多个学科在这里交叉互动，成果叠出，以开疆辟土的恢弘气势填补了国内计算机科学在该领域的空白地带，为世界学术瞩目。

姚期智院士谈量子计算

·系列

视频

文字版·中文

现在我简要的介绍一下实验方面的情况。用于量子计算的技术手段，最初可能有十几二十种，都是用于构建量子比特的基础单元，但是许多年之后，似乎只有一些方案更有希望。

如果你关注科学论文，我认为超过20个比特，固态量子计算方案会有非常高的可靠性。目前，有一些原型机获得公布，它们来自IBM和Google，有些原型机的的比特数已经快要到达100。如果有一个好的实验工作，你将可以初步展示出量子计算的强大能力，而超导量子计算就是其中一个，我认为这也是大众关注度最高的一种方案。

另外，离子阱也是一个相当成熟的技术，我认为他们俩都有希望竞争第一。这里也有一些新技术方案。其中令我印象特别深刻的是利用金刚石来构建量子计算机。出于虚荣心，我乐于看见量子计算机是由金刚石做出来的，这样我就可以把它放在桌上展示给大家看。

此外还有利用光子技术方案，我们可以做玻色采样并将10个量子比特纠缠起来。顺便提一下，最后一个工作是在中国完成的，来自于中国科学技术大学的潘建伟教授领导的团队。

现在，大众对量子计算似乎比以前乐观了许多，并且有很多有天赋的量子物理学家，正致力于构建量子计算机所需要的各个部件。就像我刚才提到的，我们研究所正在用金刚石来试图推进这方面进展。我想说，所有的技术方案都有自己的长处和不足，并不是说一种技术方案就一定优于另一种，比如，金刚石方案，目前在可以设计和制造的比特数比离子阱和超导要少，但是金刚石方案可以在室温下工作并且是固态的，这一定程度上让人想起了硅技术，所以这种方案在未来还有发展的潜力。

关于离子阱嘛，首先你要制备出这种离子，这样它们就可以用做量子比特了。离子阱在有些方面很有优势，比如它们非常稳定，目前它们的退相干时间可以达到十分钟。然后你要用磁场去稳定住它们，让它们排成一排.

现在我来总结一下，首先，我认为量子计算非常让人激动，从智力的角度而不是功利的角度来看的话，如果我们比较量子计算机和经典计算机，对于经典计算机来说，它的设计原则非常简单并符合常识，这种设计原则最早可以追溯到古代科学家，比如欧几里得，阿基米德和达芬奇。

当你这样想时，也许会认为最早提出科学计算机的不是图灵，而是古希腊的科学家-欧几里得，如果你认为自己可以利用尺子和圆规来研究几何学。你可能问这样的问题，你能计算哪些几何问题? 这真的是一个非常漂亮的计算理论系统，你可以加减乘除或者求平方根，现在量子计算是非常不一样的，它基于20世纪的现代物理学，它的设计原则非常晦涩并且反直觉。它在智力上是一件很让人激动的事情，某种意义上，我们都有机会成为第二个图灵，我认为 Turing，Church 和 Kleanie 所在的时代是一个伟大的时代。

那时Kurt Gödel刚刚发表了一个不完全性定理（哥德尔于1931年提出，他证明了任何一个形式系统，只要包括简单的初等数论描述，而且是自洽的，它必定包含某些系统内所允许的方法既不能证明真也不能证伪的命题）, 计算的本质是什么似乎成了一个悬而未决的问题, 对很多研究者来说，那是一个思考， “计算的意义是什么”这一重要问题的黄金时期，这是该问题第一次被深入研究。

它吸引了极大的兴趣，就像一个从未被探索过的原始森林,任何进展都显得非常重要，现在，量子计算给了我们第二次这样的机会。当你考虑量子计算机能做些什么的时候，你将要回到最开始设计版图的阶段，因为任何有潜力的物理器件，量子器件各类实验方法，实验过程，对你来说都是建造量子计算机的潜在方案。

所以，这就是为什么量子计算机具有无限的可能性，就像刚才我所提到的，可以用X光来进行上述的计算，也可以考虑利用它做其他类型的计算，并且能从中获得什么。因此这是一个非常广阔的领域，等待着去探索。

第二个结论是，我认为量子计算机我认为在过去的十几年间的进展是巨大的。现在我们谈一谈，一些公司已经制造出接近100量子比特的原型机，这项工作既需要物理实验，也需要工程技术。因为在工程学角度，如果你要建立一个超导量子计算机，你需要非常多的电子学工程师来协助你，所以工程技术也扮演着重要角色。

当工业界开始介入，这个领域会充满希望，因为这是一个积极的信号，另一个乐观的迹象是越来越多的国家，开始投入研究经费到量子计算中，而且更能说明问题的是，非常多顶尖的IT公司启动了量子计算项目。所以可以看出，这些公司是非常有眼光的，可以确定的是，这个领域即将会有丰硕的成果产出。

但是我们也必须沉下心来意识到，建造一个实用的量子计算机，即使到了最后时刻，它也可能还有一段非常冗长困难的距离。

最后，让我们回到量子计算上，因为大家知道量子计算有它自己的运行逻辑，并且将很快被实现，但现在最好再用我们的智慧，思考一下量子计算机的定位，以及量子计算对未来信息科学领域的影响。在我的观念中，我认为自然界有两个非常伟大的成就，第一是自然界设计了一个非常复杂的规律，也就是量子定律，它的因果逻辑如此复杂以至于无法解析计算，但它却让这个世界有了许多精彩纷呈的现象。

第二件事，大自然通过演化创造了一个物种——人类，人类的大脑几乎是宇宙中最复杂的东西，在创造和推理方面，它拥有非常惊人的能力。

现在，来看看我们如何追赶上大自然的脚步。如果我们成功的建成了一个通用的量子计算机，这意味着我们最终有能力去求解量子力学方程，也就是说我们可以利用已知的物理定律，像自然界一样，去创造一些东西。所以我认为，我们已经进入一个值得期待的时期，在这个世纪里取得伟大的进步。

至于另外一个问题，我们是否可以创造一个类似人类大脑的智能？事实上，现在的计算机科学家已经在这方面做出了很多工作，即人工智能领域，但现在的水平还不能与人脑相提并论，但我们是否有可能建立一个与人脑智力相当的系统呢？

如果在本世纪接下来的时间里，我们能够取得极大的进步，这将会是一件很棒的事情。我们可以宣称，我们可以做到与自然界一样的事情，虽然我们现在正努力在这方面取得进步，但我们还有很长的路要走。

正如我提到的，量子计算和人工智能是两大热门方向，一个非常鼓舞人心的问题是，随着人工智能和量子计算机的发展，是否有机会将将量子计算和人工智能结合起来。这样也许就可以利用量子算法，来理解或创造超自然智能。如果我们可以做到，那我们就可以成功建造出自然界都无法创造的超智能系统。

我不知道这有多大的可能性可以成功，也许我们需要仰望星空来获取灵感，来帮助我们保持谦虚并不断的提高自己，谢谢！

英文版

And now I’m going to have a very succinct survey on the experimental side.And so, the leading technology for quantum computing, initially they were a dozen or twenty possible technologies for implementing the qubits and the processing unitsand then it seems that after a number of years they are some of the more promising ones.

I think that if you look at scientific papers, I think up to twenty bits. It’s fairly solid and with very high reliability. And now there are some announcement of some prototype in the making by IBM and Google that would go substantially toward the 100s-qubit mark. And if you really can do a good job, you will start to have some demonstrably very powerful ability. And the superconducting one, I think it's the one received the most attention in the public eye.

But the ion trap is also a technology that's fairly mature, and I think that they are both contenders. There are some newer types of technologies, and the one that I’m particularly impressed with is to use diamonds to do computers. Basically is the vanity factor that I would like to see quantum computers made of diamonds so that I can put it on my desk to show it to everyone.

And there are photonic computations that we can do the boson sampling and up to ten entangled (qubits). And by the way, this last piece of work was done in China, in the University of Science and Technology under professor Pan’s leadership.

People are now seem to be much more optimistic than before, and there are many talented, quantum physicists are working on building up the components of the quantum computers, and I’m just going to mention a few that come from our institute,their progress using diamond. I mean all these technologies -- they all have strengths and they have weaknesses. So they are not really saying that one is dominating the other.

So for example, the diamonds, I think right now the number of bits you can design and manufacture is smaller than the ion trap and the superconducting. But the diamonds have the property that can work in room temperature and it's solid. So therefore it makes it kind of remind us of the silicon. And so it has potential for development in the future.

And now the ion trap. First you have to you kind of produce the ions, and then you have the magnetic field to stabilize them and the line them up. And so they can work with qubits. And the ion trap, it’s good in some respect. It's very stable. The coherence time is now ten minutes. And so now I’m coming to the conclusion.

And the first one is that, I think that quantum computing is very exciting from the intellectual point of view aside from the utilitarian point of view. If we think about the quantum computing and classical computing that the classical computing the design principles are simple, logical and common sense.

And actually, I think the design principles can be appreciated, even by ancient scientists like Euclid, Archimedes, certainly Da Vinci. When come to think of it, perhaps the first scientific computer was not proposed by Turing, but actually by Euclid，I think the ancient Greeks. if you think that you can do geometry by using rulers and compasses, you can ask the question, what are the geometric things you can compute?

That really is a very beautiful system of computational theory. You can add, subtract, multiply and divide and actually take square roots. Now quantum computing is very different. It's based on the twentieth century modern physics. And its design principles are pretty opaque and are very non-intuitive. It's exciting intellectually, because in a sense, we are getting a second opportunity as Turing.

I think that the time of Turing and Church and Kleanie, they were -- those were great, big times. Kurt Gödel just published the incompleteness theorem and what is a computation becomes a question that seems to be in the air. And it's a great time for researchers to think about what's the meaning of computation – I mean such a huge concept.

The first time being doing research on. So it's immensely interesting it's like going to a virgin forest where nobody has been there before: any progress you can make is going to be important.

And now quantum computing gives us a second coming. So this is a second chance. And when you think about what can be done by quantum computers, you're going back to the drawing board. because potentially any physical device, any quantum device kind of experiment, process, is potentially a way for you to utilize and build quantum computers.

So that's why it's intellectually very -- as you can see there's immense possibilities -- just from the example I mentioned to you, you can think about the optical x-ray computer. And then you can think about how to utilize it to do extra computations, and what do you get. And so there is a very big intellectual space to explore.

And the second conclusion is that the progress in quantum computing is, I think the last fifteen or twenty years, I think it's remarkable. And now we are really talking about there are companies are making prototypes of up to 100 qubits. We are sort of into a mixture of the physics experiment and plus the engineering. Because the engineering aspect, say, when you're -- if you have to build superconducting quantum computers, you need a lot of electrical engineer to help you in order to do it.So engineering is starting to come into play.

Whenever engineers come into building something, I have immense hope and because I think that's a good sign. And another evidence for being optimistic is that there are many many countries that have put a lot of research money into quantum computing.

And even more tellingly, many of the leading IT companies are putting up a big quantum computing programs, so you can see that these are players who, I think, typically have good eyes.

And so it must be that they are also sensing that there is something on the verge of fruitful outcome. So, but we have to temper this with the caveat that building a really practical quantum computer, even though it could be entering the final stretch, but however it will be a really really tedious and difficult last mile. 

And finally, I’m going to go above quantum computing because you know that kind of computing has its own logic now that is going to happen sooner or later. And but we would like to lift ourselves a little bit more and intellectually and think about what is the position and the impact of quantum computing in the future of information science.

And in my view, among other things, I think nature has two really outstanding achievements.

One is that nature designs a very intricate law, namely the quantum laws whose consequences are so difficult to figure out analytically, making the world so interesting with many intriguing phenomena. And the second thing is that the nature, through evolution, has created a species (the humankind) whose brain is among the most complex in the universe, has really really amazing power to create and to make inferences.

And now let's see that how well we are on our way to sort of try to catch up with nature. If we are successful in building a quantum computer with a very large scale, then it means that that we finally have the ability to solve quantum equations. And it means that the physical laws as we know it -- we can actually also create things by making use of that knowledge, like nature.

So I think that we are entering the phase where we can hope certainly in the rest of the century, we will be able to make great progress. And what about the other one? Are we able to create intelligences that have the deviance ability like human minds? And now, actually there are computer scientists working hard on this.

I think it's called artificial intelligence. And so, I mean it's not exactly the brain’s intelligence, but is it possible for us to build up systems that would match the human brain? And so if in the next -- in the rest of the century, we make tremendous progress, then it'll be really nice that we also can claim ourselves that we can make things that match the nature in that regard. And we are making some progress on it, but we are still far from it.

And now, I mentioned that quantum computing and AI are two of the hottest trends there. And the question that really is inspiring for us to ask is that is there a chance through the progress of AI and quantum computing, if we can combine quantum and AI together so that using quantum algorithms to understand or to create new intelligence, maybe we can go beyond nature. 

If we can do that, then we have succeeded in building some super-intelligent systems

that nature has so far been unable to create. And I don't know how much that will be successful, but I think that is always good for us to have ideals so high up in the sky 

that we will keep ourselves humble and strive to improve ourselves.

Thank you very much.

🔽点击查看往期内容

想要和更多志同道合的人一同讨论科普问题，获得墨子沙龙小秘书贴心服务，及时获取各类科普活动通知吗？扫码加墨子沙龙小秘书，拉你入墨子沙龙超大群。

墨子沙龙是由中国科学技术大学上海研究院主办、上海市浦东新区科学技术协会及中国科大新创校友基金会协办的公益性大型科普论坛。沙龙的科普对象为对科学有浓厚兴趣、热爱科普的普通民众，力图打造具有中学生学力便可以了解当下全球最尖端科学资讯的科普讲坛。

关于“墨子沙龙”