The promise of quantum computers | Matt Langione

Transcriber:

翻译人员: 思辰 李

21世纪最大的挑战会是什么？

如今，人们可能会猜测 气候变化、公共健康、资源不平等。

What will the biggest challenges of the 21st century turn out to be?

实际上，我们还无法下定论。

我们所知道的是，

Today, one might guess climate change, public health, inequality,

超级计算必将是解决方案的一部分。

but the truth is we don't yet know.

近一百年来，

越是在面临最危机的时刻，

What we do know

is that supercomputing will have to be part of the solution.

我们对高性能计算机

越是依赖。

For nearly a hundred years,

从破译纳粹密码到人类基因测序。

our reliance on high-performance computers

in the face of our most urgent challenges

年复一年，像施了魔法一般，

has grown and grown,

from cracking Nazi codes to sequencing the human genome.

计算机处理器 变得更小、更快、更好，

Computer processors have risen to meet increasingly critical and complex demands

以满足愈加挑剔、复杂的需求。

但有一个问题，

by getting smaller, faster and better year after year,

正当我们对计算机的依赖 前所未有地增长时，

as if by magic.

运算能力的提升却止步不前。

But there's a problem.

At the very moment that our reliance on computers is growing faster than ever,

这股魔法好像用尽了，

还是在最糟糕的时机下。

progress in compute power

我们很少谈论这个话题： 尽管我们用计算机实现了许多成就，

is coming to a standstill.

The magic is just about spent.

仍有相当多的事 是计算机无法完成的，

The timing couldn't be worse.

We rarely talk about it, but for all that we've accomplished with computers,

而这对商业和社会来说代价巨大。

there remain a startling number of things that computers still can't do,

举个例子：50年前， 人们曾设想要实现

可以即时设计出药物的计算机， 但至今未果。

at a great cost to business and society.

The dream of near-instant computational drug design, for instance,

这一困境从未如此显而易见，

因为现在的世界 正处于隔离与瘫痪的状态，

has yet to come to fruition, nearly 50 years after it was first conceived.

而我们还在苦苦等新冠疫苗的出现。

Never has that been clearer than now,

但是药物发现仅仅是 研究者们受困的一个领域，

as the world sits in a state of isolation and paralysis,

as we await a vaccine for COVID-19.

在其他一些领域 研究则可能完全停滞，

即便是现在最快的超级电脑 也无法解决问题。

But drug discovery is just one area in which researchers are beset --

这就给例如气候变化、

and in some cases, blocked entirely --

by the inadequacy of even today's fastest supercomputers,

金融与物流等价值创造领域 带来了很大的限制。

putting great constraints in areas like climate change,

过去，

随着部件每年变得越来越小，

and in value creation in areas like finance and logistics.

我们可以单纯地期待 超级电脑能变得更快更好。

In the past,

可现在却不行了。

we could rely on supercomputers simply getting better and faster,

我们正面临着物理上的极限。

as parts got smaller and smaller every year,

晶体管已经小到几乎与原子同体积。

but no longer.

For now, we're drawing up against a hard physical limit.

事态如此发展 会带来一个后续的问题。

Transistors have become so minuscule

这个问题也是过去几年

that they're fast approaching the size of an atom.

我鼓励商业领导 和政策制定者解决的问题：

Such a state of affairs invites a natural follow-up question,

如果没有基本意义上的超级计算机，

and it's one that I've spent the last several years

那么帮助我们应对 21 世纪挑战的

encouraging business leaders and policy makers to address:

将会是哪种科技？

If not traditional supercomputers,

答案是：量子计算。

what technology will emerge

类似我身边这台的量子计算机，

to arm us against the challenges of the 21st century?

通过利用百年前人类尚未知晓的 亚原子物理属性

Enter quantum computing.

Quantum computers like this one

来突破原子的限制。

promise to address the atomic limitation

量子计算是如何工作的呢？

量子计算可以摆脱

by exploiting subatomic physical properties

that weren't even known to man a hundred years ago.

传统半导体计算的两个主要限制。

传统计算机 运行的方式是既定式的：

How does it work?

Quantum computing enables a departure from two major constraints

一切状态要么是、要么不是， 要么开、要么关，没有中间状态。

of classical semiconductor computing.

此外，传统计算机是按顺序工作的： 它们一次只能做一件事。

Classical computers operate deterministically:

everything is either yes or no, on or off, with no in between.

但量子计算机 则基于概率的模型运作，

而且它可以同时处理多项任务。

They also operate serially; they can only do one thing at a time.

Quantum computers operate probabilistically, and most importantly,

这得益于它的三个特性： 叠加、纠缠和干涉，

they operate simultaneously, thanks to three properties --

它们才能够一次探索多种可能。

为了理解它的工作原理， 请想象一台电脑正在尝试走迷宫。

superposition, entanglement and interference --

which allow them to explore many possibilities at once.

传统电脑将会按顺序

To illustrate how this works, imagine a computer is trying to solve a maze.

尝试每一种可能的解法。

如果第一种方法行不通，

它就会排除这种解法，

The classical computer would do so by exhausting every potential pathway

走回到初始位置，

再尝试其他解法。

in a sequence.

If it came across a roadblock on the first path,

不停重复上述动作， 直至达到终点。

it would simply rule that out as a solution,

量子计算机则会 同时测算每一种可能的解法。

revert to its original position

and try the next logical path,

and so on and so forth until it found the right solution.

只要试一次，问题就可以解决了。

A quantum computer could test every single pathway

at the same time,

事实上，

很多复杂的问题都像走迷宫，

in effect, solving the maze in only a single try.

尤其是模拟和优化问题。

As it happens,

这些问题可以通过量子计算机 更快速地解决。

many complex problems are characterized by this mazelike quality,

especially simulation and optimization problems,

但所谓的量子加速真的有价值吗？

some of which can be solved exponentially faster

with a quantum computer.

为了确信 我们需要更快的超级计算机，

But is there really value to this so-called quantum speedup?

我们首先需要确定 我们的问题确实具有计算性质。

In order to believe that we need faster supercomputers,

事实证明，许多问题多少都是的。

we need to first believe that our problems are indeed computational in nature.

举个例子， 让我们来看看化肥的生产，

这是气候变化科学中的 一个标志性问题。

It turns out that many are, at least in part.

如今，大多数化肥的生产方式 是将氮气和氢气融合，

For an example, let’s turn to fertilizer production,

one of the hallmark problems in the science of climate change.

去制成作为有效成分的氨气。

The way most fertilizer is produced today

这个过程是可行的，

但却伴随着严重的代价。

is by fusing nitrogen and hydrogen to make ammonia,

对于企业来说，一年的开销 会达到 1 千亿至 3 千亿。

which is the active ingredient.

The process works,

而对于环境影响而言，

but only at a severe, a severe cost

每年世界 3% 至 5% 的天然气

to businesses, who spend 100 to 300 billion every year,

都被用于肥料合成。

and to the environment:

three to five percent of the world's natural gas

为什么科学家们 未能开发出一种更有效的工艺呢？

is expended on fertilizer synthesis

原因在于，为了做到这一点，

every single year.

So why have scientists failed to develop a more efficient process?

他们需要模拟构成作为关键催化剂的

氮素酶的静电场中 迷宫般的分子的相互作用。

The reason is that in order to do so,

they would need to simulate the mazelike molecular interactions

科学家们实际上知道 如何做到这一点，

that make up the electrostatic field of the key catalyst, nitrogenase.

但这目前需要世界最快的超级计算机 花上 80 万年运算出来。

Scientists actually know how to do that today,

而用一台满配置的量子计算机， 则只需要不到 24 小时。

but it would take 800,000 years on the world's fastest supercomputer.

另一个例子， 让我们回到药物发现领域，

新冠让我们大多数人 第一次认识到这一领域。

With a full-scale quantum computer,

less than 24 hours.

For another example, let's return to drug discovery,

设计一种针对新冠 这样的传染性疾病的疫苗，

a process COVID-19 has brought into sharp focus for most of us

从确定疾病的驱动因素，

for the very first time.

Designing a vaccine for an infectious disease like COVID-19,

到筛选数百万个 候选用的激活剂和抑制剂，

每一种药物 就要花10年或更长时间，

from identifying the drivers of the disease,

to screening millions of candidate activators and inhibitors,

而其中 90% 的药物 无法通过临床试验。

is a process that typically takes 10 or more years per drug,

对制药公司来说，

每种批准的药物的成本 是 20 亿至 30 亿美元。

90 percent of which fail to pass clinical trials.

但开发延期和失败 只会带来更高的社会成本。

The cost to pharmaceutical companies is two to three billion dollars

每年有超过 800 万人 死于传染病。

per approved drug.

But the social costs of delays and failures are much, much higher.

这是新冠大流行的头 6 个月 的死亡人数的 15 倍。

More than eight million people die every year of infectious diseases.

那为什么计算机驱动的药物设计 没有达到预期指标呢？

That's 15 times as many people as died during the first six months

原因同样在于计算资源是有限的。

of the coronavirus pandemic.

So why has computational drug design failed to live up to expectations?

如果说解析身体中的疾病 就像解开一个锁，

Again, it's a matter of limited computational resources, at least in part.

那么为其设计药物 就好像在茫茫一片化学空间中搜寻，

If identifying a disease pathway in the body is like a lock,

就好像是走一座分子结构的迷宫，

designing a drug requires searching through a massive chemical space,

从而找到适合的化合物，

相当于找到开锁的钥匙。

effectively a maze of molecular structures,

问题在于， 追踪这整个化学空间，

to find the right compound,

to find a key, in other words, that fits the lock.

并将其转换为可供搜索的 药物设计数据库，

The problem is that tracing the entire relevant span of chemical space

需要耗费 5 万亿亿亿年，

and converting it into a searchable database for drug design

还得在世界上最快的 超级计算机上完成。

would take 5 trillion trillion, trillion, trillion years

而在量子计算机上， 只需要半个多小时。

量子计算的胜利 不仅仅止于实验室。

on the world's fastest supercomputer.

各种行业的进步和发展

On a quantum computer, a little more than a half hour.

But quantum computing is not just about triumphs in the lab.

都受到了零散但棘手的 计算限制的阻碍，

The flow of progress and industries of all kinds is currently blocked

对企业和社会产生了不小的影响。

银行是一个出乎意料的例子。

by discreet, but intractable computational constraints

如果银行能够更自由地向个人、企业 和企业家提供贷款会怎么样呢？

that have a real impact on business and society.

For what may seem an unlikely example, let's turn to banks.

如今的主要阻碍之一

What if banks were able to lend more freely

是银行将 10 %至15 %的 资产留作现金储备，

to individuals, entrepreneurs and businesses?

一定程度上是因为 很难通过计算模拟投资风险。

One of the key holdups today

is that banks keep 10 to 15 percent of assets in cash reserves,

银行无法考虑到 全球或整个市场的风险状况，

in part, because their risk simulations are compute constrained.

极其严重的风险虽然罕见， 但也的确不可预测。

黑天鹅事件就是一个例子。

They can’t account for global or whole-market risks that are rare

10% 到 15% 的确是一大笔钱。

but severe and unpredictable.

尤其是考虑到 每降低 1 %的现金储备，

Black swan events, for example.

Now, 10 to 15 percent is a whole lot of money.

就意味着额外 1 万亿美元的 可投资资产。

When you consider that for every one-percent reduction in cash reserves,

这意味着， 如果银行能变得足够自信，

it would lead to an extra trillion dollars of investible capital.

他们可以通过量子计算来模拟风险，

从而将现金储备 降至资产的 5 %至 10 %，

What this means is that if banks ultimately became comfortable enough

这将会每年为个人和企业

with quantum-powered risk simulations

提供新冠时期一样的补贴。

to reduce cash reserves to, say, five to 10 percent of assets,

the effect would be like a COVID-19-level stimulus

一旦量子计算 可能带来的变革变得明朗，

for individuals and businesses

就只剩一个问题： 我们到底要等多久？

every single year.

Once the transformative power of quantum computing is clear,

研究人员对于量子运用 启动的时间有着谨慎的估计。

the question then becomes: Well, how long must we wait?

这么做是有道理的， 毕竟还有一些关键的障碍需要克服。

Researchers are cautious when asked about the timeline to quantum advantage.

也不仅仅是工程方面的挑战，

还有关于量子力学 基本性质的科学问题。

Rightly so -- there remain a number of critical hurdles to overcome,

and not just engineering challenges,

因此，在量子计算完全成熟之前，

but fundamental scientific questions about the nature of quantum mechanics.

可能还需要 10 、 20 、 甚至 30 年的时间。

基于这点， 与我沟通过的一些高管决定

As a result, it may be one, two, even three decades

他们等得起，

before quantum computers fully mature.

他们打算推迟投资。

我觉得这种想法很错误。

Some executives that I've spoken with have come to the conclusion on this basis

因为虽然有些技术 得通过累计的规律稳步发展，

that they can afford to wait,

that they can afford to postpone investing.

I believe this to be a real mistake,

但许多技术的出现 几乎是一夜之间的突破，

for while some technologies develop steadily,

完全无视了任何过去的预测。

according to the laws of cumulative causation,

量子计算很有可能 迎来这样一个突破，

many emerge as precipitous breakthroughs,

almost overnight defying any timeline that could be drawn out in advance.

它达到一些里程碑式的节点的时间

已经比计划提前了几十年。

Quantum computing is a candidate for just such a breakthrough,

在80年代末，

许多研究人员认为， 量子计算的基本构件，

having already reached a number of critical milestones

decades ahead of schedule.

即量子比特，

需要 100 年的时间 才能造出来。

In the late 80s, for example,

many researchers thought that the basic building block of quantum computing,

但仅仅过了 10 年，它就来了。

现在，IBM（国际商用机器公司） 在 29 台机器上

the qubit,

would take a hundred years to build.

拥有近 500 个量子比特， 可供客户使用和研究。

Ten years later, it arrived.

Now IBM has nearly 500 qubits across 29 machines,

这意味着，我们不应该担心 量子计算机来得太晚，

available for client use and research.

而应该担心它们来得太早，

What this means is that we should worry less

而我们却来不及完成 必要的准备工作。

about quantum computers arriving too late

一位获得诺贝尔奖的物理学家说过：

and more about them arriving too soon,

“量子计算机与目前的计算机的区别

before the necessary preparations have been made.

For to quote one Nobel prize winning physicist,

比目前的计算机 与算盘的区别更大。”

"Quantum computers are more different from current computers

我们需要时间 来整合必要的工作流程。

than current computers are

我们需要时间来寻找合适的人才。

from the abacus."

重要的是，

It'll take time to make the necessary workflow integrations.

我们需要时间，远见和想象力，

It'll take time to onboard the right talent.

来确认并界定量子计算机

Most importantly,

it'll take time, not to mention vision and imagination,

能为你的企业解决的 高价值问题有哪些。

各国政府已经在对量子技术 进行大量投资

to identify and scope high-value problems

中国、欧洲和美国的投资额 总高达 150 亿美元。

for quantum computers to tackle for your business.

Governments are already investing heavily in quantum technologies --

而风险投资公司们也在效仿。

但现在需要的 用来加速变革的投资

15 billion dollars among China, Europe and the US.

该用于开发案例、培养人才、

And VCs are following suit.

But what's needed now to accelerate innovation

并实验如今真实可用量子计算机中。

is business investment in developing use cases,

在今天这样的世界里，

in onboarding talent

创新的需求不能再被推迟了。

and on experimenting with real quantum computers

领导者必须现在就采取行动，

that are available today.

我们推动了近 70 年 不断革新的处理器加速

In a world such as ours,

the demands of innovation can't be put off for another day.

终将停滞不前。

Leaders must act now,

而迈向魔法和超级计算新时代的竞赛 已经在进行中。

for the processor speedups that have driven innovation for nearly 70 years

这是一场我们输不起的比赛。

are set to stop dead in their tracks.

The race toward a new age of magic and supercomputing is already underway.

量子计算机正领先于杆位， 等待着我们去突破。

It's one we can't afford to lose.

谢谢！

Quantum computers are in pole position. They're the car to beat.

Thank you.