Can AI Help Solve the Climate Crisis? | Sims Witherspoon | TED

翻译人员: jing lin 校对人员: suya f.

You may have had the experience of unboxing furniture

你应该有家具拆箱的经验吧，

and come across instructions that go something like this:

你会看到这样的安装说明：

"Assemble the bookshelf according to the provided diagram."

“请按照示意图组装书架。”

Yes, I know what a bookshelf looks like.

是的，我知道书架是什么样子。

Probably wouldn't be reading the assembly instructions

但如果不是因为在这个 过程中我需要一些帮助，

if I didn't need a little more help with the process.

我大概不会去阅读组装说明。

Or maybe you've opened a cookbook

或者当你打开一本食谱，

with an author who thinks you're already somewhat of a chef.

食谱的作者似乎默认 你已经有点厨师的基础了。

"Deglaze the pan."

“给锅除釉。”

What?

这是什么意思？

(Laughter)

（笑声）

OK, off I go on a separate search to understand whatever that means.

好吧，我只好开始另外的搜索 来弄清楚这是什么意思。

Instructions that tell you what to do and not how to do it

一个操作说明仅 告诉你“该做什么”，

而不是“怎么做”

are pretty useless.

它是毫无用处的。

And yet, even when we're talking

但是，当我们在谈论

about something as important as climate change,

像气候变化这样重要的事情时，

we hear them all the time.

也总能听见：

“Transition to renewable energy.”

“向可再生能源转型。”

“Electrify everything else.”

“让一切电气化。”

“Deploy solutions that are equitable and fair.”

“要部署公平、公正的解决方案。”

Yes, let's do all of that.

现在，让我们一起来做这些事。

But how?

但是，怎么做呢？

Answering how is where we understand which solutions are actually feasible,

要回答怎么做，就要了解 哪些方案是切实可行的，

whether that be with today's infrastructure,

无论是当今的基础设施、

our evolving regulatory environment

不断变化的监管环境、

or any of the other number of dependencies and constraints

还是我们必须考虑的

that we have to consider.

任何其他依存和限制条件。

How we solve climate change

我们如何解决气候变化问题

also depends on our very definition of the problem.

还取决于我们对这个问题的定义。

It's a scientific challenge, a sociopolitical issue,

这是一个科学挑战、 一个社会政治问题、

an economic problem and so much more.

也是一个经济问题，等等。

And how we solve it will depend on how we frame it.

而我们如何解决它 将取决于我们如何构建它。

There is no single answer.

没有单一的答案。

I'm a scientist,

我是一名科学家，

so I approach climate change as a scientific challenge.

所以我把气候变化 当作一项科学挑战来对待。

I'm also a techno-optimist

我也是一名技术乐观主义者

and artificial-intelligence product manager,

和人工智能产品经理，

so I also approach it as a technological one.

所以我也把它当作一项技术挑战。

When it comes to a sustainable future,

谈到可持续的未来，

artificial intelligence can help us do three critical things.

人工智能可以帮助我们 做三件关键的事情。

First, it can help us understand climate change

首先，它可以帮助我们了解气候变化

and its effects on Earth's ecosystems.

及其它对地球生态系统的影响。

Second, it can help us optimize current systems and infrastructure,

其次，它可以帮助我们优化 当前的系统和基础架构，

because we can't just start over from scratch today.

因为我们今天不可能从零开始。

And third, it can help us accelerate the breakthrough science we need,

第三，它可以帮助我们加速科学的突破，

such as fusion as a carbon-free energy source.

例如核聚变作为无碳能源。

Today, I'd like to talk about that second one,

今天，我想谈谈第二个问题：

optimizing current systems,

即优化当前的系统，

and specifically, how we can use AI to harness a superpower

特别是人工智能如何帮助我们利用

we already have in this fight:

我们在这场斗争中已经拥有的超级能量：

wind energy.

风能。

Renewables are unquestionably a key to a sustainable future,

毫无疑问，可再生能源 是可持续未来的关键，

but the problem is they're unpredictable.

但问题在于它们是不可预测的。

Sometimes, the sun shines and the wind blows,

有些时候，阳光普照、风力强劲，

and sometimes, it just doesn't.

但另一些时候，就并非如此。

Now, for an electricity systems operator,

对于电力系统运营商来说

who needs supply to meet demand in real time, 24-7,

他需要能够满足全天 候实时需求的电力供应

this is hugely problematic.

那这就是个大问题。

Renewables can't be 100 percent reliably scheduled.

可再生能源无法被百分之 百可靠地预先安排。

Now, unfortunately, fossil-fuel plants are the opposite.

不幸的是， 化石燃料发电厂却恰恰相反。

You can burn a specific amount of coal at a set time

你可以在设定的时间 燃烧特定数量的煤炭，

to deliver exactly the amount of electricity you want

就能在可预测的时间范围内

in a predictable time window.

准确提供你想要的电量。

So ...

所以...

if you're a power systems manager

如果你是一名电力系统经理，

whose job is to literally keep the lights on,

你的工作就是要保证电灯能亮起来，

which source are you more confident depending on?

那么你更有信心依赖哪个来源？

But here's one of the places where AI can come in.

这就是人工智能可以 发挥作用的地方之一。

It is a powerful tool for forecasting.

它是一个强大的预测工具。

AI systems can ingest vast amounts of historical data

人工智能系统可以摄取 大量的历史数据，

and help us predict future events.

帮助我们预测未来的事件。

So, while we can't eliminate the variability of wind,

虽然我们无法消除风的多变性

we can use AI to more accurately predict its availability.

但我们可以用人工智能来 更准确地预测其可用性。

That was my team’s “what” to do.

那就是我的团队要正在做的事。

Use AI to accelerate the transition to renewables, like wind energy.

使用人工智能加速向 可再生能源（如风能）的转变。

The tough part was the “how” to do it.

困难的部分是 “如何” 去做。

First, we researched the challenge.

首先，我们仔细研究了这个问题。

We read papers, we spoke to domain experts,

我们研读论文、 与领域专家进行交流，

we found out everything we could about the problem.

找到我们所能找到 的关于这个问题的一切。

Our team, which is a mix of research scientists,

我们的团队由研究型科学家、

engineers, a product manager, a program manager

工程师、产品经理、项目经理

and an impact analyst,

和影响分析师组成。

decided that a neural net trained on historical weather data

我们认为，用历史天气数据 和涡轮发电量信息

and turbine power-production information

训练过的神经网络

would likely help us accomplish our goal.

很有可能会帮助我们实现目标。

Next, we needed to find two core elements:

接下来，我们需要找 到两个核心要素：

data to train the system

用于训练系统的数据

and a partner who was willing to deploy it.

和愿意部署系统的合作伙伴。

Both of these can be major obstacles

在现实场景中部署人工智能时，

when it comes to deploying AI in real-world scenarios.

这两者都可能成为主要障碍。

Let's start with data.

让我们从数据开始。

There are massive gaps in climate-critical data --

有关气候的关键数据存在巨大缺口，

not just in electricity,

这缺口不仅在电力领域，

but in agriculture, transportation, industry and many other sectors.

还在农业、交通、 工业和许多其他领域。

Some of our data, we could purchase or download for free --

有一些数据我们可以免费购买或下载，

weather forecasts, for instance.

例如天气预报。

But some of the data we needed was proprietary,

但是我们需要的一些数据是专属的，

and this would be, like, turbine power-production information

比如风力发电场的涡轮发电信息

and other operational data from the wind farms.

和其他运营数据。

Now, we needed that proprietary data so that we could train our models

我们需要这些专有数据， 这样才能训练人工智能模型

to learn the relationship between historical weather

了解历史天气

and historical power production,

与历史发电量之间的关系，

so it could then then make predictions about future power availability

这样它才能根据天气预报的数据

based on what data said about future weather.

预测未来的电力供应情况。

Now it's probably worth mentioning here

还需要提及的一点是：

that we were looking at a few years of data

我们查看的是这几年的

on hourly resolution,

每小时数据

not historical data at a timescale

而不是与目前气候有巨大差异的

that would have massive climactic differences from present day.

的历史时期的数据。

In addition to data,

除了数据之外，

we needed to find a partner with domain expertise

我们还要找到既具有领域专业知识、

and the willingness and scale to test new systems.

又具备测试新系统 的意愿和规模的合作伙伴。

You know, surprisingly, this can be a major hurdle

令人惊讶的是， 在现实世界中部署人工智能，

when it comes to deploying AI in the real world.

这可能是一个主要障碍。

Believe it or not, it's not every wind-farm manager

不管你信不信，并不是 每个风电场经理都想

that wants to let a bunch of AI researchers

让一群人工智能研究人员

test on their multimillion- or multibillion-dollar systems.

在价值数百万或数十亿美元 的系统上进行测试。

But the thing is, in order to prove that AI works,

但问题是， 为了证明人工智能行之有效，

we have to have deployment opportunities in the real world.

我们必须在现实世界中拥有测试机会。

Luckily for us, Google was a ready and willing partner.

幸运的是，Google就是一个 现成且有意愿的合作伙伴。

OK, yes, DeepMind is a part of Google,

是的，DeepMind是谷歌的一部分，

but it's not a given that they would let us test on their systems.

但可以让我们在他们的系统上 进行测试并不是理所当然的。

Yet they let us test on 700 megawatts of their wind-power capacity,

他们让我们测试了 700兆瓦的风力发电量，

which is equivalent to a large wind farm in the United States.

这相当于美国的一个大型风力发电场。

This made them an excellent proxy for external wind-farm operators.

这使它们成为外部风力 发电场运营商的绝佳代理。

They also lent us an expert team to advise on metrics and benchmarks

他们还借给我们一个专家团队， 就指标和基准提出建议，

and to share the data that we needed.

并分享我们需要的数据。

This is another critical component of AI for the real-world deployments.

这是人工智能能在现实世界中 使用的另一个关键因素。

Working with a domain-expert team that can tell you what they need,

与领域内专家团队合作， 才能知道他们需要什么，

how they need it to work,

需要人工智能如何工作，

which constraints keep the system safe,

哪些限制条件可以保证系统的安全、

what quantifiable metrics to use to measure AI performance

用哪些可量化的指标 来衡量人工智能性能、

and how much better that AI performance needs to be

以及人工智能的性能需要

than their previous systems

比他们的前一代提升多少，

to make the cost of switching over even worth it.

才能使切换的成本物有所值。

And that's just to name a few.

这只是其中的几个例子。

So at this point, we have our idea,

现在这个时候，

we have our data, we have our deployment partner.

我们具备了想法、数据、合作伙伴。

Now, to test and deploy our system.

现在，来测试和部署我们的系统。

Improving the accuracy of electricity-supply forecast

提高电力供应预测的准确性

is incredibly important.

极为重要。

If predictions are higher than actual generation,

如果预测量高于实际发电量，

renewable electricity managers may not have enough supply to meet demand.

可再生电力管理公司可能 没有足够的供应来满足需求。

This, in turn, drives the purchase of carbon-intensive fossil fuels

这反过来又推动购买碳密集型化石燃料

to cover that gap,

以弥补这一缺口，

because they're largely what makes up backup generation.

因为它们在很大 程度上构成了备用发电量。

Now, the good news.

接下来是好消息。

Our AI system performed 20 percent better than Google's existing systems.

我们人工智能系统的准确度 比谷歌的现有系统高20％。

Even better news is that Google decided to scale this technology.

更好的消息是，谷歌决定扩大 应用这项技术的规模。

And scaling is so important.

扩大规模非常重要。

We will run out of time in the climate countdown

我们在气候倒计时中的时间越来越少了，

if we aren't deploying solutions that are widely applicable.

我们需要部署广泛适用的解决方案 。

This particular solution is being developed

这种特殊的解决方案

into a software product

正在开发成一款软件产品。

that French company Engie is among the first to pilot.

法国Engie公司是最早 试用的公司之一。

But, you know,

但你知道的，

it doesn't even take a major research organization to do this kind of work.

它甚至不需要大型研究机构 就能完成这样的工作。

Where we focused on AI for supply-side forecasting,

我们的人工智能专注于“供应侧预测”

a small UK-based nonprofit called Open Climate Fix

一个叫Open Climate Fix的英国小型非营利组织

is focusing on AI for demand-side forecasting.

则专注于“需求侧预测”的人工智能。

They found a willing partner in the UK National Grid,

他们在英国国家电网 中找到了合作伙伴，

and are currently deploying forecasts that are two times more accurate

并且目前测试中的预测准确度

than the UK grid's previously used systems.

是英国电网之前使用系统的两倍。

Now, all of this is to say is that AI can help us

现在的所有这些都告诉我们

with the transition to renewable energy,

人工智能是可以协助向可再生能源的转变，

but scientists and technologists,

但仅靠科学家和技术专家，

we're not going to be able to do that alone.

是无法单独做到这一点。

We need to be working with partners and experts

我们需要与能够教我们

who can teach us the “how.”

“如何做” 的合作伙伴和专家合作。

So for those of you interested in this space,

因此，对于那些对此感兴趣的人，

if you're a domain expert,

如果您是领域内的专家，

please share the problems you face and the challenges that you have

请分享您面临的问题和挑战，

so that our sector can ensure

以便我们能够实现

that AI pursuits will have impact in the real world

人工智能的发展对现实世界中产生影响，

and not be purely academic.

而不仅仅是学术性的。

Even better,

更好的是，

if you want to incentivize ML researchers to work on your problems,

如果你想激励 机器学习研究人员研究你的问题，

I'll let you in on a little secret:

我会告诉你一个小秘密：

build a competition, and they will come.

举办一场竞赛，他们就会来的。

(Laughter)

（笑声）

It's true.

这是真的。

Just don't forget the datasets and metrics.

只是不要忘记数据集和指标。

If you are a data holder, where it’s safe and responsible to do so,

如果您是数据持有者，并且 可以安全且负责任地这样做，

please share data related to those challenges.

请分享与这些难题相关的数据。

If you're not sure whether the data you have is even climate-critical,

如果你不确定自己拥有的数据 是否对气候至关重要，

you can check out Climate Change AI's website,

你可以访问

“Climate Change AI”网站，

where they have published a wish list of climate-critical datasets.

那里发布了“气候关键 数据集”的愿望清单。

Access to these datasets

获得这些数据集

would unblock crucial research and innovation in AI for climate.

将解锁人工智能气候 领域的关键研究和创新。

If you're a deployment partner,

如果您是可部署系统的合作伙伴，

please, let us know who you are,

请告诉我们您的身份，

especially if you're willing to test innovative systems.

特别是如果您愿意测试创新系统。

And for everyone who's interested in this space,

对于所有对这个话题感兴趣的人，

please know you do not have to be technical to work in tech.

请知道，从事科技相关工作 并不必须是技术人员。

AI for climate action requires a variety of skill sets

用于气候行动的人工智能 需要各种各样的技能组合

and a diversity of backgrounds

和不同的背景，

that, yes, includes research scientists and engineers,

是的，包括研究科学家和工程师，

but it also includes ethicists and policy experts,

但也包括伦理学家和政策专家、

communication teams, product managers,

沟通团队、产品经理、

program managers and so many more folks.

项目经理等等。

Now for the warning label.

现在，警示标签来了。

AI is not a silver bullet.

人工智能不是灵丹妙药。

It will not solve all problems driving climate change.

它无法解决推动气候变化的所有问题。

It isn't even the right tool for many of the challenges that we face.

它甚至不是应对我们面临 的许多挑战的正确工具。

AI is also not a technology without tensions.

人工智能也不是一项 不会带来紧张关系的技术。

It needs to be deployed safely and responsibly.

它需要安全、负责任地部署。

Not to mention,

更不用说在我们的 电网使用清洁能源之前，

until our grids are run on clean energy, AI itself will carry a carbon footprint,

人工智能本身将产生碳足迹，

as will any energy-intensive technology we use.

正如我们使用的任何能源密集型技术。

But AI can be a transformational tool in our fight against climate change --

但是人工智能可以成为我们 应对气候变化的变革性工具，

it's just on all of us to wield it effectively.

我们所有人都有责任有效地使用它。

The “why” we need to is absolutely harrowing.

我们使用它的原因令人痛苦，

The “what” we can do is really exciting.

但我们能做的事情却令人兴奋。

But it’s the “how” we can do it

正是我们 “如何做”的方法

that will illuminate feasibility and help us drive impact.

将会证明它的可行性， 并帮助我们推动它的影响力。

So, in your next climate action conversations,

因此，在你的下一次气候行动对话中，

when someone presents you with an exciting "what,"

当有人向你介绍一个 激动人心的“做什么” 时，

please help to advance the conversation

请帮助将对话推进

to the impactful "how."

到有真正影响力的 “如何做”。

Thank you.

谢谢。

(Cheers and applause)

（欢呼和掌声）