How AI Could Empower Any Business | Andrew Ng | TED

翻译人员: Yip Yan Yeung 校对人员: Grace Man

When I think about the rise of AI,

当我想到 AI （人工智能）的崛起之时，

I'm reminded by the rise of literacy.

我联想了读写能力的崛起。

A few hundred years ago,

几百年前，

many people in society thought

社会上的很多人觉得

that maybe not everyone needed to be able to read and write.

也许不是每个人都得会读会写。

Back then, many people were tending fields or herding sheep,

那时候， 很多人从事农业或者牧羊，

so maybe there was less need for written communication.

对书面交流的需求没有那么多。

And all that was needed

只有主教和僧侣

was for the high priests and priestesses and monks

to be able to read the Holy Book,

需要读得懂《圣经》和最高经典，

and the rest of us could just go to the temple or church

其他人只要去寺庙、教堂

or the holy building

或者圣所

and sit and listen to the high priest and priestesses read to us.

坐等主教读给我们听就行了。

Fortunately, it was since figured out that we can build a much richer society

幸运的是，人们后来发现 如果很多人能读能写，

if lots of people can read and write.

我们的社会会富裕得多。

Today, AI is in the hands of the high priests and priestesses.

如今，AI 被掌握在 “主教”手中。

These are the highly skilled AI engineers,

这些主教就是 那些技术高超的 AI 工程师，

many of whom work in the big tech companies.

其中很多就职于科技巨头公司。

And most people have access only to the AI that they build for them.

很多人只能接触到 为他们设计的 AI。

I think that we can build a much richer society

我认为，如果我们能让 每个人参与谱写未来，

if we can enable everyone to help to write the future.

我们就能创造一个更富裕的社会。

But why is AI largely concentrated in the big tech companies?

但是为什么大部分 AI 技术 都集中在科技巨头手中呢？

Because many of these AI projects have been expensive to build.

因为开发这些 AI 项目太贵了。

They may require dozens of highly skilled engineers,

这些项目需要一大群 技术高超的工程师，

and they may cost millions or tens of millions of dollars

要开发一个 AI 系统 可能要花上几百万几千万美元。

to build an AI system.

And the large tech companies,

这些大型科技公司，

particularly the ones with hundreds of millions

尤其是手握几亿 几十亿用户的公司，

or even billions of users,

have been better than anyone else at making these investments pay off

最擅长套回这些投入，

because, for them, a one-size-fits-all AI system,

因为对于它们来说， 一个普适的 AI 系统，

such as one that improves web search

比如优化搜索引擎

or that recommends better products for online shopping,

或者为网购推荐更佳商品的系统，

can be applied to [these] very large numbers of users

可以直接适用于庞大的用户群体，

to generate a massive amount of revenue.

产生巨额收益。

But this recipe for AI does not work

但是一旦你走出科技互联网行业，

once you go outside the tech and internet sectors to other places

去向别的领域， 这个 AI 的秘方可能就不会奏效，

where, for the most part,

因为在大多数情况下，

there are hardly any projects that apply to 100 million people

几乎没有一个项目 可以覆盖一亿人，

or that generate comparable economics.

或产生相当的经济效益。

Let me illustrate an example.

我来举一个例子。

Many weekends, I drive a few minutes from my house to a local pizza store

我总会在周末从家里开车去 当地一家披萨店

to buy a slice of Hawaiian pizza

向店主买一块夏威夷披萨。

from the gentleman that owns this pizza store.

And his pizza is great,

他的披萨很不错，

but he always has a lot of cold pizzas sitting around,

但是总是有一大堆披萨滞销到冷掉，

and every weekend some different flavor of pizza is out of stock.

每个周末都会 有几个口味的披萨缺货。

But when I watch him operate his store,

但是当我看着他 运营他的小店的时候，

I get excited,

我激动万分，

because by selling pizza,

因为在他卖披萨的过程中，

he is generating data.

也产生了数据。

And this is data that he can take advantage of

如果他能用上 AI， 就可以从这些数据中获益。

if he had access to AI.

AI systems are good at spotting patterns when given access to the right data,

如果输入了合适的数据， AI 系统就会很善于识别规律，

and perhaps an AI system could spot if Mediterranean pizzas sell really well

也许能有一个 AI 系统识别出 周五晚上地中海披萨

on a Friday night,

卖得特别好，

maybe it could suggest to him to make more of it on a Friday afternoon.

也许这就能告诉他 周五下午多做一点地中海披萨。

Now you might say to me, "Hey, Andrew, this is a small pizza store.

你有可能想这么对我说： “嘿，安德鲁（Andrew），

这只是个小披萨店。

What's the big deal?"

有什么了不起的？”

And I say, to the gentleman that owns this pizza store,

而我想说，对于店主来说，

something that could help him improve his revenues

如果有什么可以帮他每年

by a few thousand dollars a year, that will be a huge deal to him.

多赚几千美元， 那就很了不起了。

I know that there is a lot of hype about AI's need for massive data sets,

我知道，人们普遍认为 AI 需要大量数据集，

and having more data does help.

有了更多数据确实会有帮助。

But contrary to the hype,

但是如果没有大量数据，

AI can often work just fine

AI 通常也可以在

even on modest amounts of data,

只有少量数据的情况下正常运作，

such as the data generated by a single pizza store.

比如一家披萨店产生的数据。

So the real problem is not

真正的问题不是

that there isn’t enough data from the pizza store.

披萨店没有足够的数据。

The real problem is that the small pizza store

真正的问题是 这小小的披萨店

could never serve enough customers

没有足够的客源

to justify the cost of hiring an AI team.

平衡雇佣一组 AI 人员的支出。

I know that in the United States

我知道美国

there are about half a million independent restaurants.

有大约 50 万家独立餐厅。

And collectively, these restaurants do serve tens of millions of customers.

这些餐厅总计服务了几亿顾客。

But every restaurant is different with a different menu,

但是每一家餐厅都是不同的， 有着不同的菜单，

different customers, different ways of recording sales

不同的顾客， 不同的记账方式，

that no one-size-fits-all AI would work for all of them.

没有一个通用的 AI 系统 可以适用于全部的餐厅。

What would it be like if we could enable small businesses

如果我们可以让小型企业

and especially local businesses to use AI?

尤其是本土企业都能用上 AI， 会怎么样呢？

Let's take a look at what it might look like

我们来看看 AI 应用于一家

at a company that makes and sells T-shirts.

制造、销售 T 恤的公司 会是什么样的情形。

I would love if an accountant working for the T-shirt company

如果这家 T 恤公司的会计

can use AI for demand forecasting.

可以用 AI 预测需求， 那就会很不错。

Say, figure out what funny memes to prints on T-shirts

比如，通过研究 社交媒体上的潮流，

that would drive sales,

锁定一些印在 T 恤上增加销量的

by looking at what's trending on social media.

好玩表情包。

Or for product placement,

就上架策略而言，

why can’t a front-of-store manager take pictures of what the store looks like

门店经理可以拍下店铺情况，

and show it to an AI

提交给 AI，

and have an AI recommend where to place products to improve sales?

让 AI 推荐商品的摆放位置， 提高销量。

Supply chain.

供应链。

Can an AI recommend to a buyer whether or not they should pay 20 dollars

AI 是不是可以推荐 买家是否应该

per yard for a piece of fabric now,

以 20 美元一码的 价格购入一块布料，

or if they should keep looking

还是应该货比三家，

because they might be able to find it cheaper elsewhere?

因为别家的价格 有可能会更低廉呢？

Or quality control.

质量管理。

A quality inspector should be able to use AI

一名质检员 应该能够使用 AI

to automatically scan pictures of the fabric they use to make T-shirts

自动扫描 T 恤的面料照片，

to check if there are any tears or discolorations in the cloth.

检查布料是否有裂缝或褪色。

Today, large tech companies routinely use AI to solve problems like these

如今，AI 已经成为大型科技公司 处理此类问题的常规手段，

and to great effect.

成果显著。

But a typical T-shirt company or a typical auto mechanic

但是现在没有一家普通的 T 恤公司、普通的汽修店、

or retailer or school or local farm

零售店、学校、本地农场

will be using AI for exactly zero of these applications today.

会用 AI 运营。

Every T-shirt maker is sufficiently different from every other T-shirt maker

每一家 T 恤制造商的情况 都是截然不同的，

that there is no one-size-fits-all AI that will work for all of them.

没有一个通用的 AI 系统 可以适用于全部商家。

And in fact, once you go outside the internet and tech sectors

其实，如果不看互联网和科技领域，

in other industries, even large companies

去看一些别的领域， 就算是一些大公司，

such as the pharmaceutical companies,

比如医药公司、

the car makers, the hospitals,

汽车制造商、医院，

also struggle with this.

都会饱受这个问题的困扰。

This is the long-tail problem of AI.

这就是 AI 的长尾效应。

If you were to take all current and potential AI projects

你可以把所有 已有和潜在的 AI 项目

and sort them in decreasing order of value and plot them,

以价值降序排列后作图，

you get a graph that looks like this.

就会得到这样一张图。

Maybe the single most valuable AI system

也许最有价值的 AI 系统

is something that decides what ads to show people on the internet.

决定了在网上 给人们展示什么广告。

Maybe the second most valuable is a web search engine,

也许第二有价值的系统 是网络搜索引擎，

maybe the third most valuable is an online shopping product recommendation system.

第三有价值的系统是 网购商品推荐系统。

But when you go to the right of this curve,

但是如果你看向曲线的右侧，

you then get projects like T-shirt product placement

就会看到像 T 恤商品陈列、

or T-shirt demand forecasting or pizzeria demand forecasting.

T 恤需求预测和披萨店需求预测 这样的项目。

And each of these is a unique project that needs to be custom-built.

每一个这样的项目 都需要定制。

Even T-shirt demand forecasting,

就算是 T 恤需求预测，

if it depends on trending memes on social media,

如果它由社交媒体上的 流行表情包决定，

is a very different project than pizzeria demand forecasting,

也与披萨店需求预测 是两种泾渭分明的项目，

if that depends on the pizzeria sales data.

披萨店的预测由销售数据决定。

So today there are millions of projects

如今成千上万的项目

sitting on the tail of this distribution that no one is working on,

就处于这个无人问津的分布长尾上，

but whose aggregate value is massive.

但是它们的合计价值是不可小觑的。

So how can we enable small businesses and individuals

我们该如何让小型企业和个人

to build AI systems that matter to them?

有能力搭建对他们 十分重要的 AI 系统呢？

For most of the last few decades,

在过去的几十年中，

if you wanted to build an AI system, this is what you have to do.

如果你想搭建一个 AI 系统， 你需要做这些事。

You have to write pages and pages of code.

你需要写长篇累牍的代码。

And while I would love for everyone to learn to code,

虽然我觉得人人都该学写代码，

and in fact, online education and also offline education

线上和线下教育也确实

are helping more people than ever learn to code,

让学习编程的人数达到了高峰，

unfortunately, not everyone has the time to do this.

不幸的是， 不是人人都有时间学习编程。

But there is an emerging new way

但是，我们现在 有了一个全新的方式，

to build AI systems that will let more people participate.

创造 AI 系统， 让更多人参与编程。

Just as pen and paper,

就像纸笔

which are a vastly superior technology to stone tablet and chisel,

是比石板和凿子 先进得多的科技，

were instrumental to widespread literacy,

在普及读写的过程中功不可没，

there are emerging new AI development platforms

现在也有一些 新的 AI 开发平台

that shift the focus from asking you to write lots of code

不再让你写一大堆代码，

to asking you to focus on providing data.

而是只让你提供数据。

And this turns out to be much easier for a lot of people to do.

这对大规模人群来说更容易实现。

Today, there are multiple companies working on platforms like these.

现在有很多公司在做这样的平台。

Let me illustrate a few of the concepts using one that my team has been building.

我的团队也在做这类平台， 我来给大家介绍其中一个。

Take the example of an inspector

举个例子，检测员

wanting AI to help detect defects in fabric.

需要 AI 的帮助 检测布料瑕疵。

An inspector can take pictures of the fabric

检测员可以拍下布料的照片，

and upload it to a platform like this,

上传到这样的平台上，

and they can go in to show the AI what tears in the fabric look like

然后他们可以用矩形做标记，

by drawing rectangles.

告诉 AI 布料裂缝长什么样。

And they can also go in to show the AI

他们也可以通过标记矩形，

what discoloration on the fabric looks like

告诉 AI 布料褪色长什么样。

by drawing rectangles.

So these pictures,

这些图片

together with the green and pink rectangles

与检测员标记的绿色和粉色矩形框

that the inspector's drawn,

are data created by the inspector

就是检测员创建的数据，

to explain to AI how to find tears and discoloration.

告诉 AI 如何检测裂缝和褪色。

After the AI examines this data,

AI 检查了数据之后，

we may find that it has seen enough pictures of tears,

我们会发现， AI 已经读取了足够的裂缝图片，

but not yet enough pictures of discolorations.

但是没有足够的褪色图片。

This is akin to if a junior inspector had learned to reliably spot tears,

这就类似于一个初级检测员 已经学会了如何准确地识别裂缝，

but still needs to further hone their judgment about discolorations.

但是还得再磨练一下对褪色的判断。

So the inspector can go back and take more pictures of discolorations

这个检测员可以回去 再拍几张褪色的照片，

to show to the AI,

提交给 AI，

to help it deepen this understanding.

加深它对褪色的理解。

By adjusting the data you give to the AI,

通过调整输入 AI 的数据，

you can help the AI get smarter.

你可以让 AI 变得更聪明。

So an inspector using an accessible platform like this

检测员使用这样容易操作的平台，

can, in a few hours to a few days,

在几小时至几天内，

and with purchasing a suitable camera set up,

再采购一套合适的摄影设备，

be able to build a custom AI system to detect defects,

就能在搭建起一个 定制化 AI 系统，

检测工厂中所有 T 恤面料上的 瑕疵、裂缝和褪色情况。

tears and discolorations in all the fabric

being used to make T-shirts throughout the factory.

And once again, you may say,

你可能又想说：

"Hey, Andrew, this is one factory.

“嘿，安德鲁，这就是一家工厂，

Why is this a big deal?"

有什么了不起的？”

And I say to you,

我想告诉你，

this is a big deal to that inspector whose life this makes easier

对那个减负的检测员来说， 这很了不起，

and equally, this type of technology can empower a baker to use AI

同样，这项技术可以让 一名烘焙师使用 AI

to check for the quality of the cakes they're making,

检查手中蛋糕的质量，

or an organic farmer to check the quality of the vegetables,

让一名有机农场主 检查蔬菜的质量，

or a furniture maker to check the quality of the wood they're using.

让一个家具制造商 检查木材原料的质量。

Platforms like these will probably still need a few more years

这类平台也许还需要一些时间

before they're easy enough to use for every pizzeria owner.

将操作难易度调节至 适用于每一个披萨店店主。

But many of these platforms are coming along,

但是很多平台都在进步，

and some of them are getting to be quite useful

有些平台只需要少量培训，

to someone that is tech savvy today,

就已经对如今懂技术的人来说 非常有帮助了。

with just a bit of training.

But what this means is that,

这也就意味着，

rather than relying on the high priests and priestesses

我们不需要再依赖于主教

to write AI systems for everyone else,

为所有人编写 AI 系统，

we can start to empower every accountant,

我们的每位会计、

every store manager,

每位门店经理、

every buyer and every quality inspector to build their own AI systems.

每位买家、每位质检员都有能力 搭建自己的 AI 系统。

I hope that the pizzeria owner

我希望披萨店店主

and many other small business owners like him

和其他像他这样的小型企业主

will also take advantage of this technology

都可以用上这项技术，

because AI is creating tremendous wealth

因为 AI 创造着巨大财富，

and will continue to create tremendous wealth.

也将在未来持续创造巨大财富。

And it's only by democratizing access to AI

只有让人人都有机会用上 AI，

that we can ensure that this wealth is spread far and wide across society.

我们才能将这样的财富 播撒到社会的每个角落。

Hundreds of years ago.

几百年前。

I think hardly anyone understood the impact

我觉得几乎没有人懂得

that widespread literacy will have.

普及读写的重要性。

Today, I think hardly anyone understands

我认为现在几乎没有人懂得

the impact that democratizing access to AI will have.

让每个人有机会 用上 AI 的重要性。

Building AI systems has been out of reach for most people,

大多数人没有机会 搭建 AI 系统，

but that does not have to be the case.

但是未来不一定会是如此。

In the coming era for AI,

在接下来的 AI 时代中，

we’ll empower everyone to build AI systems for themselves,

我们会让每一个人有能力 为自己搭建 AI 系统，

and I think that will be incredibly exciting future.

我觉得这就是我们 振奋人心的未来。

Thank you very much.

谢谢。

(Applause)

（掌声）