Close your eyes and picture a shoe.

閉上你的雙眼，然後想像一隻鞋子

OK.

好，

Did anyone picture this?

有人想像的鞋子長這樣嗎？

This?

這隻？

How about this?

還是這隻？

We may not even know why, but each of us

我們可能不知道為什麼，但我們每個人

is biased toward one shoe over the others.

偏偏想到了某款鞋子，而不是其他款式

Now, imagine that you're trying to teach a computer

現在試著想像你要試著教導一部電腦，

to recognize a shoe.

去辨認一隻鞋子

You may end up exposing it to your own bias.

最後電腦接收到的資訊可能都會受你的偏見左右

That's how bias happens in machine learning.

這就是為什麼人類偏見會影響機器學習

But first, what is machine learning?

但首先，什麼是機器學習呢？

Well, it's used in a lot of technology we use today.

在今日的科技裡，我們很容易見到它的應用

Machine learning helps us get from place to place,

機器學習引導我們前往目的地

gives us suggestions, translates stuff,

給我們建議、翻譯語句，

even understands what you say to it.

甚至能進行語音辨識

How does it work?

它是如何運作的？

With traditional programming,

傳統上，程式編碼需要

people hand code the solution to a problem, step by step.

人手動調整每一個步驟，提出問題的解決方案

With machine learning, computers learn the solution by finding patterns in data

有了機器學習，電腦能找到資料中的模式，進而學習提出解方

,so it's easy to think there's no human bias in that.

因此我們很容易以為，沒有人類偏見會參與這個過程

But just because something is based on data doesn't automatically make it neutral.

但我們不能理所當然地因為它的基礎是數據資料，就認為機器學習很中立，

Even with good intentions, it's impossible to separate ourselves from our own human biases,

即使立意良好，我們永遠也無法剝除自我偏見

so our human biases become part of the technology we create in many different ways.

所以在許多方面，人類偏見變成我們科技發明的一部份

There's interaction bias, like this recent game

有所謂的互動偏見，像剛剛這場小遊戲，

where people were asked to draw shoes for the computer.

參與者必須為電腦畫出一雙鞋子

Most people drew ones like this.

多數人畫出了這樣的款式

So as more people interacted with the game,

當越來越多人參與遊戲，

the computer didn't even recognize these.

電腦就會變得認不出這種鞋款

Latent bias-- for example, if you were training a computer

再來是內隱偏見，舉例來說，如果你要訓練一部電腦

on what a physicist looks like, and you're using pictures of past physicists,

去學習辨認一位物理學家，用的是過去物理學家的相片，

your algorithm will end up with a latent bias skewing towards men.

你的運算法則最後會出現偏向男性的內隱偏見

And selection bias-- say you're training a model to recognize faces.

再來談談揀選偏見，假設你要訓練一個人臉辨識模型

Whether you grab images from the internet or your own photo library,

無論你是從網路上或自己的相片庫中選取圖片，

are you making sure to select photos that represent everyone?

你確定你挑的照片能夠代表每個人嗎？

Since some of our most advanced products use machine learning,

既然某些最先進的科技產品裡已運用了機器學習，

we've been working to prevent that technology from perpetuating negative human bias--

我們一直為終結人類偏見而努力

from tackling offensive or clearly misleading information

對於冒犯性或高誤導性的資訊，

from appearing at the top of your search results page

我們避免它出現在你的搜尋結果頁面頂端

to adding a feedback tool in the search bar

此外，我們在搜尋欄位中加上了一個回饋工具，

so people can flag hateful or inappropriate autocomplete suggestions.

讓使用者可以舉報仇恨性或不適當的自動填入搜尋

It's a complex issue, and there is no magic bullet,

這是一個複雜的議題，而且並沒有特效藥

but it starts with all of us being aware of it,

但我們可以從正視它開始做起，

so we can all be part of the conversation,

讓大家都能參與對話，

because technology should work for everyone.

因為科技應該是為了我們每個人而發展