ADM-201 dump PMP dumps pdf SSCP exam materials CBAP exam sample questions

谷歌Duplex AI的原理是什么? – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

谷歌Duplex AI的原理是什么?

Google Duplex A.I. - How Does it Work?

这里是ColdFusion TV
You are watching ColdFusion TV.
大家好 欢迎收看这一期ColdFusion视频
Hi, welcome to another ColdFusion video.
我相信你们大部分人都听说了谷歌的Duplex
So I’m sure by this stage, most of you guys have heard of Google Duplex.
这则新闻爆发时 我还在保加利亚
I was actually in Bulgaria when the news break about this.
因此我迅速回到澳大利亚 制作这期视频
So I quickly had to make my way back to Australia to make this video.
这就是我气喘吁吁的原因
So if I’m out of breath, that’s why.
此视频中 我们将深入了解一下谷歌的Duplex
Okay, so in this video we’ll take a deeper look at Google duplex.
那什么是Duplex呢
So what is duplex?
它基本上是谷歌助手的扩展
It’s basically an extension of Google assistant
它可以根据你的指令 跟真人通电话
that can make phone calls to real humans just by you asking it to do so.
是一种建立在Wavenet科技上的深度神经网络
It’s a deep neural network that builds off wavenet technology.
Wavenet是一个语音合成程序
Wavenet is a speech synthesis program
通过聚合短声音 生成一段讲话
that worked by joining very short units of sound together to create a speech.
它被发布时 被认为是自然语言合成的突破
It was a breakthrough in natural speech synthesis when it came out.
我们已经在上一视频中了解了它
We’ve already taken a look at it in a previous video,
这是一个其功能的简略回顾
but here’s a quick recap of what it can do.
从1770到1850 在英文诗歌和绘画中出色的方面
Aspects of the sublime in English poetry and painting, 1770 to 1850.
这是没有任何输入时 Wavenet说的话
This is what happens when you don’t type in anything for wavenet to say.
它依然会生成随机模仿人类声音的元音频
It’s still generating raw audio to randomly imitate human sounds
那依然听起来很像人类说话
and it still sounds like words.
归属于
Belong to.
“作为最华丽的人 我们得到了两英尺 那是一个垫子”
He as the head Sheen we’ve gained two feet so it was a pat.
“她只是打字 但她没有”
She did just type, dic she isn’t.
“说话”
Tell
[嘴唇声和呼吸声]
[Lip sounds and breathing]
“宁静祥和”
Serenity silent.
“使用铲子挖掘会挖得很慢”
Digger with a shavette to dig could dig slow.
Duplex是又一个建立在Wavenet之上的神经网络
Duplex is another neural network built on top of wavenet.
最终结果就是AI可以进行真实对话
The final result is an AI that can have realistic conversation,
并使用了Wavenet准确的发声
but with wavenet vocal precision.
这是2018年5月在Google I/O 上Duplex的华丽登场
And here’s the grand unveiling of Duplex at Google’s I/O event in May of 2018.
但即使在美国
But even in the US,
六成的小型企业没有在线订购系统
60% of small businesses don’t have an online booking system set up.
你接下来听到的是谷歌助手向一家沙龙打电话
So what you’re going to hear is the Google assistant actually calling a real salon
为你进行预定的对话
to schedule the appointment for you.
我们来听一下
Let’s listen.
[掌声]
[Applause]
[电话嘟嘟声]
[Telephone Beep]
你好 我能为你做什么
Hello, how can I help you?
你好 我要为客户预约一个女士发型
Hi, I’m calling to book a women’s haircut for a client.
我想订在五月三日
I’m looking for something on May 3rd.
好 稍等一下
Sure, give me one second.
嗯哼
Mm-hmm.
[观众大笑]
[Laughter]
好 你想要订在什么时间
Sure, what time are you looking for around?
中午十二点
At 12 pm.
好的 我们在十点有空
Ok, we have a 10 o’clock.
上午十点也可以
10 am is fine.
好的 预定在五月三日 上午十点 客户是丽萨
Okay, perfect. So I will see Lisa at 10 o’clock on May 3rd.
对 谢谢你
Okay great, thanks.
好的 祝你一天愉快 再见
Great. Have a great day. Bye.
[口哨和掌声]
[Whistles and applause]
[电话嘟嘟声]
[Telephone beep]
你好 我能为你做什么
Hi, may I help you?
你好 我想在七号周三预订餐桌
Hi, I’d like to reserve a table for Wednesday, the 7th.
一共七个人吗?
For seven people?
嗯…… 一共四个人
Umm… It’s for 4 people.
四个人 什么时候 今天还是今晚
Four people? When? Today? Tonight?
嗯……下个周三 下午六点
Umm… Next Wednesday, at 6pm.
事实上 我们只接受五人以上的预定
Actually, we reserve for like upwards of five people.
四个人的话 你们可以直接来
For four people, you can come.
等待有位置通常需要多久
How long is the wait usually to be seated?
什么时候 明天还是周末
For when? Tomorrow? Or weekend? Or…
[大笑]
[Laughter]
下个周三 七号
For next Wednesday, the 7th.
不 人不是很多 四个人可以直接来 可以吗
Oh no, it’s not too busy. You can come for four people, okay?
嗯 我懂了 谢谢
Oh, I gotcha. Thanks.
再见
Yep. Bye-bye.
[口哨和掌声]
[Whistles and applause]
我们的助手能理解上下文中的细微差别
But our assistant understands the context to nuance,
这次它有了询问等待时间的新行为
it new acts for rate times in this case
并出色地进行了交流互动
and handle the interaction gracefully.
根据谷歌的博客 你听到的”嗯”和”啊”
According to Google’s blog, the ums and ers that you hear
有时是合成时添加进去的
are put in sometimes synthetically,
但有时是系统正在像人一样
but actually it sometimes is there to signal that the system is still processing,
处理信息而发出的声音
just like a human would.
公众对此的反应为震惊和恐慌
The public reaction to this was on the side of shock and horror
有的人甚至很愤怒
and even anger by some.
有的人认为这是欺骗
Some people thought it was very deceitful
因为有人跟人工智能通了话 却不知道对方不是人类
to have an AI talk to someone over the phone without them knowing.
但谷歌声明 它们会保持透明
But Google has made it very clear that they’re going to be transparent.
他们会让对话另一端的人知道 他们在和Duplex通话
They’ll be letting the people on the other end of the phone know that they’re talking to duplex.
Duplex是在预约这一狭窄领域中训练的
Duplex has been trained in the narrow field or scheduling appointments or bookings,
包括询问生意在放假期间的的营业时间
and inquiring about a business’s opening hours on holidays.
要澄清的是 Duplex不能进行通常意义的谈话
To be clear, Duplex cannot have general conversations.
但我确信未来几年 其覆盖领域会越来越大
But I have little doubt that the scope will widen in the coming years.
谷歌说训练此类神经网络时 有特殊的挑战
Google stated that there were unique challenges when it came to training such a neural net.
如何让AI靠谱地理解自然语言
How do you get an AI to robustly understand natural language
并使用现实的方式回应
and reply in a realistic manner?
这是很难做到的
This will be pretty hard to do.
比如 相比计算机 每个人的说话方式都不同
For example people tend to talk differently to one another than with computers.
我们说得更快 会在对话中自我纠正
We talk faster, correct ourselves mid conversation
而且对话过程中还依赖上下文
and even a mid parts of conversation and rely on context instead.
通过一条低质量且有噪音的线路通话
Throw in the poor quality and noisiness of a phone line,
这是一个高难度的通话挑战
and you have a pretty hard challenge on your hands.
为了让这一点更明显
To solidify this point a little bit,
我们使用句子”四 可以”来举例子
Let’s think about the phrase, “Ok for 4”.
句子很简单 但它依赖于之前的句子作为上下文
It’s such a simple sentence but it relies on many previous sentences for context.
这句话可以指时间或人数
This phrase could refer to a time or an amount of people.
我们人类对这类句子习以为常
We as humans take such things for granted,
但从AI或解决此类问题的研究团队的角度来看
but it’s interesting to take a brand new look at this through the eyes of an AI
这将会很有趣
or a team of researchers trying to solve this problem.
好 Duplex是怎么工作的
OK, so how does Duplex work?
Duplex使用了一种递归神经网络
Duplex uses something called a recurrent neural network.
如果你不知道什么是神经网络
If you don’t know what a neural network is,
它基本上很多矩阵相乘的函数集合
it’s basically a massive matrix multiplication function
矩阵的每部分建立在叫做节点的人工神经元上
where each part of the matrix is built up of artificial neurons, called nodes.
节点包含一个数学公式并分层排布
The nodes contain a mathematical formula and are arranged in layers.
每个节点都有一个输入和输出
And each node has an input and an output.
在接受到可能有的输入后
After receiving the inputs, whatever they may be,
整个矩阵或神经网络的最终目标就是
The end goal of the whole entire matrix or neural network
基本上是找到减少时间的方法
is basically to find out how to reduce how long it is,
或者说减少错误的数量
or in other words reduce the amount of error.
也许神经网络最奇怪的地方在于
Perhaps the strangest thing about neural networks
没人确切知道它们如何得出了结果
is that no one actually knows how they come to their conclusion.
你给它输入 它不知怎么的就得出了结果
You just give them the inputs and they somehow get an answer.
神经网络已经出现一段时间了
Neural networks have been around for a while.
实际上 1971年帮助创建第一个CPU的人 Ted Hoff
In fact, Ted Hoff, the guy who helped create the very first CPU at Intel back in 1971,
在早年间研究过神经网络
actually worked on neural networks in his early career
但可用的神经网络在最近大约五年间才出现
but practical neural networks have only been possible in the past five years or so.
自从2012年以来 神经网络的复杂度大体上增加了五百倍
Since about 2012 the general complexity of neural networks has advanced 500 times,
我认为这是计算机科学中最有趣的领域之一
and I think it’s one of the most fascinating fields of computer science.
如果想更详细地了解神经网络的工作原理
If you want a more detailed explanation of how neural networks work.
Youtube频道3Blue1Brown有很多出色的视频
There’s an absolutely brilliant video by the YouTube channel 3Blue1Brown.
它们非常通俗 我会在简介中留下地址
It’s a remarkable explanation. I’ll leave a link in the description below.
好了 都快跑题了
But anyway, I’m getting sidetracked.
Duplex具体使用的神经网络
So the specific type of neural network that Duplex is using
如我提到的是递归神经网络
as I mentioned is called a recurrent neural network.
此类神经网络有一小块内部储存
These type of networks have a small internal memory
可以保存特定输入 用于理解上下文
that allows them to remember specific inputs to help understand context.
因此它算得上是语音合成的最佳神经网络
For this reason, it’s pretty much the perfect neural network for speech recognition
它也是大多语音识别算法的核心
and it’s at the heart of most speech recognition algorithms.
正如我之前所提
So as I mentioned earlier,
Duplex经过了很多电话通话的训练
Duplex was trained on a whole bunch of different phone conversations.
那它是如何知道正在发生的事呢
But how did it learn to understand what was going on?
一通对话会启动一个模拟讲话
Well, the conversations, of course, start an analog speech
这个讲话的音频成为谷歌自动语音识别系统的输入
and then this speech in audio form is fed into Google’s automatic speech recognition system.
这时音频为解释为了文本
From this point on this audio is now understood as text.
文本被转换为神经网络可以理解的形式后
In this text, once converted into a format of the neural network that can understand
就输入到系统中
is then fed into the system.
通话中的其他元数据和上下文信息
Other metadata and other wider context information from the calls,
比如谈话中预定的正确时间
such as the correct time for the appointment schedule in question,
还有一天中的几点都被输入到系统中
all the time of day is also fed into the system.
把它们结合起来
So tying it all together
当神经网络听取一堆电话录音输入时
when the neural network listens to a whole bunch of phone recordings as inputs,
最终它会学会如何减少自己的错误量
it eventually learns how to reduce the amount of error it has,
这意味着跟它对话时 它的反应更好
meaning that it has better responses when spoken to.
这一切神经网络训练出来的最终结果就是谷歌Duplex
The final train neural network that comes out of all of this is Google Duplex.
[电话嘟嘟声]
[Telephone beep]
我能帮你什么
How can I help you?
– 你好 -你好 怎么了
– Hello?- Hello, what’s up, man?
嘿 我想知道你们今天的营业时间
Hey, um… I wanted to know what are your hours for today.
上午十点到下午六点半
10 a.m. to 6:30 p.m.
好的 明白了 谢谢
Okay, got it. Thank you for your time.
不客气 再见
No problem, sir. Bye.
那Duplex可以干什么呢
So what can Duplex do?
除了谷歌示范的功能外
In addition to what was shown in the Google demo,
Duplex也能做些有趣的事情
Duplex can also do some pretty interesting things,
比如处理讲话被打断的情况
such as handle interruptions.
你的电话号码是多少
Ok, what’s your phone number?
– 223 -22什么
-Two two three-Two two what?
– 223 – 好 223
-Two two three.-Ok, two two three.
详细解释
Elaborate.
你好 我想在五月二十五日预定餐桌
Hi, I would like to reserve the tables for May 25th.
不好意思 哪一天
Sorry, what day?
周五 嗯 五月二十五日
For Friday, um… May 25th.
还可以通过电话回答问题
And respond to audio issues over the phone.
– 你在吗 – 在 我在这儿
-Are you here?-Yeah, I am here.
因此有很多关于这个AI通过图灵测试的讨论
So there was a lot of talk about this AI passing the Turing test.
如果你不知道图灵测试 它是一种
So for those of you that don’t know, the Turing test was a thought experiment
1950年计算机之父之一阿兰 图灵发明的思想实验
brought up by Alan Turing, one of the fathers of computer science, back in 1950.
他提出了如下的测试
He proposed a test that goes as follows:
一个人通过文字跟一台看不到的机器交流
A person would interact with a machine that they couldn’t see over text.
这个人不知道与他对话的是人还是机器
This individual doesn’t know if they’re talking to another person or a machine.
如果机器可以跟这个人交流
If the machine can interact with this individual
并不让那个人发现自己是台电脑
without them suspecting that they’re talking to a computer
那么那台机器就通过了图灵测试
then that machine passes the Turing test.
阿兰·图灵预言了到2000年
Alan Turing made the prediction that by the year 2000,
我们就会发明能够通过图灵测试的机器
we would have machines that would be able to pass the test.
他的预言偏差不是很大
And he wasn’t that far off.
第一台通过测试的机器是在2014年的一台文字交流机器人
The first machine to pass the test was a text-based chat bot in 2014.
图灵测试起初只是针对文本
The Turing test was originally just for text.
谷歌的Duplex似乎在很狭窄的对话领域中通过了图灵测试
And Google Duplex seems to have passed the Turing test in the very narrow field of conversation
是在它在提供预约服务的时候
when it comes to making appointments,
而且是通过声音 不仅仅是文本
but through voice, not just text.
我打赌这是很多计算机科学家现在还没意识到的事
I bet that’s definitely something that a lot of leading computer scientists didn’t see happening just yet.
我认为这点很有趣
So I found that very interesting on that point.
我有一些最后的想法
So some final thoughts:
有人认为未来这可能会让电话零售商
Some people may think that this may take jobs away from telemarketers,
或其他电话相关行业的人失业
and other phone based work in the future.
这非常有可能
This may very well be possible,
但如果我说这是肯定的或
but I’d just be being reactionary if I was to say where this was a certainty.
它会对社会有副作用的话 就有点过分了
Or if I was to say that this will be a net negative on society in any way,
因为现在得出结论还为时尚早
because at this point it’s far too early to tell conclusively.
记住了 Duplex不是为了日常对话而创造的
A reminder: Duplex isn’t for general conversation.
它只专门负责预定和询问假期时的营业时间
It’s only specifically for booking reservations and inquiring about open hours during holidays.
但我确信它能做的领域会不断扩展
But I’m sure its scope is going to grow.
就现在谷歌助手的能力来说
As far as the assistant goes,
因为领域有限 它还改变不了我们现在的生活
due to its limited scope, it doesn’t really change how we live our lives,
除非你是残疾人
unless you have a disability I’d say.
如果一家商店有网页
If a store has an online page,
通过网页预定会快得多
it’s far quicker just to book that way,
但再提一遍 Duplex使用了神经网络
but then again Duplex is using a neural network.
随着时间流逝 我们在这个频道已经见过一些惊奇事
And time and time again on this channel, we’ve seen how those things can surprise us.
它的覆盖领域越来越广 速度比我们想象的要快
The scope may just grow quicker than we think.
但总的来说 这项科技非常炫酷
But on a wider night, this technology is pretty cool.
可以肯定是的 我们在AI上到了一个转折点
It’s becoming clear that we’re at an inflection point when it comes to AI.
过去几年 我们在AI领域卓有建树 像是AlphaGo等
In the past couple of years, we’ve been marveling at AI breakthroughs like AlphaGo and others.
现在 似乎可以进入真实世界的AI出现了
But now it seems like we’re starting seeing real world applications of AI coming into view.
每一次都让不可能成为了可能
Each time making all this once- thought-impossible possible.
我认为这样的事会越来越常见
I think this will start to get more and more commonplace
同时也会有不断地突破
as breakthroughs keep occurring in parallel.
问我担心Duplex吗
So am I worried about Duplex?
一点也不 没什么让我担心的
Not at all, really. I don’t see anything that should make me.
谷歌声明当你跟Duplex AI对话时 信息会透明
Google has stated that they’re going to be transparent when you talking to a Duplex AI.
等到这种科技变常见了
But if this technology becomes commonplace,
再问我这个问题吧
ask me that question again.
总之 这个社会也要接受我们正处于新时代风口的事实
But I think all in all, society has to accept that we’re at the doorstep of a brand-new era,
一个有无限可能的时代
a time with unimaginable possibilities.
我为站在这里感到荣幸
I think that’s a bit of a privilege as it stands right now.
好 这就是视频的总结 谢谢观看
Anyway, that just about wraps up this video. Thanks for watching.
这里是ColdFusion 我是Dogogo
This has been Dogogo you’ve been watching ColdFusion.
如果偶然看到这个视频的话 请订阅此频道
Feel free to subscribe if you just stumbled across this channel.
我要感谢所有来到迪拜区块链峰会的人
And I want to thank all of you guys that came to the Dubi Blockchain Summit.
能见到你们很开心
It was really cool meeting some of you guys.
好 就这样吧 我们下个视频见
Anyway that’s it. I’ll catch you again soon for the next video.
干杯 玩的开心
Cheers guys, have a good one.
ColdFusion 新的想法
ColdFusion. It’s new thinking.

发表评论

译制信息
视频概述

叙述了谷歌Duplex AI的原理,讲解了其使用的神经网络、图灵测试,并分析了这一事件对社会的影响。

听录译者

One静茹

翻译译者

Reed

审核员

审核员YX

视频来源

https://www.youtube.com/watch?v=IuIpgArEZig

相关推荐