如果你年纪大了还记得在 90 年代尝试过语音听写软件(voice dictation software),那么你就会知道乏味的真正含义。如果您想用语音输入电子邮件或备忘录,则需要数小时的培训、令人痛苦的缓慢语速以及最终相当无用的文本。
今天的情况再不一样了。由于机器学习和云(machine learning and cloud)计算能力的重大进步,数字语音识别(voice recognition)系统几乎可以像理解另一个人一样理解你。
这就是为什么可以为每部智能手机、笔记本电脑和平板电脑(laptop and tablet)配备内置数字助理(Digital Assistant)的原因。三个主要参与者是来自苹果的(Apple)Siri、来自微软的(Microsoft)Cortana和来自谷歌的谷歌(Google)助手(Google Assistant)。万一从名字上看不出来。
这三位助理都从事类似的工作,也有自己独特的价值主张。客观地说哪个助手是“最好的”可能是不可能的,但是将它们与特定的常见任务进行比较是一个有用的练习。对于每个阅读本文的人来说,这些助手中的一个最适合您的具体情况。
因此,对于这种三向比较,目标将是展示每个选项的弱点和优势,以便您可以就与谁一起去做出明智的选择。然而,在我们这样做之前,让我们介绍一下我们三位可爱的参赛者。
Siri——真正的原创(Siri – A True Original)
Siri最初是作为iPhone 4S的独有功能推出的。考虑到iPhone 11将于 2019 年推出,这使得Siri 成为(Siri one)当今最古老的数字助理之一。
当Siri的第一个预告片播出时,它看起来有点像科幻小说(science fiction)。惊人的语音合成(voice synthesis)和看似完美的语音识别(voice recognition)是真正的亮点。回想起来,Siri涉及到一定程度的烟雾、镜子和相对有限的高质量功能领域。从那以后她的智力和五官都(intelligence and features)变大了,但流行的观点似乎是,老姑娘不再是技术领导者(technology leader)。
无论如何,这个助手仅限于苹果生态系统(Apple ecosystem),所以更大的问题是苹果爱好者(Apple aficionados)是否应该坚持他们的设备 附带的商店品牌助手。(store-brand assistant)
Cortana – 从游戏女神到私人助理(Cortana – From Game Goddess To Personal Assistant)
Cortana以 Bungie 举世闻名的Halo 系列(Halo series)游戏中的角色命名。她是协助士官长的AI,是(Master Chief)游戏系列(game series)的英雄。现在,您也可以让她标志性的声音(voice talk)在您的脑海中畅谈。好吧,那就是如果你戴着耳机和那个全尺寸的Halo Master Chief 收藏家(Halo Master Chief collector)的头盔。
由于微软的移动操作系统(operating system)实际上已经不复存在,Cortana主要是Windows 10的一项功能,在Android 和 iOS 商店(Android and iOS stores)都有应用程序。这意味着她提供了一些很棒的Windows 集成(Windows integration)功能,但当然不是人们实际使用的任何移动系统的原生功能。
谷歌助手——进行图灵测试(Google Assistant – Taking On The Turing Test)
谷歌(Google)无疑是世界上最顶尖的科技公司之一。尤其是在高级人工智能算法(AI algorithms)方面。所以毫不奇怪,谷歌助手(Google Assistant)被最新技术塞进了鳃里。
在某些方面,这可能有点问题。功能集不断变化,并不总是很清楚Google Assistant的结束位置和其他Google服务的开始位置。
例如,Google Duplex是一项令人印象深刻的服务,Google 的AI 聊天机器人(AI chatbot)会打电话给(例如)当地的美发师并为您预约。你只要问就行啦。然而,Duplex是作为自己的项目开发的,然后作为扩展被引入到Assistant中。(Assistant)
使用Google Assistant的好处是它与Google服务紧密集成,其中包括与Android的完美集成。
我们如何测试(How We Tested)
这种比较背后的想法不是根据纸上的规格来比较这三个助手。相反,我们的想法是对它们进行比较,根据您每天实际使用它们的方式进行实际操作。
这里重要的是用户体验以及这些出色的机器人是否真的按照您的意愿行事。如果特定的助手(specific assistant)对您的特定用例(use case)具有杀手级功能(killer feature),那么无论如何这都是没有意义的。
我们使用两个单独的设备来测试这里有问题的三个数字助理。对于Siri,我们使用的是 iPad Pro 9.7英寸平板电脑(” tablet),在撰写本文时安装了最新版本的 iOS 12。对于其他两款设备,我们使用的是Samsung Galaxy Note 10+。
由于这个想法是为了感受日常使用,每个人都被赋予了完全相同的常见任务,包括进行互联网搜索(internet search)、创建日历提醒(calendar reminder)和启动应用程序。
语音合成质量(Voice Synthesis Quality)
这可能不是一个硬性的技术规范,但数字助理的生死取决于说话时听起来多么自然。这些应用程序的重点是将用户转移到对话界面。这(Which)意味着刺耳的机械声音是(voice isn)不可接受的,并且会让大多数人放弃使用这样的产品。
在我们对这三者的测试中,谷歌助手(Google Assistant)名列前茅,Siri位居第二,Cortana位居第三。谷歌和苹果的产品听起来都很自然,至少在他们的北美英语(English)迭代中是这样,但可怜的Cortana听起来她需要升级。
触发短语反应性能(Trigger Phrase Reaction Performance)
任何现代数字助理(digital assistant)的一大特点是免提使用。您所要做的就是说出这个神奇的词,助手随时准备提供帮助。至少在理论上。所以我们想看看触发词(trigger word)的反应如何。
请注意(Please note),对于 iOS,只有Siri支持免提触发。虽然你可以在 iOS 上同时安装Cortana 和 Google Assistant(Cortana and Google Assistant),但你不能只用一句话来启动它们。这里最简单的解决方法是让Siri启动另一个助手。这很奇怪,但如果你绝对需要,这只会增加几秒钟的额外工作。
让我们从 Apple 的Siri开始。立即使用“Hey Siri ”触发短语(trigger phrase),您无需等待提示即可提问。Google Assistant的表现也一样,但默认情况下它的行为并非如此。至少不是在非谷歌手机上。在菜单中进行大量挖掘以找到隐秘命名的“语音匹配”设置,它可以让你从手机上的任何地方召唤数字精灵,即使它被锁定。
Cortana 的应用仅(only )在应用打开时响应触发词。(trigger word)所以说实话并不是那么有用。
无聊但有用的东西(The Boring But Useful Stuff)
设置提醒、打开网页、启动应用程序和其他类似任务是使用数字助理的基础。(bread and butter)那么这三个票价如何?
在进行网络搜索时,这三个助手都非常好。Cortana使用Bing,这对于喜欢 Google 算法的人来说可能不是一件好事。三人都乐于设置提醒、做备忘录并在日历上安排约会。
在 iOS 上, Siri可以从容地启动应用程序,并且乐于使用Apple Music 服务(the Apple Music service)播放音乐。在Android上,Google Assistant也很乐意满足应用启动请求,但奇怪的是不能与YouTube Music一起正常工作。这有望很快得到解决。Cortana 坚决(Cortana)拒绝启动任何应用程序,也(apps or step)拒绝整体走出她的应用程序泡沫。
这就是Cortana将为我们做的所有事情。这(Which)一点也不多。
设备集成与控制(Device Integration & Control)
如果数字(Digital)助理可以访问正确的权限,那么他们非常适合免提操作您的设备。最新版本的Siri几乎可以完全控制她所在的Apple 设备。(Apple device)她可以管理大量的切换,例如打开或关闭WiFi 或蓝牙(WiFi or Bluetooth)。您也可以使用Siri激活辅助功能,例如VoiceOver。
在Android上,谷歌助手(Google Assistant)也做得很好,可以切换常用设置而不会跳过一个节拍。然而, Cortana(Cortana)无法管理,这在这一点上并不令人惊讶。
非语言使用(Nonverbal Use)
虽然这些设备的主要展示功能(showcase feature)是您可以简单地与它们交谈以完成工作,但与您的数字助理进行大声交谈并不总是可能或可取的。您可能在公共汽车上、开会或其他您宁愿保持安静的地方。
如今,大多数助手都有文本聊天模式(text chat mode),您可以在其中停止对话,但仍然可以获得自然语言响应。
然而,Siri 默认情况下不会这样做——您必须转到 iOS辅助功能菜单并激活(accessibility menu and activate) Type to Siri。Cortana 和 Google Assistant(Cortana and Google Assistant)都默认提供文本输入(text input),因此他们在这里获得了少量的额外功劳。
语音识别性能(Voice Recognition Performance)
数字助理对您的理解程度(assistant understand)如何?这个问题有两个部分。第一个与您的语音到文本的准确转换有关。如果您说的话变成了类似自动更正的乱码,那也没多大用处。问题的第二部分与软件对您所说的内容的解析程度有关。
三位助手都毫无问题地理解了我们所说的并正确解析了它。这里没什么可报告的,但2019 年的语音识别和 AI(voice recognition and AI)令人印象深刻。
判决——谁胜出?(The Verdict – Who Came Out on Top?)
对于苹果(Apple)用户来说,判决是明确的。尽管Siri在(Siri)尖端技术(cutting-edge technology)方面客观上落后于Google Assistant ,但它与 iOS 和(Google Assistant)苹果(Apple)硬件(s hardware)的深度集成意味着它提供了最佳的整体体验。
特别是考虑到其几乎完美的免提性能。如果您还使用Apple Music等(Apple Music)Apple服务并花一些时间微调Siri快捷方式,那么很难想象会忍受其他数字助理在 iOS 设备上的限制和束缚。
对于无论如何都无法访问Siri的(Siri)Android用户来说,事情变得更加有趣。在我们的Galaxy Note 10+测试设备上的(test device)Cortana 和 Google Assistant 之间(Between Cortana and Google Assistant),Google Assistant显然占了上风。Cortana 使用Bing作为其搜索引擎(search engine)是可以理解的,但Bing始终给出不太令人满意的搜索结果,这是对它的一个相当严重的标记。
然而,最后,Cortana在(Cortana)Android上是一个非常有限的应用程序。Windows上的Cortana存在的跨功能可能使其成为一个有用的附加应用程序,但当Google 助手(Google Assistant)就在那里并且集成良好时,它不能被推荐为你的主要Android 助手。(Android assistant)
Siri, Google Assistant, & Cortana – Three Digital Assistants Compared
If you’re old еnough to remember tryіng voice dictаtion sоftware in the 90s, thеn you know the true meaning of tedіum. If you wanted to type out an email or a memo using your voice, it required hours of training, a painfully slow speaking speed and ultimately a pretty useless piece of text.
Today things couldn’t be more different. Thanks to major advances in machine learning and cloud computing power, digital voice recognition systems can understand you nearly as well as another human being.
Which is why it’s possible to ship every smartphone, laptop and tablet with a built-in Digital Assistant. The three main players are Siri from Apple, Cortana from Microsoft and Google Assistant, from, well, Google. In case it isn’t apparent from the name.
All three assistants perform similar jobs and also have their own unique value propositions to offer. It’s probably impossible to objectively say which assistant is the “best”, but comparing them along specific common tasks is a useful exercise. For every person reading this, one of these assistants is the best match to your specific circumstances.
So for this three-way comparison the goal will be to showcase the weaknesses and strengths of each option so that you can make an informed choice about who to go with. Before we do that however, let’s introduce our three lovely contestants.
Siri – A True Original
Siri first launched as an exclusive feature of the iPhone 4S. Consider that the iPhone 11 is launching in 2019, that makes Siri one of the oldest digital assistants today.
When the first teasers for Siri were aired, it seems a little like science fiction. Amazing voice synthesis and seemingly flawless voice recognition were real showstoppers. In retrospect, Siri involved some measure of smoke, mirrors and a relatively limited domain of high-quality functions. Since then she’s bulked up with intelligence and features, but the popular opinion seems to be that the old girl is no longer the technology leader.
Regardless, this assistant is limited to the Apple ecosystem so the greater question is whether Apple aficionados should stick with the store-brand assistant their devices ship with or not.
Cortana – From Game Goddess To Personal Assistant
Cortana is named after the character from Bungie’s world-famous Halo series of games. She’s the AI that assists the Master Chief, the hero of the game series. Now you too can have her iconic voice talk in your head. Well, that is if you wear headphones along with that full-size Halo Master Chief collector’s helmet.
Since Microsoft’s mobile operating system is effectively no more, Cortana is mainly a Windows 10 feature, with apps on both Android and iOS stores. This means she offers some great Windows integration features, but is of course not native to any mobile system people are actually using.
Google Assistant – Taking On The Turing Test
Google is undoubtedly one of the top technology companies in the world. Especially when it comes to advanced AI algorithms. So it should be no surprise that the Google Assistant is stuffed to the gills with the latest technology.
In some ways this can be a bit of a problem. The feature set is constantly shifting and it’s not always clear where Google Assistant ends and other Google services begin.
For example, Google Duplex is an incredibly impressive service where Google’s AI chatbot will phone (for example) the local hairdresser and make an appointment for you. All you have to do is ask. However, Duplex was developed as its own project and then rolled into the Assistant as an extension.
The upside of using Google Assistant is that it’s tightly integrated with Google services, which includes great integration with Android.
How We Tested
The idea behind this comparison is not to compare these three assistants according to their on-paper specifications. Instead the idea is to compare them, hands-on in terms of how you’d actually use them on a daily basis.
What matters here is the user experience and whether these brilliant bots actually do what you want them to. If a specific assistant has a killer feature for your specific use case, then this is all moot anyway.
We used two separate devices to test the three digital assistants in question here. In the case of Siri we used an iPad Pro 9.7” tablet, with the latest version of iOS 12 at the time of writing. For the other two devices, we used a Samsung Galaxy Note 10+.
Since the idea is to get a feel for day-to-day use, each one was given exactly the same common tasks, which include doing an internet search, creating a calendar reminder and starting an application.
Voice Synthesis Quality
It might not be a hard technical specification, but a digital assistant lives and dies by how natural it sounds when speaking. The whole point of these applications is to shift the user over to a conversational interface. Which means a jarring, robotic voice isn’t acceptable and would put most people off from using such a product.
In our testing of these three, it’s the Google Assistant that comes out on top, with Siri second and Cortana a rather distant third. Both Google and Apple’s products sound natural, at least in their North American English iterations, but poor Cortana sounds like she needs an upgrade.
Trigger Phrase Reaction Performance
One of the big features of any modern digital assistant is hands-free use. All you have to do is say the magic word and the assistant is ready to help. At least in theory. So we wanted to see how well the trigger word was reacted to.
Please note that in the case of iOS, only Siri supports hands-free triggering. Although you can install both Cortana and Google Assistant on iOS, you can’t launch them with just a word. The simplest workaround here is to ask Siri to launch the other assistant. It’s weird, but if you absolutely need to, this adds just a few seconds of extra work.
Let’s start with Apple’s Siri. Using the “Hey Siri” trigger phrase is instant and you don’t have to wait for a prompt before asking your question. Google Assistant performs just as well, but it doesn’t behave this way by default. At least not on non-Google phones. There’s quite a bit of digging in the menu to finding the cryptically-named “Voice Match” setting, which lets you summon the digital genie from anywhere on your phone, even if its locked.
Cortana’s app only responds to the trigger word when the app is open. So not all that useful to be honest.
The Boring But Useful Stuff
Setting reminders, opening webpages, starting apps and other similar tasks are the bread and butter of using a digital assistant. So how do these three fare?
When it comes to doing web searches, all three assistants are perfectly fine. Cortana uses Bing, which might not be a good thing for people who prefer Google’s algorithms. All three were happy to set reminders, take memos and put appointments on the calendar.
On iOS Siri launches apps with aplomb and is happy to play music using the Apple Music service. On Android, Google Assistant also happily complies with app launch requests, but strangely doesn’t work properly with YouTube Music. This will hopefully be fixed soon. Cortana flat out refuses to launch any apps or step outside her app bubble as a whole.
That’s all Cortana would do for us. Which is not much at all.
Device Integration & Control
Digital assistants are great for hands-free operation of your device if they can access the right permissions. The latest version of Siri has almost complete control of the Apple device she resides on. She can manage a large number of toggles, such as flipping WiFi or Bluetooth on or off. You can activate accessibility features such as VoiceOver using Siri as well.
On Android, Google Assistant did just as well, toggling common settings without skipping a beat. Cortana was however unable to manage, which is not too surprising at this point.
Nonverbal Use
While the big showcase feature of these devices is the fact that you can simply talk at them to get things done, it’s not always possible or desirable to have a loud conversation with your digital assistant. You may be on the bus, in a meeting or some other place where you’d rather stay quiet.
These days most assistants have a text chat mode where you can take the conversation off the air, but still get natural language responses.
Siri does not do this by default however – you have to go to the iOS accessibility menu and activate Type to Siri. Both Cortana and Google Assistant had text input available by default, so they get a small amount of extra credit here.
Voice Recognition Performance
How well can a digital assistant understand you? There are two parts to this question. The first has to do with accurate conversion of your speech to text. It’s not much use if what you say turns into an autocorrect-like garbled mess. The second part of the question has to do with how well what you’ve said is parsed by the software.
All three assistants understood what we said with no problem and parsed it correctly. Not much to report here but that voice recognition and AI in 2019 is pretty impressive.
The Verdict – Who Came Out on Top?
For Apple users, the verdict is clear. Although Siri objectively lags behind Google Assistant in terms of cutting-edge technology, its deep integration with iOS and Apple’s hardware means it provides the best overall experience.
Especially given its virtually flawless hands-free performance. If you also use Apple services such as Apple Music and spend some time fine-tuning Siri shortcuts, then it’s hard to imagine putting up with the limitations and hoops other digital assistants pose on iOS devices.
For Android users, who don’t have access to Siri anyway, things get more interesting. Between Cortana and Google Assistant on our Galaxy Note 10+ test device, Google Assistant clearly has the upper hand. Cortana’s use of Bing as its search engine is understandable, but Bing consistently gave less satisfactory search results, which is a rather serious mark against it.
In the end however, Cortana is a very limited application on Android. The cross-functionality that exists with Cortana on Windows might make it a useful additional app, but it can’t be recommended as your primary Android assistant when Google Assistant is right there and as well-integrated as it is.