Google I/O 2016: Artificial Intelligence Becomes Product DNA

Following the developer conferences of Microsoft and Facebook, Google held its own I/O conference in the early hours of this morning (Beijing time). We found that artificial intelligence has become the DNA of Google products.

[[166530]]

However, from Allo to Google Home, many new products proudly demonstrated the artificial intelligence and machine learning technologies behind Google Assistant. They are not only the brains of the core products of the conference, but also the advantage that Google has over other competitors.

1. From mobile-first to AI-first

Currently, if the mobile service thinks something is useful, Google will pop up a prompt on the phone, thanks to Google's Knowledge Graph, which understands hundreds of millions of entities and their relationships, including people, places, and things. Rishi Chandra, Google's vice president of products, said: "There will be no better engine to answer users' questions. We fundamentally believe this."

The rest of the world queries Google over 100 billion times a year, which translates to about 3 billion searches a day. Ten years ago, Google translated two languages. Today, it is speaking to people in over 100 languages, translating over 140 billion words a day.

Pichai believes that Google's success in defeating its competitors is inseparable from its tradition as an engine company that "organizes the world's information." But now, Pichai hopes that it can understand things more comprehensively, and achieve a shift from "mobile first to artificial intelligence first." This means changing the operating soul of the company. An early sign is that the former head of Google Search retired and was taken over by the company's artificial intelligence head John Giannandrea.

The “traditional” Google is an information search company. The “modern” Google is based not only on suggesting relevant information, but also on informing and assisting. Informing means that we give you information when you need it. For example, Google Now tells you that it’s time for an appointment. Assistant helps you take action - plan a trip, make a reservation; anything that can be done through the Internet, Google can help you do it. With information search, recall and precision of over 80% are good - not every suggestion has to be perfect, because users can ignore bad search results. However, assistant functions are much more difficult. You can’t use a service that has a 20% error rate for bookings, and you won’t use it even if the error rate is 2%. Therefore, the assistant function must be much more accurate, and therefore smarter and more aware of the usage situation. This is what we mean by “AI first”. -Peter Norvig

Google is moving the center from the blank white web to every piece of technology (phones, cars, smart watches) and online service that readers own. Google.com will become more personal (Google for you). All of this, Pichai said, marks a profound change in the way we use information.

1. Google Assistant (vs Siri, Alexa, Viv)

Play emoji games with Google Assistant

[[166531]]

You can chat one-on-one with Google Assistant. It will remember your name and what you are doing. You can ask it "Did my team win?" and it will understand which team you are talking about.

The assistant recognizes that she's talking about Italian restaurants, so an icon appears that says it has relevant information. Clicking the icon makes it visible to everyone in the discussion group, and they can instantly make a reservation.

Google Assistant is a personal assistant that's like a conversational version of the Google search box, with some elements of what we've seen in Google's mobile search app. It's an evolution of Google Now, competing with Apple's Siri and Amazon's Alexa.

Google Assistant can start a simple conversation, during which it can help users complete tasks such as calling a car, finding headlines, finding the fastest route, calling contacts, and even finding graduation photos directly in the photo app and sharing them with friends, searching for movies playing nearby, buying tickets, etc. Google Assistant can solve problems for you in a continuous conversation, instead of just asking and answering questions as before.

"We are making search more and more assistive. We want users to have a continuous two-way conversation with Google, helping them get more done in the real world. We think of this as building a single Google for users."

Over time, Google may allow other developers to join so that other third-party services can be integrated into the system.

Both Google Home and Allo integrate Google Assistant as a core part of the product.

For example, you can ask Google Home devices whether your package has been shipped. They will tell users that it has been shipped based on Gmail messages. If you ask about traffic conditions to the airport, the device will say that it is more congested than usual and will also send alternative route plans to the user's phone.

For example, when using the Allo chat app, if the Google Assistant sees you're discussing something with a contact, like going out for dinner, it can offer to help. Anyone in the thread can then ask the Assistant a question for advice or to get in touch.

2. Allo

Google hopes to change the way people use Google products through Allo. Google CEO Sundar Pichai said in his opening speech today, "We are trying to make the search function more like an assistant."

Allo is a mobile chat application launched by Google, representing the determination of Google's communication department to get back on track. In the past few years, products such as Google+, Wave, and Buzz have never gained a place in the market, losing to market giants such as Facebook, Twitter, and Snapchat.

The newly launched Allo has a built-in Google Assistant. In addition to basic chat functions, users can also modify text size, create personalized picture labels, play games with Google Assistant, and let it help you provide chat content, book restaurants, find information, etc. The more frequently you use Allo, the smarter it will be. Allo can also help users find contacts from the phone book. For users who have Google accounts such as Gmail, they can also use Allo to send messages directly to contacts.

When you chat with friends on Allo, it will automatically recommend "Smart Replies" to you. If a friend invites you to dinner, Allo may suggest you reply "What time?" As you use Allo more and more, it will learn to reply intelligently according to your tone. So, if you have an emoji expression that you particularly like to use, it is very likely to appear in the next reply suggestion. In addition, Smart Reply can even recognize pictures through Google Photos and suggest replies related to the pictures.

Although the virtual assistant is the core, just like the personal assistant M launched by Facebook before, there are many differences between the two.

M is being tested in California with “thousands” of tasks. But when the task is too complex, such as ordering food offline, a human team needs to step in to complete the task. However, the assistant used by Allo is completely self-run and open to everyone.

Pichai said these AI features are just the beginning, as there are still many computational challenges that have not been solved. How to balance personal privacy and consider the social impact of this human-like thinking are issues that need further exploration. "In some ways we may be pioneers, but in some ways we are willing to follow the right direction pointed out by others," Pichai said, "This is how we keep moving forward."

At this I/O conference, Google showed more of its vision for the future, because Google hopes that developers and users can understand Google's ideas-and make them a part of the Google ecosystem. "Just like we spend a lot of energy and time on search, this (artificial intelligence) is also the direction we plan to carve for many years," Pichai said, "This is a grand journey."

The app will be available to iPhone and Android users this summer.

3. Duo Video App

[[166532]]

Duo is a companion app to Allo. It has similar functions to Facetime. It uses WebRTC and QUIC protocols to communicate and can make video calls even in poor network conditions. Based on the phone number, users can directly use Duo to dial for video calls. And it has a complete encryption function to ensure the privacy of users.

The biggest feature of Duo is the knock knock function, which gives you a live streaming in advance, so that you can understand why the other party called you before you answer the video and prepare in advance. Imagine that if a boyfriend and girlfriend are arguing, the girl doesn't want to answer your call, but when she sees you prostrating and apologizing in the video, the possibility of her being soft-hearted increases greatly.

Duo will be officially launched this summer and will be available on Android and iOS platforms.

4. Google Home (vs Echo)

Simply put, Google Home, similar to Amazon's Echo and Apple's Apple TV, is the center of the smart home system, connecting various smart hardware in your home.

Google Home and Echo have many similar functions. For example, you can play and control music, add appointments, and check movie times online. Of course, you can also use the integrated Google Assistant to control the lights at home and the temperature of the air conditioner. In addition, it can tell stories and popularize science to children; it can also remind you at any time about the weather and what you should do.

Google Home is also different from Echo.

Let’s talk about the advantages first.

The brain of Google's smart home is Google Assistant, which draws on Google's vast search database, while Echo is based on Amazon's virtual assistant Alexa. So Google Home may be smarter than Echo, at least when it comes to completing basic tasks such as web searches and traffic information. In a recent comparison test (competing Amazon, Apple, Google, and Microsoft), Google performed best in completing basic tasks, largely due to the huge amount of data Google's search engine has.

In addition, Google Home has another advantage over Echo: it can communicate with other devices. Not only can it be paired with other speaker systems, but it can also communicate with Google Chromecast digital TV sticks. This means that you can synchronize playlists in different rooms, or simultaneously show TV series on TV. In terms of communicating with other devices, Echo does not have the prospect of Google Home. Similarly, Home can also send information directly to the user's smartphone. Google's demo video shows a father asking Home about traffic conditions on the way to work. Home replied that there was an accident along the way and sent an alternative route to his phone.

"Google Home ... is built on more than a decade of research in natural language processing," said Mario Queiroz, Google's vice president of product. Google Home can also accurately recognize different accents and understand children's commands.

Like Google Web Search, Google Home also comes with options for personal privacy settings. "The rules of operation, privacy and control of the system are the same," said Scott Huffman, vice president of Google's search project.

Let’s talk about the disadvantages.

While Home is smarter than Echo for basic tasks, when it comes to services from Google and other companies, such as ordering food from a restaurant, Home's success will largely depend on whether Google can convince third-party developers to use Home to provide services. Pichai said that to do this, they really need to work with developers and third parties to provide these services to product users.

In comparison, Amazon has made great progress in this regard. Alexa can call Uber, order Domino's pizza, etc. Although time will tell who will get more favor from third parties, Echo, which was launched in June last year, obviously won in terms of time.

In addition, Gartner Research analyst Brian Blau believes that the biggest challenge facing Google is not technical, but how to get its products into the hands of users.

One of the reasons for the success of Amazon's Echo is that Amazon can promote the $179 smart speaker heavily on its website, and people are ready to buy when they visit Amazon's website. However, Google does not have a similarly attractive platform to promote its products.

Let's clarify the relationship between the concepts. Google calls the virtual assistant part of the search product Google Assistant. Previously, it was called Google Now, and later it was called Google again. However, this assistant lacks a name like Siri or Alexa.

The biggest difference between Google Assistant and other similar products is that it can have natural conversations. This assistant will appear with Google Home and Allo, but the voice command is not "Google assistant" but "Google".

2. Other Products

1. Android N new system

Google has chosen to release the Android N developer preview ahead of schedule. The biggest functional improvements in Android N are the addition of split-screen multi-window support, quick replies and notification grouping, improved battery life, and everyone's favorite emoji.

Security improvements were also demonstrated at the conference, and it will support Java 8, while Vulkan becoming part of the platform will allow developers to make better games and provide better image processing. So, will the effect of playing mobile games on Android phones surpass that of iPhones in the future?

2. Daydream, a new VR platform

More than 50 million VR apps have been downloaded from the Cardboard Marketplace

Android N launched the DayDream mode for VR. This "VR mode" can be seen as a mode dedicated to smartphones.

The most important feature is that when the phone opens the VR application, it can call the "VR mode" to release extra performance, and more importantly, it can shorten the delay and avoid the difference between action and vision affecting the immersion. Currently, Samsung, HTC, LG, Xiaomi, Huawei, ZTE, Asus and Alcatel are all manufacturing mobile phones that support this standard.

New Google standard VR controller

At the same time, Google has also established development standards for VR-enabled helmets and controllers.

In addition, Google is also working hard to promote VR as a new platform. Now, Google's own products such as Youtube, Google Play, Google Streetview, Google Photos, etc. are all committed to launching Daydream / VR versions. Google has also become a VR partner with Netflix, Hulu, WSJ, New York Times and other online and traditional media.

3. Android Wear 2.0 new wearable

Directly input handwriting on the watch screen for quick recognition

Android Wear 2.0 is the biggest upgrade of the wearable operating system released by Google in 2014. The biggest feature of Android Wear 2.0 is that it can run software independently, no longer relying on smartphones. Watches using this system can connect to Wi-Fi by themselves. Android Wear 2.0 mainly focuses on applications related to fitness, diet, communication and itinerary planning.

4. Android Instant Apps

Instant Apps launched by Google are similar to DeepLink technology, but more importantly, they are backed by the powerful Google Play. Instant Apps provide a temporary installation version, and if you want to use it for a long time after trying it, you can download the full version with just one click.

Developers only need to use the API provided by Google to support this function in a very short time.

5. Mobile analytics

Firebase, a backend developer services platform acquired by Google in 2014, has become more robust. Using Google Analytics as its foundation, Firebase now adds mobile analytics. Developers will be able to categorize user groups based on certain criteria and push notifications and activities in a targeted manner. Pichai said, "Firebase is the most comprehensive developer service we have ever made, and I'm glad it's available to everyone today."

6. Google Cloud

Google has invested heavily in its cloud platform to compete with Amazon and Microsoft. Pichai has told investors in the past that he hopes to distinguish between Google Cloud and machine learning and artificial intelligence. Today, he said that developers using Google Cloud will be able to use Google's computer vision API, speech API, and translation API. "When you use Google Cloud Platform, you can use not only the great software that Google uses internally, but also the specific hardware that we use internally," he said.

In addition, he said, "When you use Google Cloud Platform, you are also taking advantage of TPUs," TPUs are Google's own custom integrated circuit chips for applications, called tensor processing units (TPUs). The name was inspired by Google's TensorFlow open source deep learning framework. But this technology is unique.

[[166534]]

These TPUs are used in the AlphaGo AI-powered Go player that beat top Go player Lee Sedol, as well as Google Search and Google Street View, and this was one of the specific pieces of hardware Pichai mentioned for internal use.

Facebook F8 Developer Conference, Microsoft Build 2016, and today's Google I/O Conference all showed us the technological trend of "AI first". This is just like Sony, which developed AI earlier (but later gave up), and now resumed its AI business, showing us that the spring of AI development seems to have really arrived.

What, some ask, would a Sundar Pichai Google search engine look like in his AI-first world?

Pichai smiled broadly and said, “It depends on who’s searching.”