At a press conference held two days ago, Google changed the previous technology company style released a Pixel phone, View VR DayDream box, Home Google and other hardware devices. But the press conference, CEO Sundar Pichai Google repeatedly mentioned a concept is Assistant Google intelligent voice assistant. Like Apple's Google, Assistant Siri has embedded in almost every new release of Google hardware and software products.
The background of this transition is the change in the competitive environment faced by Google. Google, or its parent company, Alphabet, which is currently the world's second highest listed company, is also a leader in the field of search, online advertising, mobile devices, and so on. But in the face of Google CEO last year just took Sundar Pichai the problem is, although the company in technology ahead of other big companies, but when you need to apply the technology to the product, Google has no similar to apple and Amazon and other achievements.
Compared to Amazon, Google has long been a leading position in the field of speech recognition, Natural Language Processing and machine learning and other technologies. But the only thing Google did not do is to make these technologies as early as Amazon made products, so that they go to occupy the consumer's home.
Perhaps it is already well prepared, perhaps to see the opponent's aggressive offensive, Google now decided to use their own advantages, and its use in the product, including software and hardware. And this advantage is the artificial intelligence.
10 years ago mobile-first, Google in the field of artificial intelligence, the killer is open, this time Google is also open source and Android status machine learning platform TensorFlow. TensorFlow's charm is that it is not only in the establishment and training of the speed of the neural network is 5 times faster than before, but also can support mobile devices, desktop, CPU, GPU and servers and other platforms.
Open source TensorFlow provides developers with a complete development guide and documentation, as well as a wealth of API interface, there are currently 50 a variety of Google products are using this system.
But now TensorFlow can do in the e-mail is far more than the identification of spam. Google launched in the enterprise user oriented Inbox mailbox, the message will be automatically classified as an important, travel, financial, social and other labels, and the completion of this classification is the TensorFlow machine learning system. This is the same as the principle of identifying spam, the more correct classification of the mail, the higher the classification accuracy of the system, while the Gmail and Inbox users have more than 1 billion, which brings a huge amount of data to the system.
The team was founded in 2012, when it's main goal is to carry out a deep study and neural network research experiments. Now this experiment has been completed, but the team retained, and under the leadership of Jeff founder Dean TensorFlow in a number of areas to improve the user experience of Google products.
Now Google Assistant has appeared in the Google of a variety of hardware and software products, and is likely to be open to third party developers, which makes the Google and Amazon's Alexa and Apple's Siri and Microsoft's Cortana competes in the field of voice assistant.
We can see the Google of this ambition from yesterday's new product released by Google. For example, the biggest selling point Google Pixel mobile phone is equipped with a Google Assistant, it not only makes Pixel mobile phone intelligent voice response and processing tasks like Siri, compared with the previous version in translation, image recognition has been greatly improved. The smart speaker Home Google is almost Assistant Google into a physical device, you can let it play music, check the weather and schedule.
In the software, Google focus on the introduction of the chat application Allo also embedded Assistant Google, which allows Allo to achieve intelligent recovery, image recognition and other functions. Although the Google Gmail and Inbox has 1 billion users, but the e-mail for young people are really old, so Allo has become a Google attempt to achieve a new social dream.
However, the Allo intelligent recovery function using the machine learning technology needs to send the information sent back to the user's Google algorithm analysis, and Google will store the information for a period of time. This is the point., he believes Google store this information will result in the user's privacy leak.
In the social product, Apple also encountered similar privacy issues, but Apple's approach is more respect for the privacy of the users, that is the end to end for information transmission and differential privacy technology (Differential Privacy). This technology will be cut into several pieces of user data, and adding noise mathematical conceals a user's personal information, and then the encrypted data will be uploaded to Apple's servers used for training neural network. This also allows users to use some of the local knowledge base in the absence of new vocabulary, so that the system more quickly understand its meaning, which is a good way to enhance the experience of human-computer interaction.
To be able to take into account the needs of the data and user privacy, because Apple has been in the real product for many years. Although it can not be said that Google has no respect for the user's privacy, but Snowden's criticism is not without reason. It may be that Google's engineers are in the dominant position, thus ignoring the privacy concerns of users. But in the rush to push the process of artificial intelligence to users, Google also need to learn a lot like its peers.