Sina Technology · Jun Jun from US Silicon Valley
The AI era has arrived
It is rather dramatic that the Microsoft Build Developer Conference and the Google I/O Developer Conference crashed this year and the opening time was only one day away. Microsoft is in Seattle from the 7th to the 9th; Google is in Mountain View, from the 8th to the 10th of this month. The developer community may have its own focus, but the science and technology media will have to split it up.
These two giants of technology have many business overlaps and competitions. They are just two representatives of the PC era and post-PC era. They currently have the world's largest desktop and mobile operating systems. In the PC era, Microsoft and Intel's Wintel portfolio swept the market; in the post-PC era, Google and Apple split dominated the mobile platform.
Whether it is yesterday's Microsoft Build conference, or today's Google I/O conference, there is a common theme word ——artificial intelligence. Almost everywhere, AI runs through the numerous complex new products of the two giants, redefining the experience of software services. There is no doubt that the technology industry is moving into a new era of AI.
In fact, Google and Microsoft last year defined AI as their future strategy. Whether it's what Microsoft CEOs say "AI Redefine All" or "Google Mobile First to AI First" proposed by Google CEO, the strategy of two Indian-American CEOs In fact, there is a source: AI will become the blood that runs through the two giants, penetrates and enhances all products and businesses of Microsoft and Google.
From this point of view, this year's Microsoft and Google I/O conferences do not have many new products. Instead, the original products are updated in the era of AI, and have a new experience and charm. In other words, the two major developer conferences this year are the continuation of last year's conference. They are the transcripts handed down by the two giants based on the AI strategy determined last year, which is the embodiment of the ubiquitous AI.
AI runs through all products
Yesterday Microsoft released Azure AI, IoT, Microsoft 365, Brainwave, today Google's updated family bucket software ——Android P, Gmail, Google Photos, Google Assistant, Google News, Google Maps, Google Lens, Autopilot, None First, it is not a more humanized experience brought about by AI. It is also a representation of deep learning related technologies such as speech recognition, image recognition, behavior prediction, and natural semantic understanding.
Taking today's Google conference product as an example, Gmail can help users write emails with predictability. Users don't need input of one word or one word; Photos automatically recognizes faces, and gives users the ability to retouch and automatically convert texts; Assistant can handle two users at the same time. Orders, or even direct calls back to the merchant to book; News can automatically build news topics to help users understand the ins and outs of the hot events; Waymo's automatic driving can predict those unexpected accidents, slow down in advance to avoid the red lights suddenly cross the road Pedestrian vehicles. Specific product new features introduced, you can see Sina Technology's Google Assembly feature articles.
On the other hand, artificial intelligence products require user data. Only when you trust Google and trust your personal data with Google can you enjoy the convenience of all AI-related products such as Google Assitant and Gmail. The preconditions for acquiring user data are privacy and security; this is what Microsoft and Google highly emphasized in the two keynote speeches.
As a media practitioner, I have always been looking forward to artificial intelligence technology that can liberate the work efficiency of the media. The tedious, time-consuming, and less creative work of the future news media, emergency news, handling press releases, and the construction of special topics can all be given to artificial intelligence. Media people should focus their valuable resources and energy on original content and in-depth analysis of these highly creative tasks.
There is one need to pay tribute to the two giants. While Google and Microsoft are promoting cutting-edge technologies, they have not forgotten the concern of the disadvantaged groups of the disabled. Yesterday, Microsoft launched a $25 million five-year plan "AI for Accessibility". Co-developers use artificial intelligence technology to help people with disabilities improve their living experience. Today, Google also specially introduced the Morse code input method, which can not help with machine learning technology. Expressed disabled people communicate with the outside world.
Every year, Microsoft draws five minutes from the keynote to introduce the technology to help people with disabilities. The year before last was artificial intelligence and smart glasses to help blind people. “Looking at the world” last year was a smart bracelet to help Parkinson patients re-live this year. Artificial intelligence helps blind students to overcome language barriers. These gentle stories of human nature and science and technology not only reflect the concern for the vulnerable groups, but also the best brand advertising for the technology giants.
Focus on 2B and 2C markets
The topic returned to the two giants. The two major developer conferences of Microsoft and Google are actually the developers' conference of the two platforms in the field of artificial intelligence. However, the two giants have slightly different artificial intelligence focus areas. Microsoft is mainly targeting the 2B enterprise market, while Google dominates the 2C consumer market.
Microsoft's AI product is deeply integrated into its own Azure cloud platform and Windows platform, including Azure AI and Windows ML. The Azure AI highlighted in this Build Conference is intended to promote its own "cloud + productivity + AI” full set of services to the enterprise user field. Huawei's millet embedded in the keynote speech is embedded with intelligent translation. The upgraded version of IoT represents the embodiment of Microsoft AI's specific use in the corporate market. It can be seen that Microsoft is eager to build an enterprise market ecosystem based on its own cloud AI technology.
In the corporate market, Microsoft AI's competitors mainly include Amazon AWS and IBM's Watson. Although in the field of cloud services, Microsoft Azure's market share still lags behind Amazon's AWS, Microsoft also has its own unique advantage in expanding AI services. It is the original productivity customer. Microsoft's Windows and Office occupy a dominant position in the global large enterprise market. 95% of Fortune Global 500 companies are using Microsoft's cloud services (including Azure and Office); Office 365's monthly corporate users have exceeded 120 million in October last year.
Google’s AI products are also fully integrated into their own full set of Internet services and Android platforms. These low-volume, multi-billion-magnillion user bases are the most competitive advantage for Google’s AI development and are also directly accessible to consumers in the consumer market. (Android monthly mobile devices broke 2 billion units last year, which does not include the Chinese mainland market.)
Every Google software product that passes AI's "Open" software, such as Google's almost all applications released today, can bring more user data to Google to meet the training needs of deep learning. Because of this, Google never broke its own platform and opened up almost every AI software product to the iOS platform. Compared to mobile platforms, Google is paying more attention to massive amounts of iOS user data, and even paying huge fees to buy user traffic from Apple iOS. Bernstein analysts estimate that this part of the traffic purchase cost last year was as high as 3 billion US dollars.
Google's AI competitors in the consumer market, mainly Amazon and Facebook. Although Google is far less user-friendly than Internet service users, Amazon and Facebook each have an unassailable dominant position in e-commerce and social networking. Amazon's Alexa even has its own ecological advantages in smart home and interactive entertainment, which is why Microsoft chose Cortana to work with Amazon Alexa in the consumer AI field. Although the two companies compete in the corporate market, they face the Google’s big rival in the consumer market.
Behind the computing platform contest
In addition to the eco-platform and user data, the AI strategy debate also has a cloud computing platform that supports AI operations and training. This is also the most eye-catching part of Microsoft and Google Developers Conference; yesterday Microsoft released a deep learning acceleration project, Project Brainwave (released last year), and today Google released the third generation of TensorFlow chip TPU 3.0.
It is worth mentioning that the TPU 3.0 processor uses liquid-cooled technology and its computing power is eight times faster than its predecessor. The cornerstone of Microsoft's Brinwave is Intel's Straix FPGA chip. In August last year, Microsoft officials stated that using Intel's Straitx 10 FPGA chip, Brainwave can support 39.5 trillion operations per second without batch processing.
Since 2012, Microsoft has already bet on cloud and AI's chip strategy on more flexible FPGA chips. This is an important reason for Intel to spend $16.7 billion to acquire FPGA chip giant Altera in 2015. Over the past few years, Intel has stuck to an important position in the AI era with a series of tens of billions of dollars in major acquisitions (Altera, Nervana, Movidius, Mobileye).
Microsoft's deep learning platform in the cloud is based on Intel's FPGA chip, and the mobile-side Windows ML platform is based on Intel's Movidius VPU chip. Although Intel has also announced cooperation with Facebook to build AI chips, recent series of signs show that Facebook is building its own AI processor R&D team.
In contrast, the interests of Microsoft and Intel in the AI era are more closely tied. Microsoft's expansion in the AI space also means the expansion of Intel's AI chips. In a sense, Microsoft and Intel, the golden partners of the PC era, once again joined hands in the era of AI to achieve mutual prosperity.
At this year's three major developer conferences, there is still Apple WWDC in June. The cornerstone of Apple’s AI strategy is obviously more than 1.3 billion iOS users and related massive user data. Last year, Apple released the Core ML machine learning framework at WWDC, demonstrating for the first time its own planning in the field of artificial intelligence. Since Apple DNA highly emphasizes data security and user privacy, Apple has locked its AI platform on the mobile device itself. Core ML machine learning is done locally on iOS devices. This is a double-edged sword. Although it guarantees data security, it is also a limitation of computing power and puts forward higher requirements for mobile-end chips.
Although Microsoft and Google also have their own mobile Windows ML and TensorFlow Light frameworks, this is not the core of the two giants' AI strategy; their machine learning framework is deeply tied to their own cloud platform. In the face of Microsoft's and Google's cloud AI strategy, will Apple continue to insist on the localization of machine learning? It will be clear next month.