At the beginning of 2019, when we try to review the historical events of the domestic smart speaker & voice intelligence industry last year and predict what will happen in the industry in the next new year, We bluntly named Baidu first.

Baidu's development this year can be described as a "sudden rise". It has developed from a verbal leader to a de facto "industry benchmark", which can be described as a very "typical". It can not only serve as a reference for other Internet giants that are arranging around their core competitiveness in their respective fields, but also serve as a reference for many small and medium-sized enterprises that are developing along the direction of a large ecological layout and seeking differentiated development.

This year, we have seen many seemingly gorgeous events, but for the "initiator" of the event, the event must not have happened by accident. Behind it lies Baidu's thinking about the current situation, its own competitiveness, and market positioning, as well as its efforts in hardware products, conversational AI systems, upward skill applications, downward industry applications, and ecology, etc. strategies and actions.

Assess the situation and be determined to take the first place

The domestic smart speaker & voice intelligence industry has gradually kicked off with a series of platform and product releases by major manufacturers in 2017 . Although before this, Amazon Alexa’s ecosystem had become increasingly mature, prompting many domestic small and medium-sized enterprises to follow suit, but it did not form a large ecological trend.

Because the domestic market is mainly a market for domestic players. After all, neither Amazon nor Google have plans to enter China in the short term. The developer ecosystem is not the focus of their global expansion, and selling voice AI products in China is of little significance. Therefore, among the many domestic players who started at almost the same time, it is very meaningful who can get the "first".

We have seen that after nearly a full year of "Battle of 100 Boxes" around 2017, the voice intelligence market has also undergone some changes in 2018: First, the pattern of leading players in the smart speaker market The second is that the implementation of voice AI is further developing in a diversified direction.

On the first point, the most direct manifestation is that starting from the first quarter of 2018, domestic Alibaba and Xiaomi ranked among the top five in the world for the first time, filling the gaps left by domestic brands in the global market in the past, and continuously Maintained a stable market position for three quarters. China has also become the second largest smart speaker market after the United States. Alibaba and Xiaomi have long been at the forefront, and Dingdong ranks third.

In the third quarter, this situation changed again: According to the 2018 Q3 China Smart Speaker Report released by Canalys, Baidu, the "dark horse", entered the top three in the market for the first time with strong market performance. Compared with Alibaba and Xiaomi, which are still in the top two (negative growth), Baidu, which achieved a quarter-on-quarter growth of 711% in this quarter, appears to have full potential.

△ Canalys 2018 Q3 China Smart Speaker Report

In addition to its own brand products, Baidu’s performance in enabling device volume is also very intriguing: In the third quarter of this year, IDC released According to the report, DuerOS-enabled smart speaker shipments ranked first in China (1.54 million units, including Xiaodu’s own brand and other DuerOS-enabled smart speaker products).

△Jing Kun, general manager of Baidu Smart Life Group (SLG), announced the latest business progress at the 2018 Baidu World Conference

Baidu’s official good news said: As of November 2018 , DuerOS smart device activations reached 150 million, and voice interactions exceeded 800 million times, basically achieving the rhythm of doubling data in each quarter for seven consecutive quarters. DuerOS has become China's largest, most active and most prosperous conversational AI operating system.

This is not the achievement of one person, but the whole focus of the "smart life" product line and business layout

On March 6, 2018, Baidu announced the official establishment of the smart life business Smart Living Group, referred to as SLG. It is jointly composed of Dumi Business Department, Hardware Ecosystem Channel Department and Raven Studio. On May 18, Baidu announced that Jing Kun, general manager of Dumi Division, will serve as general manager of the Smart Life Group (SLG), reporting directly to Robin Li. Among them, the DuerOS Business Division continues to focus on the construction and operation of the DuerOS platform and ecosystem, while the Hardware Ecosystem Channel Business Division focuses on the mass production of first-party hardware, e-commerce construction and channel expansion.

This change actually reveals some secrets: Baidu has elevated conversational AI to a strategic focus of the company, and is promoting the development of DuerOS from a top-down strategic level. First of all, it is reflected in the core technologies required for conversational AI. Baidu’s research results related to the field of artificial intelligence, as well as Baidu Brain, big data search, and cloud capabilities, give DuerOS the greatest support and continue to increase Baidu’s smart hardware layout.

At the same time, in the past year or so, Baidu Strategic Investment has also increased its investment in DuerOS from the group level. This is reflected in the increase in holdings or strategic investments in content companies such as NetEase Cloud Music, Baidu Video, Dragonfly FM, and Pear Video, and strategic investments in hardware and IoT companies such as Xiaoyu Home, Jimi Technology, Yunding Technology, and BroadLink, which have further contributed to the DuerOS paves the way for the development of the hardware ecosystem.

What is most obvious in the consumer market is Baidu’s subsidy policy for its own hardware products. Baidu's financial report shows that due to the increase in promotion and channel costs, sales costs, comprehensive expenses and administrative expenses increased by 54% year-on-year, and Q3 increased by 51% year-on-year.

The support from the company’s strategic level, the tilt of resources, and the stability of the team allowed DuerOS to quickly improve the core capabilities of conversational AI, deploy and implement its own brand hardware products within one year, and penetrate the platform into the third Third-party hardware, expansion of industry applications, and establishment of a complete ecosystem have laid an important foundation.

Product Strategy: How Internet Companies Lead the Smart Speaker Consumer Market from Zero to One

This year, Baidu basically released a piece of hardware at a press conference every two to three months. products, and finally launched four smart hardware products: Xiaodu Home (599 yuan), Xiaodu Smart Speaker (89 yuan), Xiaodu Smart Speaker Pro (169 yuan), and Xiaodu Voice Car Mount (starting at 69 yuan) , forming a "prosperous" Xiaodu family.

Regarding the functional details of each product, I will not go into details here. For details, you can check Shenzhen Bay’s previous reports. Here, we try to analyze Baidu’s smart hardware product strategy from these products to analyze the reasons for the product’s hot sales.

1. Combination play

In addition to the four smart hardware mentioned above, the Huawei tablet M5 (including stand) jointly launched with Huawei at the end of 2018 has also been included in the product system.

At present, the Xiaodu family has covered all product forms and completed full coverage from high, medium and low-end prices, including screen, screenless and mid-to-high-end and low-end speakers. Meet the purchasing needs of geeks, early adopters, and those who have high requirements for smart speaker sound quality.

△ Xiaodu smart speaker: a mini smart speaker, focusing on high cost performance

This combination is very similar to the Amazon Echo family:

Amazon in 2017 It has launched 11 pieces of hardware including Echo Plus, which focuses on sound quality, Echo Show, a smart speaker with a screen, Dash Wand, which focuses on voice shopping, and Echo Buttons, which are used with voice games.

In 2018, Amazon continued to lead the way. In addition to updated iterations of its original products, it successively launched Echo Sub subwoofer, children's smart speaker Echo Dot Kids Edition, Echo Smart Plug smart socket, etc. 13 pieces of hardware.

2. Segmented user group coverage

Xiaodu’s typical users include families of three, three generations living under one roof, a world of two, young people living alone, and empty-nesters. In addition to providing users with different price sensitivities with a larger range of product choices, the Xiaodu family also achieves vertical coverage from segmented scenarios and users.

With the release of Xiaodu smart speakers in June 2018, Baidu added "geek mode" and "kid mode" to the smart speakers. The former demonstrates Baidu's big data search, deep learning, and multi-round dialogue capabilities, while the latter is the application of Baidu's AI capabilities in children's scenarios. Although the children's scene is not a new proposition, it is the first segmented user scene other than music that Baidu has clearly captured and focused on.

Xiaodu Home, which originally had the advantage of sound-screen interaction, has become a tool for children to accompany and learn after adding a children's mode, combined with children's voice interaction and children's content. One device meets the needs of the elderly (video calls) and children (accompanying and learning) in the family.

3. Full scene coverage from home to car

Although DuerOS has long launched hardware solutions for car machines and vehicle-mounted equipment, the milestone event was Baidu’s launch in November The launch of the Xiaodu Voice car mount in March marked the first time that DuerOS entered the car scene as a post-installation car product.

△ Xiaodu Voice Car Mount: Standard version and wireless charging version. Used with the Xiaodu Bluetooth APP, it can provide a full-scenario voice interaction experience in car scenarios, including voice-activated navigation, music on demand, weather query, traffic conditions and other common services.

From the family living room to the bedroom to the car, family members include the elderly, children, and young people. The Xiaodu family has been able to provide corresponding high-frequency applications to users from different dimensions, so as to achieve segmented scenarios, Deep exploration of applications.

4. Dimensionality reduction attack

In an era when large manufacturers spend money to subsidize the market, Baidu is also a heavy participant. The most typical one is Xiaodu Home, China’s first smart video speaker, which hit the bottom of the industry with a price of 599 yuan as soon as it was launched. During the Double Eleven and Double Twelve periods, a group price of 299 was launched, continuing to move towards the goal of "hits".

△ Xiaodu at Home: With voice and screen interaction, Xiaodu at Home has the export of video content, the ability to communicate, and the ability to machine vision... In Baidu’s view, this It is a product that gets the user experience right and has the ability to bring artificial intelligence into thousands of households.

In the eyes of the outside world, what Xiaodu is seizing at home is the market opportunity of smart speakers with screens. In fact, Xiaodu Home is not targeting similar products from beginning to end, but screenless smart speakers that do not appear to be of the same dimension. Or simply put, its purpose may be to directly kill the entire category of screenless smart speakers (see Shenzhen Bay’s previous interpretations of events).

And the effect of Baidu’s doing this is already very obvious: in Baidu’s “Double Eleven” success report, Xiaodu dominated the sales rankings of multiple e-commerce platforms at home. Baidu was able to catch up from behind and squeeze into the third place in the domestic smart speaker market, and Xiaodu played an important role at home.

5. Third-party hardware: seize the big deal

In terms of ecological cooperation hardware, Baidu DuerOS focuses on cutting into the existing market and seizing mobile phones in mobile scenarios, set-top boxes and TVs in living rooms. Boxes and other "head" devices with tens of millions of sales have successively reached cooperation with "head" manufacturers.

Specifically listed in 2018: Skyworth mainstream models Q6/Q7 series, TCL mainstream models X/C/P series, XGIMI H2 screenless TV; Skyworth X1 PRO Xiaopai smart TV Audio, Gehua Xiaoguo set-top box, Skyworth π box; vivo X21, vivo NEX, Huawei P20, Huawei P20 Pro, OPPO Find .

Platform strategy: focus on user experience, strengthen core system capabilities, and create a closed business loop

For an Internet giant like Baidu, it is not enough to have a set of beautiful smart hardware. To illustrate their business ambitions or ambitions. Behind the Xiaodu hardware family is the trump card of DuerOS.

1. Focus on user experience

It used to be the “largest” platform, but now I prefer to say it is the “best user experience” platform. A very important aspect of human-computer interaction experience is the experience of content acquisition.

At the 2nd Baidu AI Developer Conference in July 2018, DuerOS 3.0 was released. What kind of label does the latest upgraded DuerOS 3.0 hope to put on itself? When asked this question by Shenzhen Bay, Jing Kun gave the following answer.

Baidu’s own smart hardware labeled “Xiaodu” has a greater mission to carry the core capabilities of DuerOS conversational AI. Compared with peers, the actual selling price of these products is even lower than the BOM cost. In order for users to actually use it, use it well and spread the word, so that the price subsidy can truly have a planning effect, the product experience is crucial. of.

The device cannot wake up, the conversation cannot continue, and there is no desired content... These are the main complaints that users have about the "smart capabilities" of smart speakers, and they are also the original sins that make smart speakers turn into "speakers for intellectual disabilities" . In response to these problems, Baidu continues to iterate and optimize at the core system level around "hearing clearly, understanding, and satisfying".

2. Strengthen core system capabilities

It is the mission of DuerOS to create a more natural intelligent interaction experience where "machine actively learns and adapts to humans".

As the chief spokesperson of Xiaodu Home, Robin Li demonstrated Endless Conversation at the opening of the Baidu World Conference in November 2018, which is a more natural and smooth multiple continuous interactions in human-computer dialogue. This demonstration It represents a comprehensive upgrade in the interactive experience of Xiaodu products.

△ Robin Li demonstrated the latest dialogue capabilities of Xiaodu at home at the 2018 Baidu World Conference

At the system and technical level, it is specifically reflected in:

Understanding Ability upgrade, satisfaction upgrade, reaction speed upgrade. Taken apart, Xiaodu can understand the user's complex intentions and conduct multiple rounds of clarification and guidance based on the user's statements;

With the help of Baidu's powerful knowledge graph and accumulation of search data, DuerOS can conduct large-scale Deep learning of DuerOS constructs a semantic model with a larger "brain capacity", so that the more you use it, the better it understands you;

Through improvements and optimizations at the DuerOS cloud, voice ASR, and device levels, we create an industry Leading device response speed.

3. Create a closed business loop

In the final analysis, you still have to make money by making products.

When talking about the essence of making money, Zhao Peng, who was hired by Baidu this year as the senior director of Dumi Division, always shows the calmness of a businessman.

This is an issue that SLG must think through before letting Baidu Group executives make decisions such as subsidies. "Either give users data, which will lead to rapid growth and high activity; or give me money." Such a voice is like an undercurrent, which makes DuerOS must be forward-looking.

The release of DuerOS 3.0 represents the birth of a new business model for domestic voice AI platform companies represented by Baidu.

The first step in realizing this business model is to allow content and service resources and developers from the original PC and mobile era to make money in voice interaction scenarios, so that the platform has a complete commercial closed loop.

Taking skill developers as an example, there are 4 ways to obtain income from DuerOS: in-skill payment, paid skills, 100 million yuan developer support plan, and 2018 Conversational AI Skills Competition. DuerOS’s business sharing model, and 100% of the skill income will be returned to developers within 6 months.

△ The 12-year-old youngest developer who took the lead in making money on DuerOS by developing conversational AI skills

Content resource owners can use DuerOS’s platform tools to make money on DuerOS devices. Quickly distribute content on the Internet, and then make money through "payment within skills" and "paid skills".

Let partners on the platform make money, including equipment vendors, solution vendors, chip vendors, content providers, developers, etc. In an open and active ecosystem, every role can do what they need to achieve a win-win situation.

Ecological strategy: integrate content resources, develop skill developers, and accelerate industry empowerment

In the consumer market, a huge content and service system can meet the changing needs of users, forming a DuerOS's important competitiveness. In the industry market, it promotes the commercialization of DuerOS capabilities, which in turn further promotes the prosperity of the consumer market.

1. Integrate content resources:

Content resources on DuerOS mainly include self-owned content, content converted by third-party content providers, and newly developed conversational AI skills:

p>

In terms of its own content and services, DuerOS has internally integrated Baidu Maps, iQiyi Video, Baidu APP, Miaodong Encyclopedia, Haokan Video, etc.

In terms of daily life services, it links to popular daily life services such as Meituan Takeaway to realize full voice ordering and payment.

In terms of audio-visual content, Xiaodu cooperates with QQ Music to obtain a large number of genuine music resources; Xiaodu cooperates with CITIC Academy to provide users with a wealth of high-quality audio books; Xiaodu integrates Discovery, VIPKID , Douyu live broadcast and other massive video resources.

2. Develop skill developers

With the popularity of conversational AI hardware devices, users’ demand for skills is increasing. The skill store on DuerOS provides a rich variety of skill types to meet the diverse needs of different users. At present, more than 800 skills have been released, covering news and information, leisure and entertainment, smart home, life services, parent-child education, financial management and office and other types.

Baidu spares no effort to promote the prosperity of the conversational AI developer ecosystem. On the one hand, DuerOS continuously iterates and optimizes the Xiaodu skills open platform, providing developers with a convenient development environment; on the other hand, it launches the "Conversational AI Skills Competition" and sets up 750,000 cash and 250,000 prizes. Tour" and a series of nationwide developer activities to recruit developers and guide development.

Awakening Journey: An offline technical salon for conversational AI developers. It has visited Shenzhen, Beijing, Chengdu, Shanghai, Hangzhou, Nanjing and other cities, and held 15 technical salons and development work The workshop attracted a total of more than 6,000 software and hardware developers and industry professionals to sign up, and the event site was also full.

No other platform in China can offer such generous developer feedback and intensive market education. Based on this, in the past year, the number of developers on the DuerOS open platform has increased more than three times, and there are currently more than 24,000 registered developers developing skills on the DuerOS platform.

3. Industry empowerment starts with chips

In order to expand the installation and activation volume of DuerOS on the device side, Baidu focuses on third-party application empowerment on mobile phones and TVs. , speakers, headphones, cars and other almost all household appliances and smart devices commonly used in life, and cooperate with upstream manufacturers to jointly empower them.

For OEM manufacturers, as early as the Qualcomm Snapdragon Annual Technology Summit at the end of 2017, DuerOS announced that it would join hands with Qualcomm to deeply support and jointly optimize Baidu DuerOS on mobile platforms including Snapdragon 845. artificial intelligence solutions.

In 2018, DuerOS once again increased its chip-level cooperation and made efforts in Bluetooth audio, as shown in:

It once again joined hands with Qualcomm to jointly create Xiaodu Bluetooth APP and DuerOS Version 3.0 Bluetooth audio product solution. The solution is based on the Qualcomm QCC5100 series low-power Bluetooth system-on-chip (SoC) and will support the design of wireless Bluetooth headsets with a range of AI functions.

Reached chip-level cooperation with Knowles Electronics and jointly developed a reference design solution for wearable devices such as headsets, earphones and true wireless earbuds based on Knowles IA610 smart microphone. It supports the lowest power consumption in the industry and can always wake up.

△ At the beginning of 2019, Baidu will hold the Xiaodu Bluetooth Alliance Summit in Shenzhen

4. Solutions are coming to fruition

Based on the core capabilities of DuerOS 3.0 , Baidu provides more than 20 cross-scenario and cross-device solutions, including screen device solutions, Bluetooth device solutions and industry solutions.

DuerOS has never deviated from its main battlefield of "screen-based conversational interaction". The screen device solution is Baidu's "industry first" solution, which enables DuerOS to empower everything on the battlefield, extending from smart speakers to smart TVs, smartphones, children's smart devices, car machines or car-mounted products and other categories.

△ Jing Kun introduced the DMA Bluetooth device solution at the 2018 Baidu AI Developer Conference

In response to the huge number of Bluetooth devices, Baidu launched the DMA Bluetooth device solution (DuerOS Mobile Accessories). Applicable devices include Bluetooth speakers, headphones, smart watches, etc. It supports seamless connection between the device and Xiaodu Bluetooth APP, and obtains DuerOS cloud services through voice interaction.

In the industry field, this is a typical application case of the hotel industry solution created by DuerOS. In July this year, Xiaodu Home (hotel version) settled in the InterContinental Hotel Sanlitun Tongying Center in Beijing. It can not only control lights, curtains, etc., but can also wake up artificial intelligence assistants with voice, call room ordering services, etc.

In terms of hotel industry applications, DuerOS industry solutions have been implemented in more than 10 domestic and foreign hotel and real estate groups such as InterContinental, Huazhu, Shimao, and OCT, and have jointly created more than 5,000 Xiaodu smart rooms.

△ Picture: In cooperation with Shimao Hotels and Resorts, every room in the Shanghai Sheshan Shimao InterContinental Hotel with a room price of no less than 4,643 yuan per night is connected to the DuerOS system.

2019: Let AI enter more people’s lives and lead the explosion of conversational AI

The growth of DuerOS is very rapid, and this rapid growth is expected to continue until 2019. Commercialization will begin in 2020.

At Baidu’s third quarter 2018 financial report conference call held on October 31, Robin Li clarified the commercialization time point of DuerOS when answering questions from analysts.

During the New Year’s Eve of 2019~2019, Baidu Quanmin Video and Baidu APP fired the first shot for 2019 with 200 million red envelopes on the stage of Zhejiang Satellite TV New Year’s Eve Party; while Robin Li brought Xiaodu was at home and appeared on CCTV's New Year's Day party, reproducing Baidu's AI conversational capabilities and bringing "Xiaodu Xiaodu" further into thousands of households.

△ Robin Li and Xiaodu appeared at the CCTV New Year’s Day Gala at home

Baidu is using “China Speed” to accelerate AI into more people’s lives, and the acceleration of DuerOS’s growth is worth looking forward to!

This expectation comes not only from Baidu, but also from practitioners and related people in the voice intelligence ecosystem, and also from ordinary people. They have always been full of yearning for technology to change their lives. They were once attracted by the black technologies flying all over the sky. It disrupted the direction, but now they have inadvertently discovered that AI has entered their lives and will subtly change their future lifestyle.