This article is adapted from the WeChat public account “QbitAI“. The writer is Li Gen and An Ni. Kr-ASIA has been authorized to publish this article.
Robin Li’s company is ahead of us in the field of AI. Tencent still lags far behind.”
The above remark was made by the founder of Tencent Pony Ma to co-founder of Baidu Robin Li, at an early-April event in Shenzhen, which was attended by all three founders of China’s big three internet giants (namely Baidu, Alibaba and Tencent, known as the BAT).
Over half a year later, the then rather modest Tencent, without much hype, has already reached into many aspects of AI, further enhancing its omnipresent role in the internet space.
In retrospect, when referring to AI at the April event, Ma also said some other things worth noting. One is that his company had three competing teams that were working on Go-savvy AI; the other is: “You can’t go far unless you have application scenarios for commercialization. Apparently, Ma’s company never lacks application scenarios.
In fact, today’s Tencent can hardly hide its ambition in AI any more, as evidenced by the “AI in all” slogan the company chanted at this year’s Tencent Global Partner Conference, which just concluded in Chengdu.
Tencent is clearly ushering a new stage of development as it threatens to beat Bytedance, Baidu and Alibaba.
Challenging Bytedance
Tencent has a new open project in the pipeline, this time centered on content. In addition to the launch of the “open content platform”, Tencent pledges to provide 10 billion yuan in revenue sharing and 10 billion yuan in industry resources, as well as bring 10 billion in user traffic to content creators, helping them with traffic acquisition and monetization, investing in their new businesses and building offline cultural and creative centers.
The platform will integrate with multiple existing platforms that handle 100 million or more visits on a daily basis, including WeChat, QQ, Qzone, QQ Browser, Kuaibao, Tencent News, Tencent Video, Tencent App Store, Now Live and Quanmin Karaoke. The goal is to build a content ecosystem that covers everything from content production to channel building, intelligent distribution and commercialization.
According to Lincker Lin, Vice President of Tencent, AI-powered content distribution will provide technological support for the company’s content project. The platform has been put in place, through which Tencent expects to retain its leading position in the battle of digital content, Lin added.
It can’t be more obvious at this point. From the mid-year reorganization of its Online Media Group to the recent adoption of an AI-powered strategy, investment of content resources and construction of a comprehensive content ecosystem, Tencent has one clear goal: challenging Bytedance.
As the earliest beneficiary of the rise of news distribution in China, Bytedance continues to see bright prospects. It completed a new funding round at a valuation of more than $10 billion this year, with the aim of realizing 30 billion yuan in annual advertising revenue while working towards the 50 billion yuan target.
Next up is Baidu. The company just brought content distribution back on its agenda earlier this year. The result is an estimated annual revenue of $1 billion from in-feed ads, which will account for one third of the giant’s total revenue, according to its latest quarterly report. “It’s an impressive milestone,” said the founder Robin Li.
For Tencent, however, the battle has long begun. In addition to a solid background in content distribution, the deep-pocketed company boasts a huge user base and plenty of revenue generation channels, thanks to products and services like Tencent News, Kuaibao, Penguin Platform, WeChat official accounts. For Tencent, turning profits from content is just a matter of time.
The only uncertainty at the moment is: What change exactly will Tencent’s “3 x 10 billion” project bring to the AI competitive landscape?
Foray into driverless cars
While in the digital content segment, Tencent’s target is Bytedance, in the field of driverless cars, it’s vying with Baidu.
Tencent has officially announced its AI-centered “new technology” platform strategy and proposed an implementation roadmap that contains basic research, construction of application scenarios and opening of AI technologies.
The company’s driverless car project has loomed large, finally, as well, suggesting that Tencent is entering a field that was long thought to be Baidu’s forte.
Su Kuifeng, Director of Tencent’s Autonomous Driving Lab, made his first public appearance at a forum on automobiles and discussed Tencent’s progress on driverless cars.
The day before, news that Tencent had joined the autonomous driving race sent the stock of the Hong Kong-listed company soaring.
Su briefed the audience on Tencent’s views on autonomous driving, its driverless car plan, research and development progress and prospect for commercialization.
Some of the core issues Tencent sets out to tackle, according to Su, are navigation and positioning, sensor technologies and prediction and decision-making algorithms, technical challenges that must be addressed to realize fully autonomous driving.
Tencent has already worked out a preliminary plan and established a basic framework to tackle these issues. It will start from high-precision maps, leverage data collected from actual application scenarios and use simulation platforms to continuously train and perfect its prediction and decision-making algorithms and improve its sensor technologies.
With respect to technology architecture, Tencent will dedicate itself to connecting vehicles to the cloud through technologies like 4G networks and short-range hotspots. It will also work to achieve data synchronization for some key application scenarios.
Su did not say much about hardware, except that the company will have no particular preferences for algorithm and data-related hardware since algorithms have become the bedrock of autonomous driving. Tencent cares more about data in actual applications and data aggregation is its main task at this stage. As for hardware, it prefers to work together with its partners to build platforms connecting vehicles and the cloud to enable connected driving.
Su also talked about Tencent’s plan for commercialization. At this stage, it will concentrate on developing Level 3 products while working on algorithms for level 4 and Level 5.
Level 3 solutions for scenarios like expressway driving are already possible with current technologies, said Su. Tencent will start from there and develop level 3 products in collaboration with its partners.
Fully autonomous level 4 and 5 vehicles, on the other hand, are set to shape the future of technologies. But at the moment, Tencent’s focus for level 4 and level 5 autonomous driving will be mainly on developing algorithms and building databases.
In summary, Su said Tencent will continue to put users’ demand first, give priorities to products, and work along both lines (level 3 and level 4/5 driverless cars).
This probably sounds familiar, as it’s exactly what Baidu is doing with its Apollo program.
Ever since Lu Qi took office as Baidu’s COO and was put in charge of the company’s Intelligent Driving Group, Baidu has worked out a model that enables it to concentrate on developing algorithms while its partners build ecosystems and hardware. As of now, Baidu has established cooperative ties with major industry partners including auto makers and transportation service providers.
The question now is: What impact will Tencent’s foray into driverless cars have on the industry and on Baidu in particular?
AI personal assistant
Apart from driverless cars, Tencent seems poised to compete against Baidu for an edge in the field of speech recognition too.
At the Tencent Global Partner Conference, Tencent also launched its first AI personal assistant – Tencent Dingdang. According to Chen Qian, General Manager of Tencent Intelligent Platform, the product is not just about voice interaction. The hope is that it can combine voice with visual and audio features and enable highly effective interaction in actual application scenarios by incorporating more into the traditional mobile phone-based voice interaction.
Dingdang will be used to meet a variety of demands. Specifically, the intelligent assistant will help connect hardware devices with AI networks to help them see, hear, understand and give replies.
For instance, for a simple application like photo taking, while intelligent functions were restricted to “taking pictures by voice commands” or “auto shooting after smile detection” in the past, Dingdang, as was shown at the conference, allows users to instruct the camera to track sounds and take pictures automatically. It also supports functions like auto-focus and facial recognition, realizing full automation.
Tencent Dingdang is also able to handle and switch between multiple applications. Imagine you are listening to an audiobook at home. With Dingdang, you can instruct the speaker to send the book to your phone and even make purchases by voice commands. Tasks that are originally performed by a speaker and a book reader separately are thus integrated and carried out with a single command.
To bring this further, Tencent Dingdang plans to construct a new AI hardware ecosystem by offering a standard interface that can be used to connect with everything.
Hardware technology firms including Philips, ZTE, GAC, Xiaomi, Road Rover, Codoon, Meizu and Pace became the first to “like” the new product at the conference.
According to Tencent, manufacturers can make their products “intelligent” directly through Tencent Voice Services, much like the DuerOS project Baidu is aggressively pushing currently.
Although Dingdang is still just a personal assistant for now, it’s only a matter of time before Tencent launches its own speech recognition operating system and opens platform. Meanwhile, QbitAI learned that the smart speaker Tencent has been working on is also in the final stage of development.
It seems that a battle between the two giants in the two AI areas of driverless cars and speech recognition is already inevitable.
Tencent’s ambitious AI scheme
Tencent’s appetite for AI, it seems, is still far from being satisfied with self-driving cars and AI voice assistant.
Tencent unveiled its ambitious AI scheme this time at the forum.
It is now clear that its AI ambition reaches into three arenas.
The first focus on its list is AI research, the results of which will be made available to both Tencent and other organizations.
A year ago, Tencent set up its AI Lab, which it has advocated as a hub for “delivering influential research results and yielding concrete benefits”. It has proved that the slogan is no empty talk. So far, over 80 papers presented by Tencent have been admitted at such top-level academic conferences as CVPR, ACL, ICML, NIPS and the others. This is pretty remarkable compared with the other companies in China. Meanwhile, Tencent’s facial recognition dataset in MegaFace Challenge2 and MS COCO also finds no rival.
Tencent’s pursuit for technology is evidenced through not only the apps we use daily including WeChat, QQ, QQ Music and Tencent Video, but also through its other key operations including games, content and social media. It deems technology as a weapon to improve its game experience, AI application in content, interpretation of image and video, content generation and feeds as well as interaction with terminal hardware in social networking.
The next move is to expand AI application to benefit small and medium-sized companies dedicated to AI and traditional industries.
To this end, Tencent will make the algorithms, data and models for realizing basic AI capabilities available to its partners and small and medium-sized companies in AI industry on Tencent’s open platform. In the meantime, API and SDK package will also be provided for other developers to hook into its AI applications.
In addition, Tencent also intends to expand AI application in traditional industries through its “Industry AI+X” program and nurturing startups in AI industry. Take Tencent Miying (literally “Tencent Imaging Seeker”) for example, Tencent Miying is an AI-assisited medical imaging product launched by Tencent earlier which can detect early esophageal cancer with an accuracy of 90%. At the forum, Tencent also revealed its “AI Ecological Plan” (the Plan) which seeks to nurture startups in AI with the extensive application scenarios of its AI technology, technology, talents and capital. To Tencent, an AI ecosystem is built not by Tencent alone, but by allying with its partners.
The third move is to foster talents and industry-university-research cooperation. Tencent aims to foster talents in AI by taking advantage of its resources and tested results in AI industry. To that end, Tencent has initiated a fellowship program for PhD students worldwide. The selected PhD candidates will be granted handsome scholarship and the research opportunity in Tencent AI Lab. In that case, the candidates can get their hands on real demands and reliable data, thus fostering fully-prepared talents for AI industry.
As it is pointed out by Zhang Tong, Director of Tencent AI Lab, Tencent’s AI scheme is intended to pave the way for industry-university-research cooperation. It believes that world-class AI development capabilities can only be cultivated through concerted efforts. Technology has the potential to address, in a creative fashion, all issues and challenges faced by human. Tencent’s ambition for AI has already been explicitly demonstrated in its slogan, Make AI Everywhere.
Guided by the concept, Tencent rolled out its AI strategy, i.e. AI Ecological Plan. The Plan, as the first detailed scheme for AI development, has kept pushing Tencent forward in the last seven years.
Tencent hopes that, through the implementation of the Plan, it can share 100 AI technologies, nurture 100 startups in AI industry, launch 300 “Cloud+ One Million Yuan Assistance Plan for Startups” programs and empower 1,000 partners. Yet, this is only the beginning.
Tencent’s “AI Ecological Plan” reaches far beyond. It hopes that, eventually, it will make the extensive application scenarios of its AI technology, technology, talents and capital available to all partners so that they can thrive together.
Tencent’s AI footprints can now be found in such key areas as social networking, content, game, health care, retail, finance, security and protection as well as translation.
To be clear, retail is Alibaba’s kingdom where Tencent doesn’t intend to flex its muscle.
Mark Ren, COO of Tencent, explained Tencent’s move towards retail in his keynote speech. He said that sellers on e-commerce platforms have nothing to fret about as the “Smart Retail Solution” doesn’t, in any way, amount to Tencent’s entry into e-commerce. The solution only seeks to empower brand holders, retailers and real estate agencies by way of “decentralization”.
Ren also added that Tencent strives to offer tailored solutions to retailers with extensive application scenarios, big data, AI technology and its full suite of products. A digitalized and intelligentized physical store will help remove the space barriers between products and customers.
No matter how Tencent sugarcoats its move towards retail, it still can’t deny the fact that its plan overlaps with the “New Retail”, “Smart Connection” and “Online and Offline” strategies laid out by Alibaba.
No wonder some senses a bit “Say Hi Jack Ma” challenge.
Tencent’s ambition for AI has been demonstrated clearly enough in its 225-page report. The question is, will the penetration of its AI technology, as set out in its report, overtake that of Baidu, which has, so far, registered mediocre performance?
Well, Robin, CEO of Baidu, certainly won’t agree with that. To your special notice, the 2017 Baidu World will kick off on this Thursday in Beijing.
Let’s hear out what Robin Li’s got to say.
Writter: Li Gen and An Ni from QbitAI