Are these the patents from Google’s & Tencent's licensing deal?

Google_Search_AI_Software.png It was announced on 19 January 2018 that Google and Tencent have exchanged friendship bracelets made of strings of patents. If you’re anything like me, you might have wondered exactly which patents were involved in this deal and why?

Well, this article will attempt to answer that question in a way that’s either completely unhinged or respectably novel. But before I unleash my Pinky and the Brain routine, it’s worth asking, “What do Google and Tencent hope to get out of this deal?”

Sundar Pichai recently said, “…I don’t know what the answers are,” on the subject of how Google can ignite meaningful business relationships in China. And Tencent was one of the companies tipped to benefit the most from Google’s China exit, but the two companies now seem on converging paths.

As explained by the Bloomberg article to which I’ve linked above, this deal is likely about intelligence—the artificial kind, mixed with a whole lot of street smarts (on Google’s part).

Since Google withdrew its search engine from China in 2010, it’s been in a quandary about how to exploit this huge market and generator of innovation. The Chinese government may have launched a cyberattack on Google, but the people of China (especially the engineers) still love Google’s technologies.

Particularly, Google’s TensorFlow AI project—an open source machine intelligence software library—has been a hit in China. According to Bloomberg, “It’s hard to find a place as fertile for AI as China. The country has one of the fastest growing TensorFlow developer communities in Asia, despite the fact that Google’s cloud services are unavailable there.”

The same Bloomberg report states, “Google has pitched Alibaba and Tencent Holdings Ltd. on using TensorFlow, an attempt to spread the software through the ranks of China’s largest tech firms…”

We also know that Tencent established its AI Lab in 2016—it is described as having “a vision of (sic) ‘Make AI Everywhere’,” and being focussed “on both fundamental research and practical application (sic) of artificial intelligence.” The nascence of Tencent’s AI programme is reflected in its patent coverage—the company’s AI-related patent count is about 6% of Google’s.

Click image to enlarge

Google and Tencent AI PatentsA comparison of Tencent's and Google's AI-related patents (source: PatSnap platform)

Google opened its own AI lab in 2017 and has previously said that deals such as the one with Tencent “reduce the potential of litigation over patent infringement”. Finally, Google has joined its new licensing partner’s and Sequoia China’s move to invest in XtalPi—a US-China biotech firm that uses artificial intelligence to speed up new drug development

It’s fair to say, this licensing deal probably involves several patents focussed on artificial intelligence—Google’s Trojan horse into China’s firewalled economy.

Based on the above analysis, I see 3 main reasons why Google and Tencent may have agreed this deal:

  • Google wants to prevent patent infringement issues, as it accelerates patent filings in China
  • Google wants to embed its IP and technologies in Chinese products, as a backdoor to Chinese consumers
  • Tencent wants to boost the capabilities of its young AI lab, using Google’s technologies

I will propose 3 sets of patents likely involved in this deal, based on these 3 scenarios and patent data analysis.

Scenario #1: to prevent patent infringement issues

A quick look at Google’s patent footprint in China, since it left in 2010, reveals an acceleration in filings—especially in the IPC G06F17 (Digital computing/data processing/methods). Many of the patents filed are focussed on AI technologies and Google already has a track record of citing Tencent's patents.

Click image to enlarge

Tencent Patents Most Citing CompaniesThe list of companies most frequently citing Tencent's patents includes Google (Source: PatSnap platform)

Since Google just opened an AI lab in China, it’s safe to assume it plans to increase its AI-related investments and patent filings within the Chinese jurisdiction. This increases the risk of patent infringement.

For confirmation, one need only look at a landscape with Google’s (red) and Tencent’s (blue) AI-focussed patents mapped on top.

Click image to enlarge

Google and Tencent AI Patent LandscapeA landscape featuring Google's and Tencent's AI-focussed patents (Source: PatSnap platform)

Focusing only on filings made in China, we can see Google’s patent CN106489146A (Using the session information query rewriting)—and a couple of others—is dangerously close to Tencent’s CN104915340A (Natural Language Question-Answering Method and Device).

Click image to enlarge

Google_and_Tencent_AI-Focussed_Patents_China_LandscapeA landscape featuring AI-focussed patents filed in China, by Tencent and Google (Source: PatSnap platform)

Both patents involve the application of natural language processing algorithms for analysing and responding to user-generated questions. Their proximity isn’t just cosmetic, it’s also technical. If Google intends to continue innovating in China, it would make sense for the company to negotiate a license for Tencent’s patent.

On the landscape is another Tencent patent valued at $67,000, which the company already licensed out in 2016—CN102801548B (Method, device and information system of intelligent early warning). This shows Tencent’s willingness to share this patented technology and demand within the market for it.

So far, we have two patents potentially licensed to Google: CN104915340A (Natural Language Question-Answering Method and Device) and CN102801548B (Method, device and information system of intelligent early warning).

Scenario #2: to embed Google’s patents and technologies in Chinese products

Let’s begin by figuring out the areas of technology where both Google and Tencent have been placing big bets. It’s reasonable to expect the companies will optimise areas in which they’re spending more and more money, not less and less.

Since patents (and the technologies they protect) tend to be expensive, patent filings are often a good indicator of R&D investment. If we look at technology areas where Google has increased patent filings in China since 2010 and the areas where Tencent has done the same, we see some similarities.

Click image to enlarge

Google Top Patent IPCs China Top IPCs for Google's patent filings in China (Source: PatSnap platform)

Click image to enlarge

Tencent Top Patent IPCs Top IPCs for Tencent's patent filings (Source: PatSnap platform)

Here are the top overlapping IPCs for both companies:

  • G06F17 (Digital computing/data processing equipment/methods)
  • G06F3 (Arrangements for transferring data)
  • G06F9 (Arrangements for programme control)
  • H04L29 (Apparatus/circuits/systems, not covered by a single one of groups)
  • G06F21 (Protecting computers/programs/data against unauthorised activity)

I constructed a search query identifying Tencent’s and Google’s patents within these IPC groups, then I pulled those patents into PatSnap’s licensee locator tool.

Click image to enlarge

Google and Tencent Licensee Locator CompaniesPatSnap's licensee locator tool (Source: PatSnap platform)

The licensee locator tool uses AI algorithms to find patterns and similarities within any list of patents, down to the IPC subgroup and textual level. It then recommends a list of potential licensees, based on the similarity of their patented technologies to those in the original list—and other criteria such as licensing history, licensability score and more.

Unsurprisingly, Google and Tencent appear in the list of top 25 potential licensees—alongside companies like Salesforce, Huawei and the omnipresent IBM.

This step basically helped me increase the depth of technological overlap between the list of patents I used for the following step of analysis.

Next, I copied the list of Google patents identified as being relevant by the licensee locator tool into Insights by PatSnap—then I filtered the list to show only patents filed in China, which discuss AI technologies in their titles, abstracts or claims.

Finally, of this segment of patents, I collected those with the highest valuations:

This gives us another 2 patents.

Scenario #3: to boost the capabilities of Tencent’s AI lab

I took a straightforward approach with this scenario and assumed Tencent would want to license in Google’s most sought-after AI patents, filed in China.

So, I constructed a search query to show me AI-related patents filed by Google, in China. Then I sorted the list by patent valuation (highest to lowest) and picked the top 6.

Click image to enlarge

Google China AI PatentsHigh-valuation Google patents filed in China (Source: PatSnap platform)

The methodology which assigns valuations to patents within PatSnap employs 26 indicators, covering the robustness of the R&D poured into an invention, technology trends over time, heterogeneity of applications, evidence of use and more. The valuation amount is generally a decent proxy for judging the desirability of a patent (except for situations where a patent is too young or too niche to have built up any meaningful indicators).

The process above yielded the following list:

Final list of 9 patents

At the end of my investigation, I’m left with a list of patents that looks like this:

  1. CN104915340A (Natural Language Question-Answering Method and Device)TENCENT
  2. CN102801548B (Method, device and information system of intelligent early warning)TENCENT
  3. CN106250508A (Answering questions using environmental context), which has a valuation of $194,000—GOOGLE
  4. CN105045799A (Searchable index), which has a valuation of $18,000—GOOGLE
  5. CN105210064A (Classifying Resources Using a Deep Network), valued at $1,900,000—GOOGLE
  6. CN102227724B (Machine Learning for Transliteration), valued at $1,700,000—GOOGLE
  7. CN102549603B (Relevance-Based Image Selection), valued at $530,000—GOOGLE
  8. CN105160397A (Training Distilled Machine Learning Models), valued at $18,000—GOOGLE
  9. CN105719001A (Large-Scale Classification in Neural Networks Using Hashing), valued at $18,000—GOOGLE

It’s not the largest list but I thought it would be better to err on the conservative side—I set out to show only relevant patents, not all relevant patents. Regardless, I hope it’s been informative to show the ways a more well-informed person, armed with insider tips and the right tools, might go about spying on their competitors.

And if we have any insiders reading, please get in touch and let me know if my analysis has been more reminiscent of Pinky or The Brain.

Download the Technology Transfer Winner's Blueprint 

  • Case studies of technology transfer success stories
  • Interviews with experts from industry and academia
  • Pointers as to how tech transfer works in the cybersecurity sector
  • Insight from ex-president of Association of University Technology Managers (AUTM)
  • Visualisations of intellectual property data, revealing tech transfer patterns

eBook TTO expert tips