Will Google Repeat an Android in Artificial Intelligence?

Reading Time: 4 minutes
Rate this post

Artificial Intelligence is making headway in more ways than one, entering into mainstream everyday life across quarters. As we near the end of 2015, I can look back to find three very prominent applications of Artificial Intelligence (AI) that have been implemented by three industry giants.

• In Jan-Feb 2015, Facebook open sourced its optimized deep-learning modules for Torch which is a scientific computing framework.
• On 9 November 2015, Google open sourced its Artificial Intelligence (AI) deep learning platform, Tensorflow.
• On 12 November 2015, Microsoft released the Distributed Machine Learning Toolkit (DMTK), which comprises of a parameter server-based programming framework and two distributed machine learning algorithms.

Until recently, data scientists were relying heavily on frameworks provided by universities like Caffe from University of Berkley (a deep learning framework) and Theano from University of Montreal (Python packages for deep learning) to solve AI problems. Though corporates were employing in-house frameworks, it is the first time that they have gone the open source way to attract more people to their platform.
Considering that the AI industry is gaining visibility on a larger scale now, the patent landscape is looking much more interesting. The Thomson Innovation’s DWPI classification was considered for analysis in this blog.

Patent Trend Analysis

There are 3,231 patent publications relevant to AI worldwide. Considering that it takes some time for each patent office to publish its applications, the last few years dip in the number of patents filed should be taken with a pinch of salt. However, one cannot discount the fact that this may be a genuine trend, considering it is a very young technology being discussed. While a reasonable number of patents have been filed starting 1996-1997, there have been crests and troughs. However, the trend is holding up.

AI1

Japan had a golden run between 1996 and 2000. US has been steadily filing patents in AI, which indicates an organic increase. If this trend continues, one can expect US to surpass Japan in terms of number of patents filed.

AI2

Analysis of Top 15 Patent Owners

Patent Ownership & Distribution Analysis

An individual inventor “Son Y S” has filed for 39 patents in South Korea. The list is dominated by Japanese companies. Though it is surprising that more US companies are not in this list, it may be explained by the fact that Japan focuses considerably on robotics which needs AI models to drive their decision making capabilities. IBM and Microsoft figure among the Top 5, confirming the R&D prowess in their DNA. Interestingly, Google ranks 26th with 16 patents.

AI3

An interesting point that permeates from the heat map is that even though most companies listed here are multinational corporations operating across different geographies, the filing of patents seems to be in their place of origin.

AI4

 

Patent distribution among assignees

The fragmented ownership (long tail) does not come as a surprise and this confirms that the industry has still got time to go before a clear winner emerges.

AI5

Geographic Distribution of Patents

Japan has pipped US to the top primarily on the basis of Japanese conglomerates’ considerable research in Artificial Intelligence. Interestingly, Australia has also made its entry into the top five geographies where Artificial Intelligence patents are filed.

AI6

Will Google repeat an Android in AI?

While Japan has been a pioneer in this space, the US is catching up to establish dominance. The market is fragmented and there is no clear winner in terms of an assignee having considerable IP assets as yet. Each company is primarily focusing on its home market at the moment which may mean that commercialization of these inventions may be still work in progress. Once the market develops, one can expect multiple continuations of the early and seminal patents.

There are not too many US companies that are extensively focusing on research and converting them into IP assets, except for the usual suspects – IBM and Microsoft. IBM has a predictive modeling platform called IBM Watson, which should be in a sweet spot based on the IP protection that it gets. However, not much can be said about other companies. The surprise has been Google, which has been pretty active commercially in this space, but owns limited patents.

To draw parallels with historical events, before Android was released and open sourced there were numerous mobile operating systems and the market was fragmented. The advent of Android led to a consolidation of mobile OS platforms. Today, we are at a junction where there are three most relevant mobile operating systems. Android was the winner in terms of number of users, but it also opened up a Pandora’s Box of IP litigations against it and its OEM partners. While it has managed to plug the hole by licensing patents and acquiring a portfolio from Motorola, the real question is whether Google has learnt its lesson the hard way.

Based on open sourcing of Tensor Flow, Google seems to have gone the open source way to initiate and subsequently establish its dominance in this space. However, it is too early to say if they have learnt their lesson from the Android experience.

(Featured image source: https://pixabay.com/en/artificial-intelligence-technology-3262753/)

Sidharth Vishwanathan
Sidharth Vishwanathan

Sidharth is a technology expert and specialist in product and source code analysis. Talk telecommunications, and it piques Sid's interest, translating into interesting observations on the industry on this blog.


1 Comment

  1. Very interesting read Sid! Any information on what topics of AI make up the portfolio? Is it mostly machine learning /neural networks stuff or does it include a good amount of robotics as well? Will be interesting to see how much of Japan’s patents are related to machine learning and which to robotics.

Post a Comment

Your email address will not be published. Required fields are marked *