A Force for Good or Daring to Be Evil?

An Analysis of Google's Patent Policies on Open Source AI Software (1)

2018.11.19Daeho Lee

The rapid development of deep learning technology is amazing, and open source is credited as one of the main reasons for such rapid development. In the software field in general, open source is already widely regarded as an important driver of innovation, and companies such as Apple, Samsung, and Qualcomm, who traditionally have relied on the existing patent paradigm to protect their technology, are also making efforts to create and foster open source communities.

In the field of deep learning, it is difficult to discuss innovation through open source without considering the impact of Google's “TensorFlow” open source software. Many researchers and engineers use the TensorFlow framework for their implementations of deep learning and as a basis for novel research.

To a certain degree, the patent, at least in the software field, seems to have become a relic of the past.

However, the fact that Google is registering key patents related to deep learning, despite its promotion of open source software, is a controversial topic in the open source community, as can been seen, for example, in various discussions on Reddit and Facebook.

Google has been steadily building an extensive portfolio of patents related to deep learning, including patents on essential technologies such as DQN, dropout, and batch normalization, which are commonly used by deep learning researchers and engineers. The range of Google's deep learning patent portfolio is diverse, covering core deep learning technology and applications employing deep learning.

Below are some of the more significant patents among Google’s portfolio:

- Methods and apparatus for reinforcement learning (DQN)

- System and method for addressing overfitting in a neural network (dropout)

- Batch normalization layers (Batch normalization)

Open source, which is founded on the principles of sharing and freedom to use technology, and patents, which grant a temporary monopoly over a certain type of technology, are concepts that fundamentally contradicting each other in nature. What motivates Google to pursue two contradicting strategies at the same time? Will TensorFlow users be safe from patent lawsuits filed by Google? Do open source licenses provide users a free pass to ignore patent infringement issues?

In this series of articles, we will attempt to answer these questions. First we will take a look at the open source license and how it relates to patents. We will then consider a few hypothetical scenarios and try to identify Google’s real intent. Finally, we will take a look at the implications of all this on any patents owned by users ofTensorFlow.

Stay tuned for our next update!

About the author

Daeho Lee

Daeho started his career as a patent attorney at Seoul firm Park, Kim & Partner, where he worked on ...

Read more

Contributors for this article: