The Next Generation of AI Technology
What patent trends can tell us
Recently, Facebook announced they will delete their facial recognition files and terminate their automatic image recognition program. At the same time, they have rebranded as Meta and plan to focus on virtual reality. This may be somewhat surprising, but not if you consider the 169 patents they had already filed in the Optical Elements category by early 2020, evidence of their commitment to exploring the augmented reality space.
Apple has been investing in the development of electric vehicles, now you can view a speculative look at what an Apple car could look like, based upon a number of patents filed by Apple. Not all patents are intended to be used, some may be filed just to protect a concept from being patented by others. While usage may not be guaranteed, a review of AI patent trends gives some indication about what areas of technology are generating research interest and corporate investment.
In my previous article on patents, I searched the patent database for key terms such as “machine learning”, and reviewed the leading companies in select patent categories with a large amount of AI related patents. Around the same time, the US Patent Technology Office released a comprehensive dataset of patents categorized into eight AI component technology areas with significant activity related to artificial intelligence. This data was also summarized in the report Inventing AI, which detailed a comprehensive effort to identify and classify every patent related to AI technology.
AI patents crop up in a wide variety of patent categories, creating complexity in defining and tracking what AI is. The eight types of AI defined here span across many categories of the official patent classification hierarchy. The AI categories of knowledge processing and planning/control represent the largest amount of activity, but all categories are seeing growth over time.
Every industry is now looking for ways to optimize their processes and increase productivity through the usage of AI. Patent activity for AI inventions is now present in over 40% of patent subclasses. The most frequent assignees continue to be very familiar names, IBM, Microsoft, and Google. By 2018, annual AI patent applications reached more than 60,000 annually, and 2021 is on track to finish with over 100K AI patent applications.
Searching for patents, using https://patents.google.com, or another search engine, allows you to examine patents for particular techniques, by certain companies, or for particular inventors. For example, my search for knowledge graphs + fraud, in US patents, indicates the activity of payment processors in patenting technology in this area.
Conclusion
Review of patents can give insight into how companies are creating their algorithms and logic for core business processes, what they value, who is working on what, and what technology is on the rise.
