A study of patenting trends in the field of artificial intelligence (AI), looking at filing figures for different areas of AI, top filers and filing jurisdictions.
Artificial intelligence, or AI, is a term used to describe an element of intelligence exhibited by computers or machines. One area of AI which has seen significant development since 2012 is machine learning, which involves the development or training of an algorithm or statistical model to enable a computer to perform a particular task without requiring explicit instructions. A trained machine learning model is instead able to rely on pattern recognition and inference to perform the task in respect of which it is trained.
In recent years, machine learning and, more generally, artificial intelligence, has been used for tasks in a wide range of technological fields. Machine learning can be used to train computer models to perform tasks that would otherwise be performed by human operators, and using computers in this way can improve workflows and reduce the workload of their human counterparts. This can be particularly advantageous in fields where human workers are under pressure to perform tasks accurately and quickly.
Depending on the task for which the computer model is to be trained, a machine learning model may be provided with various inputs to achieve the desired output. In the medical field discussed above, data may be provided to the model in the form of text (e.g. medical reports or test results) or images (e.g. imaging scans). For other models, data may be provided in other forms, such as spoken words via a human's voice (e.g. when you speak to a smart speaker).
Patenting Trends in AI
HLK Analytics, the patent searching and analytics function at Haseltine Lake Kempner, has been taking a look at patenting trends in the field of AI. One area of our research has focused on the form in which the training data is provided to the computer model. Worldwide published patent documents (e.g. published patents and published patent applications) in the field of AI were classified into three categories, depending on the form of the data used for training the model: images, text or voice. Other types of data used for training models that do not fall within the images, text or voice categories have not been included. Classifying the patent documents in this way corresponds with the European Patent Office's Cooperative Patent Classification (CPC) system used for classifying patent applications on filing.
It should be noted that data relating to patent applications filed in 2018 is incomplete, since patent applications typically remain unpublished until 18 months after they are filed. It should also be borne in mind that our research covers a particular set of classification headings, so AI-related applications that are not classified under those headings will not be included in our data.
To download click here – HLK IP Analytics Newsletter – Patenting Trends in AI March 2020
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
