The United States Patent and Trademark Office (USPTO) uses neural networks to validate patents related to neural networks, which have skyrocketed in recent years.
The US Patent Office trained its own neural network on a large array of patent data on neural networks. The new neural network analyzes the text of the patent and citations in all US documents filed since 1976.
She is also studying the work of neural networks in the patent, which may relate to knowledge processing, speech, AI hardware, evolutionary computing, natural language processing, machine learning, and planning and control.
The employees checked the results of the new neural network manually and did not find any errors. This approach, the department notes, seriously increases productivity compared to other alternative methods of information processing.
From 2002 to 2018, annual patent applications for neural networks in the United States increased by more than 100%, from 30,000 to over 60,000 per year.
Neural networks are becoming an increasingly popular technology – if in 2002 they appeared in only 9% of patent applications, now their share is 16%. most of the applications come from IBM, Microsoft and Google.