IBM chases most genAI patents, ChatGPT owner holds back
New figures show which companies are securing most patents in this fast-growing technology field | IBM filed 500+ more patents than Google in past five years.
The race to claim patents in generative AI (genAI) is being won by IBM, according to recent statistics from patent data provider IFI Claims, with the US company filing a third more applications than its nearest rival Google over the past five years.
IFI’s analysis, released on Tuesday (February 6), focuses on the artificial intelligence subset of GenAI—deep-learning models that take data and ‘learn’ to generate outputs of content such as text, pictures or videos.
ChatGPT maker OpenAI ranked outside the top 25 applicants for genAI patents with IFI Claims finding fewer than five patents for the firm.
While “surprising for an organisation that is knowledge-based”, this could indicate filings that aren’t public yet or a reliance on trade secrets to protect its IP, the data provider suggested.
Focus on ‘neural networks’
Of more than half a million AI-related patent applications filed in the US over the past five years, IFI Claims estimates that around 22% are linked to genAI.
The leading applicants in the field are IBM, with 1,591 applications over the past five years; Google (1,037); Microsoft (695); Samsung (609); and Intel (396).
According to IFI’s analysis, the patent class being used to secure genAI inventions the most is ‘computing arrangements based on biological models’ with a focus on deep learning technologies, particularly convolutional neural networks.
“Such networks help machines see and interpret images, for instance, as humans do, which makes this patent class [a] cornerstone to genAI,” said IFI.
Other patent classes key to securing genAI technology included ‘pattern recognition’, ‘image or video recognition or understanding’ and ‘information retrieval’.
With their patent strategies, IBM, Google and Samsung were covering the four bases of content generation—images, speech, text, and video, according to IFI.
Other firms were focusing on particular areas, such as Nvidia targeting images and video with its patent applications.
No consensus on copyright and AI
Yesterday (February 7), the UK failed in its bid to establish a voluntary AI code as talks between rights holders and AI developers broke down, according to a government announcement.
The working group, convened by the UK Intellectual Property Office to discuss the “interaction between copyright and AI”, failed to reach an agreement.
This followed a statement by OpenAI that it would be “impossible” to train AI models without using copyrighted materials, in a submission to the UK House of Lords’ inquiry into large language models.
