WIPO recently published the WIPO Patent Landscape Report on Generative Artificial Intelligence (GenAI)

Author Tainara Barbosa

WIPO recently published the WIPO Patent Landscape Report on Generative Artificial Intelligence (GenAI), which provides an analysis of global patent filings related to GenAI technologies. It highlights trends in patent filings, key players, and innovations in various sectors such as GenAI, GenAI models, modes, and applications.

The report emphasizes the rapid growth of GenAI, in patent family and scientific publication numbers, particularly since 2017, driven by advances in computers, the availability and use of large datasets as training data, and enhanced algorithms.

The European Commission defines GenAI as a type of foundation model, which is general purpose AI models trained on large and diverse datasets, used to generate content such as text, images, audio or video.

It utilizes machine learning models, particularly deep learning architectures like neural networks, to produce these outputs autonomously. GenAI is distinct from traditional AI as it can generate novel data based on learned patterns. Key GenAI models include generative adversarial networks (GANs), large language models (LLMs), (variational) auto-encoder, autoregressive models, and diffusion models. Moreover, the GenAI modes are the types of input/output data used, such as image, video; text; speech, sound, music; code; molecules, genes, proteins; and 3D image models.

GenAI patent applications represent approximately 6% among all the AI patent applications, from 2014 to 2023. GenAI grew from 4.2 to 6.1% from 2017 to 2023. Referring to scientific publications, it grew from 11517 to 34122 from 2022 to 2023, which represents an explosion after the 2022 release of ChatGPT.

The Top 5 patent owners in GenAI are dominated by China, being Tencent Holdings (China), Ping An Insurance Group (China), Baidu (China), the Chinese Academy of Sciences (China), and IBM (US). While the Top 5 entities with scientific publications is: the Chinese Academy of Sciences (China), Tsinghua University (China), Stanford University (US), Alphabet (US), Shanghai Jiao Tong University (China).

Referring to GenAI models, the GAN model is the one with the most patent applications filed, followed by VAE and LLM models. Each one with, respectively, approximately 9700, 1800, and 1300 patent families. When analyzing the GenAI modes, the image/video category is the one with the most patent families currently, followed by text, and speech/voice/music.

Moreover, the application types that dominate the patent application are: “software/other applications”, followed by “life sciences”, and “document management and publishing“.

The WIPO Patent Landscape Report on Generative AI highlights the rapid growth in AI innovation, driven by major players like Tencent. With significant patent activity in regions like China, the U.S., and South Korea, GenAI is transforming several sectors, such as life sciences, manufacturing, software, transportation, and security, among others.

Companies are investing heavily in AI technologies that push the boundaries of creativity and problem-solving. This report underlines the global competition and the increasing significance of GenAI in diverse industries. To access the report in its entirety, please see: Patent Landscape Report – Generative Artificial Intelligence (GenAI).

For more information, do not hesitate to contact us at international@simoes-ip.com