Went through this analysis of 2,398 generative AI patents filed globally from 2017 to 2023 breaks down into 6 categories. The biggest category was financial and information security applications at 22.8% of all patents. Image generation and processing comes second at 21.7% and medical applications comes after at 14.6%.
Conversational AI, the chatbot offered by OpenAI/Antrhopic/Meta/Google etc actually came at fourth place.
Object detection, crucial for autonomous vehicles and robotics, is declining in patent activity (Not sure if tech has matured, or they can't see the ROI anymore). So is financial security despite being the largest category. Image generation peaked around 2021 and dipped after 2022.
On the other hand, medical applications and cyber physical systems are growing fast.
What I want to say here is that if you're long on AI, the patent trends suggest medical and cyber physical applications have more room to run. The financial security space might be mature or hitting diminishing returns. Image generation looks like it's consolidating.
One last thing, the analysis found all correlations between application categories were negative. What this means is that innovation is siloed and not interrelated. In simpler terms, a breakthrough in medical imaging doesn't transfer to financial fraud detection even though both use similar neural network architectures. So GenAI isn't one market, it's 6 separate markets that happen to use related technology.
If you're investing in AI through geographic ETFs or regional indexes, this ranking tells you where actual innovation is concentrated versus where the headlines come from. Not investment advice obviously, but patent trend analysis is one of those alternative data sources that might give you an edge before quarterly reports catch up.
Source (Open Access) - [https://www.sciencedirect.com/science/article/pii/S2444569X24000702](https://www.sciencedirect.com/science/article/pii/S2444569X24000702)