Ostensive definitions for legally correct analysis of design

Explainable AI provides legally valid design/trademark identification through the use of ostensive features, creating a robust foundation for AI Gemini's use in intellectual property protection.

Comment

Inspiration

Design patents are a type of industrial design right. Examples of objects protected by design patents include decorative elements of jewelry, furniture, beverage containers, and computer icons. For WIPO member states, protection is provided through registration with WIPO and examination by designated member states in accordance with the Hague Agreement. Official services are required to carry out identification in an explainable manner, whereas the artificial intelligence models currently used do not meet this requirement, as such models hallucinate and do not guarantee legal accuracy.

What it does

We used a proprietary mathematical model of explainable artificial intelligence (XAI) based on the embedding of ostensive features, published in scientific works since 2013. The use of ostensive features instead of word embeddings makes it possible to reliably verify and legally validate designs and trademarks, which subsequently enables the use of legal procedures to protect the specified intellectual property in court. Methods of information identification using explainable artificial intelligence (XAI), based on the comparison of arrays of ordered special vectors of ostensive features, unambiguously define designs and trademarks in accordance with their legal definition. https://github.com/Mykola2056/design-representation-and-identification/tree/main

How we built it

The number of ostensive features is strictly defined in accordance with Kant and fully determines observable reality and, according to the proprietary model, is assembled into arrays of special vectors, the number of which is strictly limited and predefined. In the case of designs and trademarks, no vectors other than visual and spatial data vectors are required — that is, only 7 core vectors, in contrast to the infinite set used in artificial intelligence models based on digitized abstract concepts and words. Python-based software was developed to identify designs from a design database, with full validation of the process, which makes it possible to support legal procedures for protecting the copyright of these intellectual property objects without using their metadata. By comparing these vectors within the specified arrays, the target objects can be reliably and unambiguously identified. https://github.com/Mykola2056/design-representation-and-identification/tree/main

Challenges we ran into

The main problem is debugging and testing the program on existing design bases.

Accomplishments that we're proud of

Developing a design patent database for a single country. Testing the performance of our XAI model.

What we learned

We studied the main difficulties of artificial intelligence models based on word embedding.

What's next for Ostensive definitions for legally correct analysis of design

We want to develop a search engine based on our artificial intelligence model for WIPO.

Built With

  • a
  • api
  • artificial
  • based
  • definition
  • explainable
  • intelligence
  • mathematical
  • model
  • of
  • on
  • ostensive
  • python
Share this project:

Updates