The Korean Intellectual Property Office (KIPO) announced Patent Examination Guidelines for key technology areas related to the Fourth Industrial Revolution, including machine learning based artificial intelligence ("AI"), on January 18, 2021. In the Examination Guidelines for AI, KIPO outlines specific guidelines on description and novelty/inventiveness requirements for different categories of AI inventions (e.g., AI model training invention and AI application invention, as depicted below), in addition to eligibility requirements which correspond to that of computer-related inventions. In particular, KIPO's Examination Guidelines provide examples of various AI inventions with practical drafting tips on enablement (Article 42(3)(i) of Patent Act) and inventiveness requirements (Article 29(2)).
I. Enablement Requirement
Under Article 42(3)(i), the description of an invention shall be written clearly and fully so that a person with ordinary skill in the art (POSITA) to which the invention pertains can easily practice the claimed invention. For an AI invention, KIPO suggests that the description of the technical problem, solution, and specific technical configuration (e.g., training data, data preprocessing, trained model, and loss function, etc.) be included to enable a POSITA to practice the claimed invention, unless the technical configuration is well known in the art. Some drafting tips recommended for different types of AI inventions are as follows (hereinafter, for simplicity of explanation, the AI model training invention has been further categorized into two – "AI training data invention" for inventions directed to data preprocessing, and "AI modelling invention" for inventions directed to building a machine learning model):
- AI Training Data Invention:
Describe how raw data is processed to generate, change, add, or delete training data as well as include the correlation between the raw data and the training data (i.e., explain why the raw data is used and why the training data is preprocessed in a certain way).
- AI Modelling Invention:
Specify any technical configuration or process to implement or train a model (e.g., if an ensemble of neural networks are used to train a model, identify the neural networks used and processes or means by which the model is trained using the neural networks).
- AI Application Invention:
Provide details on the correlation between the input data and output data of a trained model; that is, i) specify training data, ii) describe the correlation among characteristics of the training data for solving a technical problem, iii) indicate the machine learning model to be trained using the training data or training method, and iv) describe the generation of a trained model that solves the technical problem by using the training data and method.
II. Inventiveness Requirement
Under Article 29(2), if an invention can be easily made by a POSITA, the invention is not patentable. For an AI invention, KIPO warns that a claim merely reciting a use of AI technology is unlikely to be patentable, unless a distinguishing technical configuration (e.g., training data, data preprocessing, trained model, loss function, etc.) for solving a technical problem is specified. Otherwise, the claimed invention would be deemed as merely using a known AI technology which can be easily accomplished by a POSITA. Similar warning is given to inventions that merely use AI technology in systematizing or computerizing a process that could be or had previously been manually conducted, simply modifying the design of a conventional AI technology (e.g., simple change to a training model), and merely adding or substituting known technology to the conventional AI technology.
Accordingly, as a rule of thumb, KIPO suggests specifying the differentiating technical configuration (e.g., training data, data preprocessing, trained model, loss function, etc.) and elaborating on the technical effects directly resulting from the technical configuration that go beyond the effects of conventional AI technologies – that is, avoid merely conclusory statements regarding technical effects such as fast processing speeds, efficient processing of large amounts of data, reducing errors, or providing accurate predictions. Some drafting tips for the inventiveness requirement are as follows:
- AI Training Data Invention:
Provide specific details about how raw data is processed for training data; for example, describe how features are extracted from input data, how the training data is generated (e.g., by standardization, normalization, or vectorization); and
Explain particular effects or improvements that can be derived from data preprocessing (e.g., by performing data preprocessing on video images of a closed-circuit TV regarding "motion tracking" features, an object in video images can be more accurately identified since movement of the object is taken into consideration, whereas, the prior art merely uses the video images to identify objects).
- AI Modelling Invention:
Describe specific configuration of modelling, such as configuration of training environment, model evaluation, linking multiple models, parallel or distributed processing, and optimization of hyper-parameter; and
Provide effects regarding the training speed, prediction accuracy of the trained model, etc. that are caused by the specific configuration and thus, not expected from the conventional AI technology.
- AI Application Invention:
Describe a specific use of the output data of a trained model and the effects of using the output data in a particular manner (e.g., by using output data of a trained model (for instance, a label from an automobile part destroyed in a car accident) to calculate the cost estimate for each type of repair, a user can conveniently anticipate increases in the insurance premium depending on the type of repair the user selects).
Although there are no surprises in the Examination Guidelines, practitioners can greatly benefit from the drafting tips and examination cases provided by KIPO to ensure that AI patent applications are being properly drafted.
Related Topics