/>iOn December 31, 2024, China’s National Intellectual Property Administration (CNIPA) issued the Guidelines for Patent Applications for AI-Related Inventions (Trial Implementation) (人工智能相关发明专利申请指引(试行)). The Guidelines follow up on CNIPA’s draft for comments issued December 6, 2024 in which only a week for comments were provided. The short comment period implied CNIPA did not actually want comments and is in contravention of the not-yet-effective Regulations on the Procedures for Formulating Regulations of the CNIPA (国家知识产权局规章制定程序规定(局令第83号)) requiring a 30-day minimum comment period. Highlights follow including several examples regarding subject matter eligibility.
There are four types of AI-related patent applications:
Patent applications related to AI algorithms or models themselves
Artificial intelligence algorithms or models, that is, advanced statistical and mathematical model forms, include machine learning, deep learning, neural networks, fuzzy logic, genetic algorithms, etc. These algorithms or models constitute the core content of artificial intelligence. They can simulate intelligent decision-making and learning capabilities, enabling computing devices to handle complex problems and perform tasks that usually require human intelligence.
Accordingly, this type of patent application usually involves the artificial intelligence algorithm or model itself and its improvement or optimization, for example, model structure, model compression, model training, etc.
Patent applications related to functions or field applications based on artificial intelligence algorithms or models
Patent applications related to the functional or field application of artificial intelligence algorithms or models refer to the integration of artificial intelligence algorithms or models into inventions as an intrinsic part of the proposed solution for products, methods or their improvements. For example: a new type of electron microscope based on artificial intelligence image sharpening technology. This type of patent application usually involves the use of artificial intelligence algorithms or models to achieve specific functions or apply them to specific fields.
Functions based on artificial intelligence algorithms or models refer to functions implemented using one or more artificial intelligence algorithms or models. They usually include: natural language processing, which enables computers to understand and generate human language; computer vision, which enables computers to “see” and understand images or videos; speech processing, including speech recognition, speech synthesis, etc.; knowledge representation and reasoning, which represents information and enables computers to solve problems, including knowledge graphs, graph computing, etc.; data mining, which calculates and analyzes massive amounts of data to identify information or laws such as potential patterns, trends or relationships. Artificial intelligence algorithms or models can be applied to specific fields based on their functions.
Field applications based on artificial intelligence algorithms or models refer to the application of artificial intelligence to various scenarios, such as transportation, telecommunications, life and medical sciences, security, commerce, education, entertainment, finance, etc., to promote technological innovation and improve the level of intelligence in all walks of life.
Patent applications involving inventions made with the assistance of artificial intelligence
Inventions assisted by artificial intelligence are inventions that are made using artificial intelligence technology as an auxiliary tool in the invention process. In this case, artificial intelligence plays a role similar to that of an information processor or a drawing tool. For example, artificial intelligence is used to identify specific protein binding sites, and finally obtains a new drug compound.
Patent applications involving AI-generated inventions
AI-generated inventions refer to inventions and creations generated autonomously by AI without substantial human contribution, for example, a food container autonomously designed by AI technology.
AI cannot be an inventor:
1. The inventor must be a natural person
Section 4.1.2 of Chapter 1 of Part 1 of the Guidelines clearly states that “the inventor must be an individual, and the application form shall not contain an entity or collective, nor the name of artificial intelligence.”
The inventor named in the patent document must be a natural person. Artificial intelligence systems and other non-natural persons cannot be inventors. When there are multiple inventors, each inventor must be a natural person. The property rights to obtain income and the personal rights to sign enjoyed by the inventor are civil rights. Only civil subjects that meet the provisions of the civil law can be the rights holders of the inventor’s related civil rights. Artificial intelligence systems cannot currently enjoy civil rights as civil subjects, and therefore cannot be inventors.
2. The inventor should make a creative contribution to the essential features of the invention
For patent applications involving artificial intelligence algorithms or models, functions or field applications based on artificial intelligence algorithms or models, the inventor refers to the person who has made creative contributions to the essential features of the invention.
For inventions assisted by AI, a natural person who has made a creative contribution to the substantive features of the invention can be named as the inventor of the patent application. For inventions generated by AI, it is not possible to grant AI inventor status under the current legal context in my country.
Examples of subject matter eligibility:
The solution of the claim should reflect the use of technical means that follow the laws of nature to solve technical problems and achieve technical effects
The “technical solution” stipulated in Article 2, Paragraph 2 of the Patent Law refers to a collection of technical means that utilize natural laws to solve the technical problems to be solved. When a claim records that a technical means that utilizes natural laws is used to solve the technical problems to be solved, and a technical effect that conforms to natural laws is obtained thereby, the solution defined in the claim belongs to the technical solution. On the contrary, a solution that does not use technical means that utilize natural laws to solve technical problems to obtain technical effects that conform to natural laws does not belong to the technical solution.
As an example and not a limitation, the following content describes several common situations where related solutions belong to technical solutions.
Scenario 1: AI algorithms or models process data with specific technical meaning in the technical field
If the drafting of a claim can reflect that the object processed by the artificial intelligence algorithm or model is data with a definite technical meaning in the technical field, so that based on the understanding of those skilled in the art, they can know that the execution of the algorithm or model directly reflects the process of solving a certain technical problem by using natural laws, and obtains a technical effect, then the solution defined in the claim belongs to the technical solution. For example, a method for identifying and classifying images using a neural network model. Image data belongs to data with a definite technical meaning in the technical field. If those skilled in the art can know that the various steps of processing image features in the solution are closely related to the technical problem of identifying and classifying objects to be solved, and obtain corresponding technical effects, then the solution belongs to the technical solution.
Scenario 2: There is a specific technical connection between the AI algorithm or model and the internal structure of the computer system
If the drafting of a claim can reflect the specific technical connection between the artificial intelligence algorithm or model and the internal structure of the computer system, thereby solving the technical problem of how to improve the hardware computing efficiency or execution effect, including reducing the amount of data storage, reducing the amount of data transmission, increasing the hardware processing speed, etc., and can obtain the technical effect of improving the internal performance of the computer system in accordance with the laws of nature, then the solution defined in the claim belongs to the technical solution.
This specific technical association reflects the mutual adaptation and coordination between algorithmic features and features related to the internal structure of a computer system at the technical implementation level, such as adjusting the architecture or related parameters of a computer system to support the operation of a specific algorithm or model, making adaptive improvements to the algorithm or model based on a specific internal structure or parameters of a computer system, or a combination of the two.
For example, a neural network model compression method for a memristor accelerator includes: step 1, adjusting the pruning granularity according to the actual array size of the memristor during network pruning through an array-aware regularized incremental pruning algorithm to obtain a regularized sparse model adapted to the memristor array; step 2, reducing the ADC accuracy requirements and the number of low-resistance devices in the memristor array through a power-of-two quantization algorithm to reduce overall system power consumption.
In this example, in order to solve the problem of excessive hardware resource consumption and high power consumption of ADC units and computing arrays when the original model is mapped to the memristor accelerator, the solution uses pruning algorithms and quantization algorithms to adjust the pruning granularity according to the actual array size of the memristor, reducing the number of low-resistance devices in the memristor array. The above means are algorithm improvements made to improve the performance of the memristor accelerator. They are constrained by hardware condition parameters, reflecting the specific technical relationship between the algorithm characteristics and the internal structure of the computer system. They use technical means that conform to the laws of nature to solve the technical problems of excessive hardware consumption and high power consumption of the memristor accelerator, and obtain the technical effect of improving the internal performance of the computer system that conforms to the laws of nature. Therefore, this solution belongs to the technical solution.
Specific technical associations do not mean that changes must be made to the hardware structure of the computer system. For solutions to improve artificial intelligence algorithms, even if the hardware structure of the computer system itself has not changed, the solution can achieve the technical effect of improving the internal performance of the computer system as a whole by optimizing the system resource configuration. In such cases, it can be considered that there is a specific technical association between the characteristics of the artificial intelligence algorithm and the internal structure of the computer system, which can improve the execution effect of the hardware.
For example, a training method for a deep neural network model includes: when the size of training data changes, for the changed training data, respectively calculating the training time of the changed training data in preset candidate training schemes; selecting a training scheme with the shortest training time from the preset candidate training schemes as the optimal training scheme for the changed training data, the candidate training schemes including a single-processor training scheme and a multi-processor training scheme based on data parallelism; and performing model training on the changed training data in the optimal training scheme.
In order to solve the problem of slow training speed of deep neural network models, this solution selects a single-processor training solution or a multi-processor training solution with different processing efficiency for training data of different sizes. This model training method has a specific technical connection with the internal structure of the computer system, which improves the execution effect of the hardware during the training process, thereby obtaining the technical effect of improving the internal performance of the computer system in accordance with the laws of nature, thus constituting a technical solution.
However, if a claim merely utilizes a computer system as a carrier for implementing the operation of an artificial intelligence algorithm or model, and does not reflect the specific technical relationship between the algorithm features and the internal structure of the computer system, it does not fall within the scope of Scenario 2.
For example, a computer system for training a neural network includes a memory and a processor, wherein the memory stores instructions and the processor reads the instructions to train the neural network by optimizing a loss function.
In this solution, the memory and processor in the computer system are merely conventional carriers for algorithm storage and execution. There is no specific technical association between the algorithm features involved in training the neural network using the optimized loss function and the memory and processor contained in the computer system. This solution solves the problem of optimizing neural network training, which is not a technical problem. The effect obtained is only to improve the efficiency of model training, which is not a technical effect of improving the internal performance of the computer system. Therefore, it does not constitute a technical solution.
Scenario 3: Using artificial intelligence algorithms to mine the inherent correlations in big data in specific application fields that conform to the laws of nature
When artificial intelligence algorithms or models are applied in various fields, data analysis, evaluation, prediction or recommendation can be performed. For such applications, if the claims reflect that the big data in a specific application field is processed, and artificial intelligence algorithms such as neural networks are used to mine the inherent correlation between data that conforms to the laws of nature, and the technical problem of how to improve the reliability or accuracy of big data analysis in a specific application field is solved, and the corresponding technical effects are obtained, then the solution of the claim constitutes a technical solution.
The means of using artificial intelligence algorithms or models to conduct data mining and train artificial intelligence models that can obtain output results based on input data cannot directly constitute technical means. Only when the inherent correlation between the data mined based on artificial intelligence algorithms or models conforms to the laws of nature, the relevant means as a whole can constitute technical means that utilize the laws of nature. Therefore, it is necessary to clarify in the scheme recorded in the claims which indicators, parameters, etc. are used to reflect the characteristics of the analyzed object in order to obtain the analysis results, and whether the inherent correlation between these indicators, parameters, etc. (model input) mined by artificial intelligence algorithms or models and the result data (model output) conforms to the laws of nature.
For example, a food safety risk prediction method obtains and analyzes historical food safety risk events to obtain header entity data and tail entity data representing food raw materials, edible items, and food sampling poisonous substances, and their corresponding timestamp data; based on each header entity data and its corresponding tail entity data, and its corresponding entity relationship carrying timestamp data representing the content level, risk or intervention of each type of hazard, corresponding four-tuple data is constructed to obtain a corresponding knowledge graph; the knowledge graph is used to train a preset neural network to obtain a food safety knowledge graph model; and the food safety risk at the prediction time is predicted based on the food safety knowledge graph model.
The background technology of the program description records that the existing technology uses static knowledge graphs to predict food safety risks, which cannot reflect the fact that food data in actual situations changes over time and ignores the influence between data. Those skilled in the art know that food raw materials, edible items or food sampling poisons will gradually change over time. For example, the longer the food is stored, the more microorganisms there are in the food, and the content of food sampling poisons will increase accordingly. When the food contains a variety of raw materials that can react chemically, the chemical reaction may also cause food safety risks at some point in the future over time. This program predicts food safety risks based on the inherent characteristics of food changing over time, so that timestamps are added when constructing the knowledge graph, and a preset neural network is trained based on entity data related to food safety risks at each moment to predict food safety risks at the time to be predicted. It uses technical means that follow the laws of nature to solve the technical problem of inaccurate prediction of food safety risks at future time points, and can obtain corresponding technical effects, thus constituting a technical solution.
If the intrinsic correlation between the indicator parameters mined by artificial intelligence algorithms or models and the prediction results is only subject to economic laws or social laws, it is a case of not following the laws of nature. For example, a method of estimating the regional economic prosperity index using a neural network uses a neural network to mine the intrinsic correlation between economic data and electricity consumption data and the economic prosperity index, and predicts the regional economic prosperity index based on the intrinsic correlation. Since the intrinsic correlation between economic data and electricity consumption data and the economic prosperity index is subject to economic laws and not natural laws, this solution does not use technical means and does not constitute a technical solution.
The full text is available here (Chinese only).
