What is artificial intelligence (AI)?
The term artificial intelligence (AI) describes computer-implemented approaches to emulate human decision-making structures to enable computers and machines to process and solve problems largely independently. An essential tool for being able to arrive at independent solutions is the ability of an AI system to learn. This ability is referred to as machine learning. In this process, the AI system learns because of examples to be able to generalize given patterns after the learning phase is complete. To achieve this, algorithms of the AI system build a statistical model during machine learning, which is based on training data and can be checked against test data, for example.
Why is artificial intelligence (AI) important?
Research in the field of artificial intelligence has been going on for many decades. The practical application of AI systems has received a decisive boost recently with the widespread availability of powerful computers that enable computationally complex learning and training procedures to be carried out effectively. At the same time, the amount of data collected in almost all areas of technology and society has grown enormously recently, significantly increasing the need for automated methods to analyze these vast amounts of data. These developments have led to artificial intelligence now being regarded as one of the key technologies of digitization.
Where is artificial intelligence (AI) already being used?
AI is already being used in almost all areas of science and industry these days. For example, AI finds application in data management from the production of data to execution feedback, as well as in user interfaces, for example voice and face recognition. Furthermore, AI is increasingly used in the context of the fourth industrial revolution, such as smart consumer goods, smart health, smart factories, smart agriculture, or self-driving vehicles. In industrial production, factories can work more efficiently and consistently produce high-quality products thanks to AI-supported automated predictive maintenance, data analysis, process design and error detection.
ChatGPT – A prototype artificial intelligence (AI)-based chatbot
A lot of attention has been paid to ChatGPT, for example, a recent prototype of an AI-based chatbot developed by the US company OpenAI. Such a chatbot is an application that uses AI to converse with humans in natural language. Users can ask the system questions, to which the system responds in natural language.
Generally, the more complex a system is, the more effective an application of an AI will be compared to analytical approaches.
Another advantage of an AI is that representative training data may be sufficient for an application of the AI to a complex system, and a complete, precise analytical description of the system in question is not required.
Is artificial intelligence (AI) patentable?
Yes, inventions in the field of artificial intelligence and machine learning are amenable to patent protection. However, some basic conditions must be met for a patent to be granted in the field of artificial intelligence.
The reason is that artificial intelligence and machine learning are based on computational models and algorithms, such as for classification, clustering, regression, or dimensionality reduction, which are inherently abstract mathematical in nature. Artificial Intelligence algorithms are therefore generally considered in Europe to be "as such" abstract in nature and therefore not technical.
Inventions involving mathematical methods that are realized in software are "computer-implemented inventions." A computer-implemented invention is an invention that involves computers, computer networks, or other programmable devices, and in which at least one feature is realized by a program.
For a computer-implemented invention, and in particular an invention in the field of artificial intelligence and machine learning, to be patentable, it must have a technical character. That means, it must make a technical contribution to solving a concrete technical problem by technical means.
How is artificial intelligence (AI) patentable?
A technical contribution justifying patentability may lie, for example, in a technically advantageous form of implementation of the AI system, in a technically advantageous approach to training/learning the AI system, or in a use of the AI system as a tool for achieving a technical purpose.
For example, a technically advantageous implementation of the AI system or a particular AI design may be motivated by the technical design of the computer system on which the AI is implemented. An example would be an AI design that specifically considers the architecture of a particular CPU so that processing is accelerated.
In the case of using the AI system as a tool to generate a technical purpose, the purpose must be a "concrete" technical purpose. A general-purpose definition is not sufficient. For example, the purpose of a "data communication" is too broad and general. It needs to be more specifically spelled out, such as "using AI to determine keep-alive pulses transmitted between a server and a client."
Patents for AI need a description of the training data and the training itself, if possible
The AI functions should be linked to the specific technical purpose. Such a linkage of the AI functions to the technical purpose can be implemented via a definition of the inputs and outputs of the AI system. This linkage is implemented via the data used to train the AI system and the training itself. Therefore, if possible, this training data and the training itself should also be clearly defined. This will also ensure that the invention is disclosed in a manner sufficiently clear and complete , a basic requirement for a patent to be granted.
Patents for artificial intelligence and other patentable subjects
If an AI system is used as a tool to achieve a technical purpose, further patentable subject may result. For example, if the resulting subject supports the achievement of a corresponding technical purpose, it may also each be independently claimed and protected:
- a generation of a training data set for training the corresponding AI system,
- the corresponding training data set itself
- a training of the corresponding AI system,
- the correspondingly trained AI system or the statistical model generated by the training.
The importance of a technical contribution or effect for the patentability of inventions in the field of artificial intelligence (AI) and machine learning was emphasized by the Board of Appeal of the European Patent Office in decision T 0755/18: If neither the output of a machine learning computer program nor the accuracy of the output contributes to a technical effect, an automatically achieved improvement of the machine by supervised learning to produce an improved output is not in itself to be considered a technical effect.
However, if there is, for example, a use of the AI system as a tool to achieve a technical purpose, such as a use of a neural network to identify irregular heartbeats (T 0598/07), a technical contribution and thus the basic patentability is to be affirmed.
Regarding technically advantageous approaches to training an AI system, the Board of Appeal of the European Patent Office has ruled, for example, that a method for improving an image classification in which a semantic classifier is trained with a set of example color images representing "recomposed" versions of an example image to increase the diversity of the training exemplars is inventive (T 1286/09).
What are some examples of technical purposes of artificial intelligence (AI)?
For example, methods for image processing, such as for medical or technical analysis, are considered to be a rather technical purpose when assessing the patentability of AI. Likewise, image processing for scene understanding, for example, is considered to have a rather technical purpose. Regarding speech analytic applications, methods for automatic speech recognition or virtual assistants, for example, are classified as being of a rather technical nature. In automated control, applications in the field of self-driving vehicles, robotics or machine control, for example, are assumed to be rather technical.
Procedures for natural language processing, such as chatbots or automatic translations, are considered to be rather non-technical in themselves. The same applies in administration, such as scheduling, general planning, logistics, or in FinTechs, which offer application systems for creditworthiness, asset allocation, and stock forecasting.
However, even in the case of such intrinsically rather non-technical applications, such as chatbots, the underlying AI system, such as the technical design of the corresponding computer system, may be technical in nature and give rise to artificial intelligence (AI) patentability.
Can an artificial intelligence (AI) be an inventor?
According to the previous legal interpretation of the European Patent Office, only natural persons can be inventors. The Legal Board of Appeal of the European Patent Office has confirmed this (J 0008/20 and J 0009/20). A number of legal consequences are linked to the designation of an inventor, for the exercise of which the inventor must have a legal personality, which AI systems or machines do not have.
Other patent offices, such as the United Kingdom Intellectual Property Office (UKIPO), the U.S. Patent and Trademark Office (USPTO), the German Patent and Trademark Office (DPMA), the Australian Patent Office (IP Australia), the Intellectual Property Office of New Zealand (IPONZ), or the Korean Intellectual Property Office (KIPO), have also previously ruled that a natural person must be named as the inventor.
What should be considered when applying for artificial intelligence (AI)-based inventions?
AI-based inventions belong to the field of computer-implemented inventions. Compared to other inventions, patenting them has the additional challenge that the technical character of the invention, i.e., its technical contribution to solving a concrete technical problem by technical means, must be clearly worked out to become patentable.
Talk to us and contact us for further information at:
Phone: +49 (0)611 / 341568-0
Fax: +49 (0)611 / 341568-11
E-mail: Send mail
Author: Andreas Bechinger