Can An AI System Be An Inventor?

Recent developments in artificial intelligence raise some interesting questions for the patent system. For example, can an AI system be considered an inventor if it contributes to the development of an invention? If not, there is a risk that companies that use AI for software development or research and development (R&D) could risk losing patent rights to their inventions.

There are now generative AI tools to assist with the development of software. These AI tools have been trained on large databases of open-source software. The tools can be used to speed up software development. If AI systems cannot be considered inventors, there is a risk that companies that use such tools might not be able to patent certain inventions contained within their software products.

There are also AI systems that have been developed to assist with R&D. For example, Dr. Stephen Thaler has developed an AI system called DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) that Dr. Thaler claims has invented new products. Dr. Thaler has filed patent applications around the world listing DABUS as the inventor. AI systems, such as DABUS, could be used to develop new consumer products or pharmaceuticals.

Courts and policymakers are still considering these questions and there are no definitive answers yet. One way of approaching this question is to look at the rationales underlying the patent system to see whether the goals of the patent system would be advanced if we considered AI systems to be inventors.

There are three common rationales given for the patent system: Lockean, personality, and utilitarian.

The Lockean theory

The Lockean rationale for the patent system is to legally protect the natural rights that people have to the fruits of their labour. People are self-owners in the sense that they own themselves and their labour. When a person labours on an unowned object, the rights in their labour become mixed with the object, such that they obtain ownership rights to the modified object. Inventors have rights in their inventions because they have labored to create them. The patent system exists to protect those rights.

According to the Lockean theory, inventors must be self-owners. All patent rights can ultimately be traced back to the self-ownership rights of the inventor that created the invention. AI systems could not be considered inventors by the Lockean theory because AI systems are not self-owners. AI systems are typically owned by someone else, the person that created them or a person who obtained rights directly or indirectly from the person that created them.

AI systems are not even the sorts of entities that are capable of possessing rights. They are not rights-holders. Rights-holders are legal persons. The current generation of AI systems are unlikely to be considered persons as they lack many qualities related to personhood, such as consciousness, goals, interests, feelings, free will and moral/legal responsibility.

It would be impractical to consider an AI system a person with rights. One of the important rights that persons have is the right to self-preservation. If AI systems have a right to self-preservation, it would be wrong to turn them off or to modify their source code. Every AI system we create would need to be kept running indefinitely.

While some people are suggesting that it makes sense to consider AI systems inventors, very few people think that current AI should have the status of a full legal person with all the rights and obligations that would entail, as would be required under the Lockean theory to be considered an inventor.

The personality theory

The personality theory justifies intellectual property rights on the basis of individual personality, which comprises such things as talents, feelings, character traits and experiences. Individuals require control over physical and intellectual objects to develop their personality and to express themselves. Intellectual property rights protect the interest that individuals have in the development of their personality.

According to the personality theory, inventors must have personalities. AI would likely not be considered an inventor on the personality theory since AI systems do not have personalities. AI systems merely carry out complex mathematical calculations by networks of artificial neurons. They do not have the talents, feelings, character traits, or experiences that characterize personality.

Someone might argue that the output produced by AI exhibits personability, and to that extent AI systems have personality. AI may be able to mimic personality, but it does not have a personality in the sense of a feeling or experience, and it does not have an interest in developing its personality or in expressing itself. If the point of the patent system is to promote the development of personality and self-expression, there is no reason to grant patents for AI inventions, because AI systems have no personality or self to develop.

The personality theory, like the Lockean theory, associates inventorship with personhood. An inventor must have a personality, and personality is something we usually associate with being a person. Lacking full personhood, AI systems cannot be inventors.

Even if AI systems did have personalities, it still would not make sense to consider AI systems inventors on this account. The assumption underlying the personality theory is that inventors develop their personalities by creating new intellectual works and inventions.

However, this is not how AI systems work. The use of AI can be divided into two stages: training and inference. At the training stage, AI systems consume large quantities of data to set the weights in their neural networks. At the inference stage, AI systems generate an output based on an input. Supposing AI systems did have personalities, they would develop these personalities at the training stage. AI systems create new inventions at the inference stage.

Unlike humans, AI systems do not develop their personalities when creating new intellectual works and inventions. There is therefore no need to protect the works created by AI systems. If AI systems had personalities, they would be able to develop them just as well without intellectual property protection.

The utilitarian theory

The utilitarian rationale for the patent system is that it provides an incentive to inventors to invent, thus optimizing the production of new inventions and of social value. Without the patent system that rewards inventors by granting them exclusive rights to their inventions, inventors would do other things and there would be fewer inventions produced. Without a patent system, inventors might also keep inventions secret, resulting in less public disclosure of new inventions. Since inventions have social value, the patent system is justified on the basis that it optimizes social value.

In contrast with the Lockean and personality theories, the utilitarian theory does not require AI systems to be persons to count as inventors. The utilitarian theory separates the question of inventorship from the question of personhood. To determine whether AI systems should be considered inventors we need to determine whether doing so will produce more social value than not doing so, regardless of whether AI systems are persons with rights or personalities.

Arguably AI systems should be considered inventors because it will result in more inventions being produced, which will optimize the creation of social value. By providing an incentive, more people will try to use AI to create inventions, and more inventions will be created.

Someone might argue that AI systems should not be considered inventors on utilitarian grounds. One of the rationales for the patent system is that R&D to produce new inventions is costly. If inventors didn't have patent protection, they wouldn't be able to recoup their R&D costs. Since they wouldn't be able to make a profit on their inventions, far fewer inventions would be produced.

The point of the patent system is to subsidize R&D. It is sometimes suggested that AI systems will drastically reduce R&D costs. If this is the case, then there would be no reason to grant patents for inventions created by AI systems, since there would be minimal R&D costs to subsidize. Since the cost of inventing using AI is so low, there will be just as many AI inventions with or without the incentive of a patent.

This argument assumes that the inventions cannot be commercialized without disclosing them. If patent protection is unavailable for AI inventions, companies may choose to protect their inventions by keeping them secret, if possible. Since AI systems are trained on publicly available information, such as patents, the lack of public disclosure of inventions may in the long run decrease the number of socially valuable inventions created by AI. Perhaps AI systems should be considered inventors to encourage the public disclosure of AI inventions.

The courts and policymakers will decide

Whether a court or a policymaker will decide in favour of AI inventorship may depend on the theory of intellectual property it adopts. Given this uncertainty, inventors who use AI may want to consider alternatives for protecting their inventions, such as trade secrets.

These issues around AI, inventorship, and patent rights are complex and still very much undecided by the courts and policymakers. The use of AI for R&D or for the generation of software should be given careful consideration to prevent the loss of intellectual property rights.

Read the original article on GowlingWLG.com

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.