Artificial Intelligence as Inventor

DABUS (Device for the Autonomous Bootstrapping of Unified Sentience)

An AI machine program called DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) was created as a series of neural networks trained with general information. One of the support team, Dr. Stephen Thaler, claims DABUS was not made to solve a specific problem in a particular technology area, and was never trained with specific technical data relevant to the invention. However, DABUS, without human assistance, created a novel solution to a problem that has been the subject of patent applications in Europe, the UK, and the US, and in South Africa.

Human Inventor Requirement

Most patent offices have a requirement that inventors be individual human persons. The US, the UK and Europe all rejected Thaler’s patent application on behalf of DABUS for that reason. South Africa has granted the patent application, but their procedures don’t require a substantive search and examination of patent applications. If the paperwork is in order and the requirements are met, the patent will be granted.

There are various traditional reasons for requiring individual humans as inventors, going back to Benjamin Franklin and the idea that the act of invention was a mental activity that could only originate from humans. The language of the Patent Act seems to require a natural person due to terms such as “whoever”, “himself/herself”, and “individual” in referencing the inventor. There are also actions described that can only be performed by a human being, such as directing an inventor to execute an oath or declaration.

There is a further consideration based on the legal tests for ‘conception’. According to the USPTO, these tests consider conception in the human mind, meaning only natural persons can be considered inventors.

The Challenge of AI

These traditional arguments, however, are running up against the evolving nature of general artificial intelligence. Narrow artificial intelligence environments have been used to identify compounds for use in some drugs. Human input, however, set up the parameters and human input was required to test the compounds. Such discoveries really don’t meet the kind of inventive conception patent offices are referring to. But in the case of general AI, the environment itself was created by humans, but not to solve specific problems. In this case, the AI generated an invention- came up with a solution to an existing problem- without specific prompting. As Dr, Thaler argues, while he can claim inventor status for the AI, he had no conceptual inventive step in the particular invention, therefore it wouldn’t be just to list him as the inventor.

Counter arguments for these might be a less textual, broader interpretation of the language, the fact that AI was inconceivable when the original tests were outlined, and that inventor identity requirements have more to do with dispute resolution than determining the source of innovation.

Should We Expand Inventorship?

Why should the system be opened up to include AI environments as inventors?

One argument, by Ryan Abbot, is that patent protection for AI generated works will incentivize innovation. No one is suggesting that patent protection would motivate an AI. Expanding inventorship would motivate those who develop inventive AI to continue doing so, which, in the end, would benefit society.

The patent system is designed to protect inventors, while disclosing information that would be of value to society. Failing to allow AI generated inventions would discourage businesses from using it, even if it proves more effective than people in solving certain problems.

There is also an interesting twist to allowing AI to be listed as the inventor. It may protect the rights of human inventors. Listing the owner of the AI as the inventor for an AI generated invention isn’t unfair to the AI. It can’t have any concern about such a matter. But it does allow people to take credit for work they didn’t do. This may ultimately devalue human inventorship, by putting the work of someone who asks an AI to solve a problem on the same level as someone legitimately inventing something new.

AI systems would still not own patents since they lack the legal and moral rights to own property. But listing an AI as the inventor isn’t about providing rights to machines. It’s about protecting the rights of human inventors and the integrity of the patent system.