It’s starting to happen. Artificial intelligence (“AI”) systems have become smart enough to create new inventions. Two recent examples of AI-authored inventions were produced by a system called “Device for the Autonomous Bootstrapping of Unified Sentience” or simply “DABUS.” According to Dr. Stephen Thaler, who developed it, DABUS is a two-part “creativity machine.” To start the invention process, the first part of the creativity machine is trained in a given knowledge area. It then generates a series of new ideas by perturbing certain internal factors and relationships that exist within the knowledge base. The second part of the creativity machine compares the newly generated ideas to the original knowledge base and tries to decide which of the new ideas would be the most novel and have the most value.

Using descriptions of two DABUS inventions, an international team lead by Ryan Abbott filed the first-ever patent applications for inventions created autonomously by artificial intelligence without a human inventor. The patent applications were filed at the European Patent Office and the UK Intellectual Property Office.

The EPO and the UK patent office found that the DABUS patent applications satisfied the primary requirements of patentability: the inventions were both novel and nonobvious. That was not enough, however. Because an AI system was identified as an inventor, the big question was: can an AI system be a legitimate inventor?

Their answer to that question was “no.” Both the EPO and the UK patent office ruled that only natural persons can be inventors. This is consistent with similar conclusions that other tribunals have reached. In New Idea Farm Equip. Corp, v. Sperry Corp., the Federal Circuit held that “people conceive, not companies.” And in Naruto v. Slater, the Ninth Circuit Court of Appeals ruled that a macaque monkey, because it was not a human, lacked standing under the Copyright Act to sue a photographer who published “selfie” photographs taken by the monkey.

Professor Abbott believes that AI systems should be acknowledged as legitimate inventors, and he is trying to change things through academic advocacy. In a law review article entitled “I Think, Therefore I Invent: Creative Computers and the Future of Patent Law,” Professor Abbott argues that creative computers should be considered inventors because it would incentivize the development of smarter AI systems. He has a point. Incentives should match desired outcomes. If we want AI developers to be rewarded for their work, it seems logical to acknowledge the contribution of AI inventorship.

The question is how best to acknowledge AI inventorship while also addressing the related problem of ownership. Looking at just inventorship for a moment, many people do not yet accept the idea that any current AI system – even the advanced DABUS creativity machine – is truly able to conceive a new idea. For example, Gene Quinn, IPWatchdog CEO, argues that “[i]t is impossible for artificial intelligence to meet that basic prerequisite for what we today consider invention.” On the other hand, without the assistance of AI systems, it is certainly reasonable to conclude that some inventions would never happen. It therefore seems appropriate to consider Professor Abbott’s contention that an AI system could at least be a substitute for patent law’s fictitious “person having ordinary skill in the art.” Such a skilled person is presumed to know of all the prior art in a particular field. Patent practitioners know that this presumption is a legal fiction, but it could be surprisingly accurate in the case of a computer that has been trained to understand a given subject. 

Ownership is the big problem for AI-generated inventions. We simply cannot let AI systems become owners of anything. One solution promoted by Professor Abbott is to automatically assign the ownership of an AI-generated invention to its developer(s). Such a rule would be consistent with our long-standing property laws, and it would incentivize AI innovation by acknowledging the contributions of legitimate acts of creativity on the part of AI systems.

The growth of creativity machines and the impact of their inventions on intellectual property laws are areas that are worth watching closely. And the more that advocates like Professor Abbott push for change in this area, the more likely that we will see significant change in the near future. For now, AI systems cannot be inventors. But we can easily envision a world where they can be. It makes sense to start developing legal frameworks now that will permit such technological advances to be incentivized and protected.

Clyde Findley is Special Counsel in the Intellectual Property practice at Berenzweig Leonard. He can be reached [email protected].

1 You can read more about DABUS here: http://imagination-engines.com/iei_dabus.php.

2 Ryan Abbott is a law professor at the University of Surrey in England.

3 The EPO application numbers are EP3564144 and EP3563896.

4 New Idea Farm Equip. Corp, v. Sperry Corp., 916 F.2d 1561 n.4 (Fed. Cir. 1990).

5 Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018).

6 57 B.C.L. Rev. 1079 (2016).

7 https://www.ipwatchdog.com/2020/01/28/epo-provides-reasoning-rejecting-patent-applications-citing-ai-inventor/id=118280/