Do Machines Dream of Electric Inventions? A Technical Review of DABUS and Its Claimed Inventions.

Editor’s note: This article is the second in a series of four posts about AI and inventorship, developed as part of Professor Colleen V. Chien’s AI and the Law class at Santa Clara University School of Law. The first blog describes the Thaler v. Hirshfeld appeal pending before the Federal Circuit and related appeals. Based on a thorough review of Dr. Thaler’s patent claims, this post describes the technology of the AI system that discovered the inventions claimed in Dr. Thaler’s patent applications.

Co-author Hao-Jan Wan is an experienced patent agent with a background in chemical engineering and biotechnology. Co-author Jungyeon Kim is a technology specialist and an incoming associate at WilmerHale. Jungyeon was previously an engineer in the server industry. They are both 2022 graduates of Santa Clara Law and co-wrote this piece as part of Professor Colleen Chien’s class on AI and the Law at Santa Clara University.

While courts around the world are wrestling with the question of whether an AI system can be named as an inventor on a patent application, skepticism abounds about Dr. Thaler’s claim that his AI system is actually inventive — and about the true inventiveness of AI in general.

In an October 2020 report by the USPTO on AI and patent policy, for example, a majority of commenters from the legal profession, academia, and private sector regarded artificial general intelligence, where AI possesses human-level intelligence, as “merely a theoretical possibility that could arise in the distant future.” Instead, in their view, current AI is limited to “narrow” AI, where systems perform individual tasks in well-defined domains and therefore “could neither invent nor author without human intervention.” Popular European IP law blog IPKat has also thrown cold water on the hoopla around Dr. Thaler’s case, comparing it to the parable of the “Emperor’s New Clothes,” where, by that analogy, everyone realizes that AI is far from being truly inventive, but pretends otherwise to avoid looking like a fool.

So, where do things actually stand? This article examines the technology behind Dr. Thaler’s AI system through the lens of Dr. Thaler’s patent applications and descriptions of DABUS, allowing the reader to scrutinize whether DABUS is truly inventive.

What is DABUS?

DABUS is an artificial intelligence (AI) system created by Stephen Thaler. According to Thaler’s disclosure, DABUS adopted a novel form of neurocomputing that allows machines to generate new concepts along with their anticipated consequences, all encoded as chained associative memories. In DABUS, a plurality of technical solutions or ideas are represented by the neural units. Through this network, chaining topologies form together with various environmental input patterns, and DABUS randomly links them together to form chaining topologies representing novel concepts. For example, if DABUS had been tasked with inventing an oral hygiene product, it would have combined several concepts together (e.g., hog whiskers → embedded in → bamboo stalk) with consequence chains forming as a result (e.g., scrapes teeth → removes food → limits bacteria → avoids tooth decay).

A comparison of “Creativity Machines,” left, which are pattern based, and “DABUS,” right, which relies upon the geometries and topologies formed among multiple neural network modules. Source: Imagination Engines.

Thaler’s website Imagination Engines describes DABUS as “a swarm of many disconnected neural nets, each containing interrelated memories, perhaps of a linguistic, visual, or auditory nature.” Eventually, a fraction of these nets “interconnect into structures representing complex concepts.” In turn, these “concept chains” connect with other chains, producing “ephemeral structures or shapes” that rapidly materialize and dematerialize. Cameras watch the neural chains, conveying their states to “novelty filters” that detect and reinforce the worthwhile concept chains, while weakening the “geometries representing undesirable notions.” “In the end such ideas are converted into long term memories, eventually allowing DABUS to be interrogated for its cumulative inventions and discoveries.”

Electro-optical embodiment of DABUS. Source: Imagination Engines.

According to Imagination Engines, DABUS includes at least 11 patented inventions. Thaler is the sole inventor of all these patents. The first DABUS-related patent was issued in 1997 for U.S. Patent №5,659,666 “Device for the autonomous generation of useful information.” It disclosed an artificial neural network for simulating human creativity. By using a reciprocal feedback connection, the neural network was trained to produce a desirable output based upon predetermined criteria. The latest patent was issued in 2019, for U.S. Patent №10,423,875 “Electro-optical device and method for identifying and inducing topological states formed among interconnecting neural modules.” This invention disclosed an artificial neural system for avoiding processing bottlenecks while providing data about states of a plurality of interconnecting neural modules. It specially addressed a method of identifying the novel topology through optical devices.

The DABUS Patents

As shown on the website Artificial Inventor, which was created by the team of patent attorneys working with Dr. Thaler, DABUS has conceived two inventions so far: a “neural flame” and “fractal container.” The patent applications can be publicly viewed on the European Patent Office’s website, as applications EP3563896 and EP3564144, respectively. The first, titled “Devices and methods for attracting enhanced attention,” disclosed a light beacon with a pulse wave having a periodical light source and a random light source for attracting attention. The second, titled “Food Container,” disclosed a food container having a wall with a fractal profile so as to engage with other food containers.

According to Thaler’s description of how his machine works, DABUS would have selected two seemingly unrelated concepts, a fractal pattern (e.g., a collection of concave and convex patterns) and a food container (e.g., a water bottle), connected the two concepts, and arrived at the claimed invention of a food or beverage container having a wall that has a fractile profile. However, whether the invention as initially claimed is novel and nonobvious as patentable is left unsolved. The applications’ designation of DABUS as a sole inventor has led to decisions by patent offices and courts on whether a patent can be granted for an invention reportedly made by an AI system.

As discussed more thoroughly in the first post in this series on AI and inventorship, both the EU and American patent offices, among other countries, have rejected the applications for lack of an inventor. In the U.S., courts have found that under the relevant statutory law, an “inventor” must be a “natural person.” As machines are not natural persons, they cannot be named inventors on patent application.

Source: Drawings for Thaler’s patent applications with the EPO.

Potential Challenges

Putting aside the statutory language, how is inventorship determined under the Patent Act in practice? Typically, inventors are those who contribute the ingenuity necessary to create an invention. Conception is the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention, as it is hereafter to be applied in practice. Hybritech, Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1376 (Fed. Cir. 1986). In other words, the one who generated the inventive concept can qualify as the inventor, while the one who operates diligently under another’s instructions might not.

Nowadays, AI is widely used in the drug development process. For instance, Pfizer is using IBM’s Watson for Drug Discovery, a system that uses machine learning, to support its search for immuno-oncology drugs. In general, AI can assist in structure-based drug discovery by predicting the 3D protein structure because the design is in accordance with the chemical environment of the target protein site, thus helping to predict the effect of a compound on the target along with safety considerations before their synthesis or production. It has become a normal procedure for AI to predict the drug structure while human-beings conduct experiments to verify the AI’s prediction. In other words, there is a great possibility that a drug structure will be independently invented by an AI system. Absent further development in the law, there is a risk that some of these inventions will be without a legal inventor and consequently not be patented because AI cannot be the inventor by law and the human operator is also not qualified if he or she merely operated under the AI’s instructions.

AI is evolving quickly and only growing in importance to many industries. When AI switches from automating the individual tasks of human researchers to automating inventive activity itself, AI may well be contributing more toward technological development than the skilled persons in those industries. No matter what the outcome of Thaler v. Vidal is, the debate over AI inventorship will need to be revisited soon.




The Blog of the Technology, Business, Law, and Policy of AI Class at Santa Clara Law School

Recommended from Medium

4 Ways Artificial Intelligence is Reshaping the Construction Industry

5 Best Use-Cases Of Voicebots


The 2018 Machine Intelligence Landscape: A New Look at MI vs ML vs DL vs AI

Will Neuralink Really Hack Our Brains

Inventory of 3 technology trends that will invade our lives in 2022 (with shares)

Demystifying AI: Is your small-medium sized business, AI ready, yet?

How to Solve the Biggest Challenges in Enterprise AI

Customer Service in a World of Ambient Computing

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


More from Medium

Can AI Be an Inventor? Legal, Technical, and Practitioner Views of Thaler v. Vidal

Why Are High Efficiency Air Conditioners Better?

Look Around You! — Do You See A World About to Change to Sustainability?

SIDL — The Danger of Comfort