The curious case of xenobots part II – AI-assisted inventions

This article was originally published by The Lawyer's Daily, part of LexisNexis Canada Inc.

Laboratory experimentation often invokes images of scientists in white lab coats and bubbling flasks. However, the rapid advancement of computational techniques and computing power have significantly changed how research and development occurs in the 21^st century. Artificial intelligence ("AI") now allows researchers to conduct experiments by computer modeling or simulation (i.e. in silico), with those computer-implemented experiments being integral to the conception or reduction to practice of a related "AI-assisted invention". Xenobots are one such AI-assisted invention. Xenobots are living machines made from frog cells (Xenopus laevis) and developed with the assistance of an evolutionary algorithm. When individual frog skin and heart muscle cells are removed from their native embryonic microenvironments and reassembled in a specific fashion, they form xenobots with a functional morphology that exhibits distinct behaviours from the genomically specified default. Xenobots have shown to exhibit coordinated locomotion via cilia present on their surface. They can navigate aqueous environments in diverse ways (including swimming and walking), heal after being damaged, show emergent group behaviors, and reproduce (by kinematical replication). See our earlier publication: The Curious Case of Xenobots Part I – Patenting of Living Machines.

Instead of conducting laboratory-based experiments in the physical world, xenobots were designed in silico and were then built out of living tissues. Scientists at the University of Vermont, Tufts University, and the Wyss Institute for Biologically Inspired Engineering at Harvard University have developed a two-stage process that designs biological machines from the ground up. First, computers automatically design new living machines in a computer simulation. Then, the best designs are built by combining together different biological tissues. In particular, the scientists developed:

a linear pipeline that takes as input a description of the biological building blocks to be used and the desired behavior the manufactured system should exhibit [e.g. locomotion, object manipulation, object transport, collective behavior, and self replication]. The pipeline continuously outputs performant living systems that embody that behavior in different ways… Less-performant designs are deleted and overwritten by randomly modified copies of more-performant designs. Repeating this process yields populations of performant and diverse designs… The designs that successfully pass through the build filter are then built out of living tissues…. Pluripotent stem cells are first harvested from blastula stage Xenopus laevis embryos, dissociated, and pooled to achieve the desired number of cells. Following an incubation period, the aggregated tissue is then manually shaped by subtraction using a combination of microsurgery forceps and a 13-μm wire tip cautery electrode, producing a biological approximation of the simulated design.^[1]

Xenobots are living machines and a 3D embodiment of an AI-assisted invention. A recent design, which has been featured in the mainstream media for being the world's first self-reproducing living machines, is one that resembles Pac-Man. The Pac-Man shaped xenobots have been reported to assemble "child" xenobots by combining loose cells into clusters that look and move like the "parent" xenobots. Compared to the other configurations, the Pac-Man shaped xenobots are better at pile making, thereby facilitating self-replication.

Some challenging patent questions posed by AI-assisted inventions

Xenobots offer a not-so-unique example where AI has assisted with invention development and has enabled researchers to refine and test hypotheses more easily. AI is a technology of general applicability in a diverse array of fields. In industries ranging from instant messaging to the pharmaceutical industry, the use of AI has become a part of everyday life. Despite the immense value that AI now adds to our lives, AI also raises challenging questions to the current patent system.

One interesting question is about the effort exerted by human inventors as compared to the "effort" exerted by AI. In a recent panel discussion, Sam Kriegam, the scientist who created the evolutionary algorithm that designed the Pac-Man xenobots, commented that "the algorithm itself was rather simple and all it did was to predict how different xenobots designs could prolong the behavior of self-replication. The algorithm tried billions of different xenobots shapes in the simulation."^[2] Have humans contributed more than AI to the creation of xenobots?

Should we recognize AI as an inventor? Inventorship and patent ownership issues have garnered worldwide attention when patent applications naming DABUS, an AI machine, as the sole inventor were filed with the World Intellectual Property Organization (WIPO) and several national patent offices. Since the filings, the DABUS patent applications have experienced some bumps and some wins:

• Some bumps in Europe, the United Kingdom and the United States. The European Patent Office ("EPO"), the U.S. Patent and Trademark Office ("USPTO") and the United Kingdom Intellectual Property Office rejected the DABUS applications and contended that only humans can be listed as inventors. (EP 18275163 EP 18275174), holding that only humans can be listed as inventors. This position has been affirmed by EPO's Board of Appeals, the District Court for the Eastern District of Virginia, and the England and Wales High Court (Patents Court).
• Some wins in Australia and South Africa. The DABUS patents (with DABUS listed as an inventor) have been issued in South Africa and Australia (decided by the Federal Court of Australia).

Unrelated to inventorship, xenobots challenge the old "undue experimentation" concept. In modern research where computational capabilities are pervasive and experiments can be conducted entirely in silico, "undue experimentation" hurdles faced in the past may be resolved by only a click. This may make it easier to satisfy the enablement requirement of patent law, in which a patent applicant must describe for the world how to make their invention.

On a related note, an invention that is very simple to enable in view of AI may find itself under attack for perhaps not being sufficiently inventive from what came before. After all, if creating an AI-assisted invention is as easy as telling an AI to "brute force" countless combinations of a "rather simple" algorithm, should society grant patent protection for that AI-assisted invention over and above the existing protection already extended to the AI system used to assist in inventing?

Kriegam commented that "xenobots tend to raise more questions than they answer."^[3] This is true. In addition to the many scientific questions xenobots have raised, xenobots pose many challenging questions to the patent system. AI-assisted inventions may be an important class of invention going forward. It will be interesting to see how the patent system will evolve under AI's influence.

Should you have any specific questions about this article or would like to discuss it further, you can contact the authors or a member of our Patent Group.