AI-Inventions, Plant Patents, and the Forever-Spinning Plum
We’ve been here before. In the 1930’s, Congress decided that plants were patentable subject matter. Since then, we’ve pushed the boundaries of what can be patentable. But now that we’re faced with AI-inventions, we have choked on the idea that an invention was created using bits and not biotics.
Attribution
In July of this year, the USPTO has once again issued another subject matter eligibility guidance announcement on the requirements for AI technology and patent protection. A savvy patent practitioner may view this as a proactive measure by the USPTO in order to foster innovation using AI as a tool, but skeptics may see this as an attempt to quickly bootstrap a new technology into an ill-fitting protective model. The US Supreme Court has already ruled that an AI cannot be listed as an inventor under the Patents Act, but maybe that’s not the right way to frame the technology.
Consider, for a moment, a perfect plum tree: strategically designed to make use of 100% of the sunlight that hits its leaves, bears the largest and sweetest of plums, is resistant to all manner of diseases and pests, and it may even harvest itself in neat piles at the end of the season. If this species of plum tree had been made in a lab by hard-working botanists, they would be interested in ensuring that their work is profitable. There are methods to do so. Back in the 1930’s the Plant Patent Act was enacted and primarily spurred on by the work of botanist Luther Burbank. The Plant Patent Act was a monumental step in establishing the plant breeding industry and fostering innovation. Before this Act, there was no legal incentive to create new strains of plants as that property wasn’t protectable nor profitable beyond the number of seeds or plants one could directly sell. Despite its monumental nature, the authors of the Plant Patent Act carved out two huge exceptions that still hold true today: plant patents cannot be protected if they sexually reproduce or are tuber propagated.
Those two carve-outs are larger than they look at first blush. The restriction to sexually reproducible plants makes sense as it means that protection is only offered on that exact genetic copy of the plant. The restriction against tuber-propagated plants seems a bit random (a favorite example by the USPTO: Irish potato and the Jerusalem artichoke). This exception is made because the part the food is propagated from is the same part of the plant that is sold as food. This exclusion was inserted due to the difficulties inherent in the widespread nature of potatoes (at the time) and in distinguishing infringing and non-infringing uses (using the plant to propagate vs just consuming the plant). Notably, however, a savvy potato botanist could seek plant variety protection on their tubers for up to 25 years or simply a utility patent. The throughline however, is this: the patent office is only empowered to offer a plant patent on plants that need some human intervention to somehow exist, be this by simple discovery in a cultivated state or by grafting or other genetic tinkering.
Returning to the perfect plum tree, imagine for a moment that this tree bore the perfect plum. This hypothetical plum would have the capability to, for example, spin forever. Strapping a generator onto this forever-spinning plum would result in a literal perpetual motion machine. Setting aside the statutory restrictions on perpetual motion, this forever-spinning plum would be the catalyst for a new golden era of humanity. Such an invention is worthy of praise and protection.
So, what does this have to do with AI? The short answer is we’ve been here before. Plants act just as AI does if viewed from the right angle. Plants simply execute natural laws to grow and bear fruit, AI does the same only with data and executables. The natural laws at play are no different. The only difference is how we perceive what an invention is. No one would deny the botanist their Nobel prize nor their patent on their perfect plum tree and forever-spinning plums. But (at least currently) the coder that writes the perfect AI to design perpetual motion would be hard-pressed to get the USPTO to give them credit for the output of that AI.
We’ve already done the hard work thinking through what kinds of protection we’re willing to offer to things that do the hard work for us. Plants bear fruit yet an inventor can protect that fruit and plant. We use mice to test the effectiveness of drugs and the effects of different genes and are still able to patent those inventions that result from that research, despite the rodents doing the heavy lifting. Notably, the mice themselves are also patentable subject matter. We use mice as a machine to produce antibodies that can then be patented. Effectively, we throw a mouse into a situation where a solution is demanded (adapt to the infection or die) and profit from the effects. Similarly, if a plant produces the juiciest of fruit, we can protect it even when we didn’t design the fruit outright (as long as the plant meets the statutory requirements, of course). So, it seems arbitrary to stop this pattern when it comes to things that aren’t alive like AI. We throw the AI into a situation where a solution is demanded and once it solves the problem, just as the mice did, we somehow rationalize that this solution couldn’t possibly be protectable under our current IP laws.
We’ve limited the scope of what kinds of plants we want protected, namely ones that can’t grow out of control and ones that can’t generate new species without our say-so. These are good stances to take with AI. We’ve used them in the same way we use plants, and by the same token, we don’t want them to grow out of control or start reproducing with one another to create new scarier versions of themselves.
We’ve been here before, the forever-spinning plum is no different if it comes from a tree or an AI. We’ve already done the work figuring out what policies work for plants and we need to consider if they can be applied to AI-created inventions.