No one has watched Iron Man without dreaming of having their own personal assistant like JARVIS to run their lives. In fact, even the small robot Dum-E that helps Tony Stark clean up his mess would be a keeper to most. In the last decade, Artificial Intelligence ("AI") and Machine Learning ("ML") have made great advances in the world of technology and computer programming. The number of uses for these tools are growing exponentially from personal and medical assistants to cars that can drive by themselves. It is no wonder that developers and inventors are trying to protect the intellectual property of their creations.
When it comes to patenting, mathematical methods and computer programming form the basis of AI and ML which are excluded from patent protection. However, this may be overcome in certain circumstances. If a computer-implemented invention, were to contain a "technical effect" it may well be patentable. The obvious question to then ask is what is this technical effect? And just as important, at what point is this requirement satisfied?
In looking at examples in this field and at patent examiners reports to patent applications we can get a preliminary understanding to what examiners (the people who will ultimately allow an application to proceed to grant) feel will satisfy this technical effect*. Since South Africa takes its lead from the EU, the following EU examples provide us with a road map that may can lead us to patentability of AI and ML programs.
Our first road maker is an application for a neural computer containing an ML program that could be used to test for heart arrythmias. It was known that heart monitors at the time presented false positive results when used. This neural computer aimed to overcome this problem by training the ML program not only to look for signs of arrythmias, but also to look for and exclude ECG patterns that were irregular but not indicative of arrythmias. In testing, the system successfully reduced the number of false positives while providing accurate positive results. Under examination it was noted that this invention provided a technical solution to a current technical problem. This thereby satisfied the technical effect requirement for patentability.
Next on route is an application for an insulin pump with a user-specific interface. The pump contains an ML function that updates the user interface options, depending on the user and the time of day, thereby saving space on the very small screen. So, the question is whether there was a technical solution to a technical problem. In looking at insulin pumps, by their very nature they have very small screens, and it was a technical challenge to display all the necessary options that a user would require. The examiners noted that the interaction between the ML-integrated pump and the user interface did provide a technical solution for the user, and therefore the pump would be patentable.
Finally let us look at an application for a computer program that uses AI learning to reduce wasted space when cutting out shapes from a sheet of material. On first glance this is simply a computer program which would be excluded from patentability. But once this program is implemented into a machine controller, that controlled the cutting of the steel, say for cutting out car doors, that could be an invention having a technical effect. This invention therefore provides a technical solution, being a better way of cutting car doors, and is therefore patentable.
Based on what we have learnt, and as an experiment, let us examine whether JARVIS is in fact patentable. If the JARVIS ML program is implemented within a device and can overcome a technical problem, it should be patentable. It would not be hard to imagine that the technical problem was that a user was only able to type in instructions via keystrokes and receive a wide selection of results through the presentation visual data. In that case, I would suspect that a ML device like JARVIS that is able to receive audible commands, and trained to obtain results depending on user-specific parameters and location, and to immediately give the most accurate result, using an audible, conversational interface, would be providing a technical solution to our problem. Accordingly, it would have the technical effect that would warrant JARVIS patentable*.
As an AI or ML developer, maybe you have created a device that is providing a technical solution by overcoming some technical problem. Have you thought about protecting it using a patent? A patent could give you the monopoly you need to boost your device into the market. It could also give you substantial leverage in dealing with investors, licensees and even buyers of your tech.
Based on our road map, we see that it is important to fully disclose all innovative technical uses of the invention. More that the AI or ML computing needs to be implemented into a user-interactive device to increase its chances of patentability. A point to remember is that the combination of machine learning functionality and the provision of an intuitive and interactive interface to a user is what would provide the technical effect. From here, having a patent attorney to carefully draft your claims to ensure that any exclusions to patentability are overcome, as well as to clearly explain and distinguish your invention from those that currently exist, would surely take you over the finish line. Feel free to call us if you have any questions.
It is not surprising that a patent was filed in 2001 and granted for an Artificial Assistant. No prizes for guessing who one of the inventers is… I will give you a clue. His name starts with a T! Find it here https://patents.google.com/patent/US20120016678A1/en
*For the sake of this article, we will assume that the inventions discussed here have the novelty and inventiveness required for patents in general.
Please note that the legal topics informally discussed here are general discussions of certain aspects and therefore certainly not intended as legal advice. We look forward to discussing your particular case with you.