In my view, Supreme Court jurisprudence on software invention eligibility has made little net progress since Diamond v. Diehr in 1981. The major advance in jurisprudence underlying that decision, determining the Section 101 question by looking at the claimed invention, “as a whole”, to determine if it is “otherwise statutory” without the mathematical algorithm (or more generally, abstract idea) in question, largely was ignored if not eviscerated by Bilski v. Kappos (2010). The diametrically opposed conclusions recently obtained by different factions of the CAFC in CLS Bank starkly demonstrate that the Bilski “abstract idea” analytical framework, for determining patent eligibility under Section 101, is fundamentally flawed and drastically overbroad. The problem is that, in its attempt to weed out business method inventions from patent eligible subject matter, the Bilski Court devised an analytical framework that, taken to its full extent, is readily capable of kicking virtually any software patent out of patent eligibility. Most troubling is that the Bilski Court gives absolutely no guidance as to where to draw the line between what should be regarded as objectionable and eligible subject matter.
So, how do we get out of this mess? Not very easily, and not very likely, but if we do, it will have to built on a solid understanding of the contours of the statutory subject matter challenge, so that we can finally make some real progress, instead of going around in circles for another 30 years. At this juncture, I think it is pointless to try to decipher and harmonize the multitude of software subject matter cases – that is a fool’s errand. The better approach, in my opinion, is to take a fresh run at the issue, and get more insight into what a good Section 101 test should achieve from a policy perspective, and then try to figure out what that test would look like. If that test then matches up with an existing line of reasoning, we can follow that reasoning. Or, perhaps we need to introduce a new or modified test. Clearly, the Supreme Court would benefit from taking a step back and studying the policy issue divorced from the convoluted reasoning of its prior decisions on the topic.
Because the inner workings of a computing system are not directly observable, it is helpful to draw an analogy to another form of programmed technology that will no doubt raise similar statutory subject matter questions. Namely, let’s assume that the invention is embodied in a programmable robot, and that the robot is an industry standard robot. Let’s further assume that robots such as these are owned by hundreds of millions of businesses and consumers. This industry standard robot moves about with battery power, understand verbal commands, speaks, detect objects in its environment using machine vision, and takes action on those objects, such as move them from one place to another, rearrange their order, or alter their state, for example by painting them different colors.
Now, let’s explore the patent eligibility for some hypothetical programmed configurations for the robot.
First, let’s assume the robot is programmed to perform a novel method to efficiently install a bumper on an automobile assembly line. The method relies on a novel algorithm that determines an optimum force to use to hold the bumper against the chassis while it is welded in place. Let’s also assume that the only novel element of this programmed robot is software loaded into the robot and used to program its operation according to the algorithm.
As a policy matter, should this be patentable subject matter? Setting aside legal jurisprudence, I would say most engineers would believe this to be patentable subject matter, as it is most certainly technology, most certainly useful, and most certainly applied to an industrial application. What could be more historically quintessential American technology than automation and a mass production assembly line?
But is this patent eligible subject matter under the current Bilski “abstract idea” test? I would submit that a court could easily find this invention ineligible subject matter using that test. First, the industry standard robot can be ignored, just the same way the court in Bilski ignored the computer in that case. That leaves only the disembodied software and algorithm that, according to Bilski, are nothing more than an ineligible abstract idea.
Next, let’s say the robot is programmed to perform a novel surgical procedure that is impossible to manually perform, again employing a novel algorithm embodied in software loaded into the robot and used to program its operation. Certainly, if this procedure was performed manually, it would be considered patentable, at least by the vast majority of patent attorneys. Yet again, however, the Bilski abstract idea test could easily be used to deem it non-statutory. Again, a court need only ignore the industry standard robot, and declare the disembodied algorithm abstract.
Most patent attorneys and policy makers would probably agree that the subject matter posited above should be eligible for patent protection. But these inventions are readily filtered out from patent eligibility using the Bilski abstract idea analysis.
Now consider another hypothetical- let’s say this time that the robot is programmed to automatically design a specialized surgical procedure for a specific patient, after reading the patient’s digital x-ray’s. But in this case, the claimed robotic configuration only requires the robot to input the x-ray data, determine the procedure, and store it in a memory, so it is ready to be used later. This also seems like a solid technical and scientific medical advance even without actually performing the surgery, but again, the Bilski abstract idea test could easily be used to deem it non-statutory.
To explore the over-breadth of the abstract idea test further, let’s now assume that the robot is programmed to move flags of different colors between different areas of an airfield, where the flags use a standard signaling scheme to signal to airplanes passing overhead. Different combinations of flags of different colors signify different wind conditions on the field. Let’s further assume that a robot is programmed with a novel algorithm to choose the most efficient route the robot travels to move the flags between a storage area and a display area, saving energy and improving the speed at which the flags can be reconfigured to signal different conditions.
Again, from a policy perspective, most would expect that a method to move signal flags around an airfield more efficiently, to improve the speed at which signals can be switched, and save energy, would be considered patent eligible subject matter.
Now, let’s assume that the algorithm in the above example is a novel artificial intelligence algorithm that mimics the human thought process. Here again, I would argue that from a policy perspective, this invention, being an engineering advance, should be eligible subject matter. But again, Bilski would almost seem to require this invention to be disqualified, as it mimics what many would consider an abstract process – human thought (although that assumption in itself will probably be proved a fallacy at some point in the future).
Now, let’s assume that the flags represent not weather conditions, but financial information, and in particular the cost of landing at the airfield on that particular day, the cost of jet fuel at the airport, and the cost of parking overnight at the airfield. Does this change our feeling about this invention from a policy perspective? The algorithm still increases efficiency in moving the flags around, and saves energy, so it still has a technical advantage that is independent of the information conveyed by the flags. But here again, the abstract test could easily be used to dismiss this invention from patent eligibility, even more so because it’s purpose is to signal financial information.
So let’s now modify the hypothetical to make it more challenging from a policy perspective. Now assume that the novelty resides not in the manner in which the robot moves the flags around, but rather in a novel algorithm to set the pricing of the airport services based on anticipated weather conditions. Also, here again, the Bilski abstract idea test can be used to disqualify this invention from patent eligibility. Yet, the robot is performing a useful physical activity (signaling price) and is definitely a machine. However, from a policy perspective, one can argue that this invention is a financial innovation, not a technical innovation or at least an innovation that provides a technically functional advantage, and therefore perhaps should be treated differently.
OK, now let’s assume that the novelty of the algorithm is an artistic choreographic performance, and the robot is programmed to perform this choreography using the flags. Let’s also assume there is no mathematics involved in the algorithm – it is strictly a series of moves for the robot to perform. Again, the Bilski abstract idea test can certainly be used to eliminate this from patent eligible subject matter. And again, like the previous hypothetical, there is a substantial qualitative difference between this, an artistic advance, and a technical advance. We can make this hypothetical even more apparently patent ineligible by assuming that the innovation is a musical score that serves only an artistic purpose, and that the robot is simply programmed to play the score on a conventional speaker.
So, what can we observe about the choreographic subject matter hypothetical that might help guide us to a better test? Well, we can’t say the algorithm doesn’t control the operation of the robot, because it does. Essentially it is integral with the physical operation of the robot. Nor can we honestly say the activity is abstract, as the robot is making physical movements that, if applied to a utilitarian purpose, would be considered patent eligible in nature. In fact, about the only meaningful observation we can make to differentiate the posited invention from one that is more clearly patent eligible, is that it is artistic as opposed to technical in nature.
Of course, were the posited invention to include an artificial intelligence algorithm to automatically create the artistic choreography, it would at least have a substantial technical nature, being a product of software engineering and scientific research, notwithstanding the algorithm being put to an artistic purpose.
Now let’s consider what the pricing hypothetical teaches us. In this case the robot signals, in a purely physical way, the pricing information generated by the pricing algorithm. The pricing algorithm also receives data that represents weather conditions, and for the purpose of the hypothetical, we could even say that the robot actually detects the weather conditions using one or more detectors that are standard features of the robot. But here, as assumed, the innovation lies in a purely financial algorithm that produces pricing information that in essence is completely unrelated to the operation of the robot, other than the robot is used a mechanism to signal the resultant prices. Unlike the choreograph example, the novel pricing algorithm does not inherently require the pricing information to be signaled in any particular way. But by definition of the hypothetical, the signaling of the pricing is definitely a required element of the invention, so according to the machine test, cannot be ignored. And, since a major component of the posited invention is that it informs pilots flying overhead, it’s hard to say that the signaling is “insignificant post solution activity”, although it may be post solution activity that is known in the prior art. Moreover, if we were to say that the novel algorithm predicted weather vs. determined pricing, we would again likely think differently of it even though virtually all aspects of the invention are identical other than the nature of the algorithm.
So what can we surmise or conclude from the above?
The central problem for the Court in CLS Bank (and the problem for the Bilski Court) is that it is confronting a hybrid invention – one that is not purely technical in nature at least according to traditional notions of patent eligibility. The simple fact is that such inventions are a fusion of traditionally patent ineligible subject matter such as financial algorithms or “business methods” and traditionally patent eligible subject matter, that is, automated systems or processes. These “hybrid” inventions span across a continuum. On one end of the continuum, the novelty lies almost entirely in the financial algorithm, and the technical portion that is arguably “otherwise statutory” owes itself almost entirely to the automation of the algorithm on a programmed computer. On the other end of the continuum, the “otherwise statutory” subject matter may be a purely technical innovation, such as an artificial intelligence algorithm in combination with a novel technical architecture that is based on a scientific advance that is put to use to make a business decision. In between there are a multitude of combinations that defy simple categorization as patentable or unpatentable.
The Bilski abstract idea test is vastly overbroad in its reach – it can too easily be applied to disqualify seemingly deserving innovations from patent eligibility. In particular, by isolating an algorithm from its physical instantiation, the Bilski test can be used to inappropriately disqualify innovative electromechanical activity, and other deserving subject matter, from patent protection. As is evident from the above examples, the technical algorithms are all equally abstract to the financial algorithm or artistic algorithm, when conceptually disembodied from their respective robot embodiments. Accordingly, the abstract idea test, at least as applied by Bilski, is clearly unable to properly differentiate between traditionally worthy, physically embodied, technological innovations, and abstract ideas that truly lack sufficient technological or physical instantiation to be properly patent eligible.
The defective reasoning used in Bilski is made even more evident considering that most of the electromechanical activity described above could, in theory, be achieved with a purely mechanical device, albeit an exceedingly complex one. If that were the case, it becomes logically impossible to separate the algorithm from its physical instantiation, and the Bilski abstract idea test becomes ineffectual as a tool to disqualify such an inventions from patentability. My point here is that if the abstract idea test were logical and technology neutral, it would apply equally well to a purely mechanical device and an electromechanical device, which it does not, at least as presently formulated.
By the same token, the machine test, as a positive litmus test for patent eligibility, is also overbroad. It can too easily be applied in the above example to find patent eligibility for an innovation that would not normally be considered eligible, namely an artistic performance by the robot where the only novelty lies in the choreography as opposed to a technical advance.
Probably the Diehr analytical framework – that essentially holds that an invention cannot be excluded from patent eligibility just because the point of novelty is an abstract idea, such as a mathematical algorithm, if the invention as a whole is “otherwise statutory” — offers the most sound approach. Diehr, in my opinion, seems to require an analytical framework that is directly opposite to the analytical framework applied in Bilski, which focuses on the abstract idea and ignores the invention as a whole by playing make believe – i.e., pretending that a computer programmed to perform financial or business method tasks is indeed not a machine at all.
The Diehr approach to determine patent eligibility under Section 101 is at least logical, and does not ignore reality. It essentially determines if the invention as a whole is otherwise statutory. Rather than requiring us to arbitrarily “disappear” the machine part of an invention to metaphysically transform it into an unpatentable abstract idea, Diehr uses a holistic approach that does not require tricked up legal reasoning.
Of course Diehr would seem to require us to find that a computer programmed to perform a financial algorithm is “otherwise statutory” because it is indeed a machine in the absence of the algorithm. I don’t believe that this is the case however, because as we can see from the choreography invention posited above, if the “invention” is artistic in nature, it should not be patentable whether performed by people or by programmed machines. As such, Diehr would say that a computer programmed to perform choreography, where the invention as a whole lies in the realm of “art” and not in the realm of “technology”, is not eligible for patenting under Section 101.
This leads me to my conclusion that the more honest, logical and clear approach to dismissing business method innovations from eligibility is to hold that a machine programmed solely to perform a non-technical business method is not “as a whole” within the useful arts.
On the other hand, because computers programmed to perform business methods essentially are machines that automate business tasks, it is not illogical to say that, “as a whole”, they are within the useful arts. This does not appear, however, to the be direction the Supreme Court is heading.