Slashdot Mirror
New USPTO Test Could Limit Software-Based Patents
bizwriter writes "The high tech industry has been waiting for a Supreme Court decision in the Bilski case to decide fundamental questions, like when you can patent software. But there's a new test from the Board of Patent Appeals and Interferences (PDF) that just became precedential, meaning that it offers new grounds on which the US Patent and Trademark Office can deny patents on machines that use mathematical algorithms."
Probably? Reworked? All algorithms are already mathematical formulas.
The natural world may only be approximated using a mathematical framework (at least at present), but the devices that we create based on that knowledge is a direct application of the mathematical models.
No, they are a direct application of the laws of physics (and its manifestation in chemistry, biology, ...). These laws and their applications are described by mathematical models because those are a lot more efficient to work with (and to automate) than equivalent natural language descriptions, but the underlying novel insights are based on experimentation in the real world, not regarding maths (maths are just used to describe/generalise the observations from said experiments).
So say you have an algorithm, and claim that it is nothing more than math. You can implement that algorithm in a programming language, convert it to machine language for a general purpose processor, implement it as a giant logic table in an FPGA, implement it using TTL logic circuits, or make a custom ASIC of the same logic. Many algorithms, such as signal processing, can be implemented to within a desired precision with either digital or analog filters. There are also mechanical implementations of the same ideas - mechanical calculators are obvious, but even simple things like automatic valves may be more robust in for certain applications than converting to the electronic/digital realm and back again. At what point do you draw the line and say that it is no longer just a mathematical algorithm?
There are definitely grey areas and there is no way to draw a 100% clear line (even if you don't consider the issue of software patents at all; it's simply the nature of civil law), but as far as I am concerned the cases that you mention are not necessarily hard to classify.
Whether you implement an algorithm in software or in an ASIC or in an analog filter should be irrelevant. If your contribution only lies in the algorithm, then this is what should be considered for testing the patentability requirements. If you also contribute a completely novel way of building an ASIC or an analog filter, then you'd be crazy to only claim that novel hardware in combination with a particular algorithm rather than the hardware on its own (since the latter would cover any usage).
Conversely, if you figure out that you need 2 parts of Pb and 3 parts of Na to make 2 parts of Au, the fact that you can express this using a mathematical formula does not render this knowledge unpatentable (just like describing it in English does not render it unpatentable, even though text cannot be patented -- what you are patenting is not the description, but what is described). And performing this chemical reaction under control of a computer program would not render this process non-infringing simply because the basic knowledge is described in the form of a computer program and applied under computer-control.
There are other examples, such as the psycho-acoustic model used by MP3 compression. While MP3 compression is usually implemented completely in software and a psycho-acoustic model is a mathematical model, this model is based on new knowledge about the physical world and hence a patent on that would not necessarily be a math patent (of course, there are other problems that pop up in this case, such as interoperability/network effects).
Another example is anti-lock braking: it's virtually always performed under software control, but the actual invention is that by measuring the heat caused by the friction between the wheel and the brake you can determine whether you are skidding or not (and again, whether you perform this process via software, hardware or anything else is irrelevant as to whether or not it infringes, as long as the patent claims are drafted properly).
Donate free food here