Patenting maths

I’ve finally watched the excellent video “Patent Absurdity: how software patents broke the system”, and I highly recommend it. It contains several interviews surrounding the Bilski case in the U.S. Supreme Court, with commentary from some of the biggest names in the software patent debate. I will let the video speak for itself, it really is an excellent documentary. I will just mention that the Bilski patent application is for a mathematical method of hedging risks in commodities trading, which was rejected by the USPTO.

I just wanted to highlight one specific section, which to me is the most concise and well-stated demolition of various software and process patents.¬† In one instance, one of the attorneys for Bilski was asked whether their “invention” was just “picking up the phone and calling other people”. The attorney responds that it could be reduced to that, but it is much more. He says that:

“If you look at Claim 4 (we have these things called claims which describe what the patent is for), there is a long mathematical formula in there that didn’t exist in nature or anywhere in the literature, that these very inventive folks came up with.”

Really? Let us forget for a moment the wider and important question of whether or not it is a good idea to patent mathematical formulae (unsurprisingly, I think it is a very bad idea). Is it true that Bilski’s patent application discovers a new formula? Apparently not.

Ben Klemens, the author of Math You Can’t Use (which I own but haven’t read, and has now moved up my must-read list), has different ideas. He is the one who I think thoroughly destroys several math-related patents. He explains that a the heart of writing software there must be an algorithm that explains the process that will produce a result. These algorithms can take the shape of mathematical formulae, and you just change the variables according to what you want to achieve. Klemens explains for example that there is a common formula in linear algebra called the singular value decomposition (SVD), which is an “important factorization of a rectangular real¬† or complex matrix, with many applications in signal processing and statistics.” In plain English, it is a tool that can be used to describe various processes in mathematical terms. Klemens explains that it is possible to assign all sorts of variables to the equation, and goes ahead to prove his point. He breaks down the variables in a given SVD and assigns them different values:

We have the matrix:

x1 = sexuality
x2 = cats
x3 = affection

And then the vectors:

J1 = Jane’s responses
J2 = Joe’s responses

Then you substitute the values and enter them into the SVD equation, and you have just derived U.S. patent 6,735,568 which protects a “Method and system for identifying people who are likely to have a successful relationship”, eHarmony’s infamous patent.

Klemens has in one step completely demolished a patent that should have never been awarded, and the worst part of it is that he claims this happens all the time, that there are clever patent attorneys and patent trolls out there who are pretty much conning a mathematically-illiterate system by taking well-understood mathematical equations, assigning values to the variables, and filing for a patent.

For anyone who supports business method patents, this is a point that requires reply. There are thousands of patents out there that are nothing more than a clever mathematical con. Let us hope that Bilski delivers the right result.