Ridiculous Patent: Patenting Math

Today we point you to US Patent #6,434,582, "Cosine algorithm for relatively small angles." You read that right. The USPTO granted a patent on generating cosines for small angles.

If you read the claim, it appears that it applies only to a physically embodied circuit inside of a arithmetic logic unit (ALU). If you look at the diagrams, it certainly looks like it covers schematics. But this is almost certainly misleading. The talk of AND gates and OR gates and wired inputs and outputs describes precisely the logic required to perform the computation. And, if you read claims 11-14, it is clear that the patent is meant to apply more broadly than to just the physical embodiment of the circuit, that it is intended to extend to the algorithm itself. Any theoretical computer scientist or electrical engineer did homework excercises in college whose answers looked a whole lot like this.

Why is this dangerous? The patent covers pure logic. Pure logic is mathematics. Math is considered to be a fact of nature. Under 35 U.S.C. §101 patents can't be granted that cover the "laws of nature, natural phenomena, and abstract ideas." But here you have it: a government granted monopoly on a natural phenomena.

This is but one example. All computer software reduces to math (this too is a fact of nature, resulting from theoretical work of men like Alan Turing and Alan Church). All algorithms are math. All computer software, therefore, should be excluded from being patentable. Yet it is not. In fact, the USPTO has granted over 170,000 patents on software.

Why is software so fundamentally different from other types of inventions? Software is easily authored by anyone who wants to spend the time to do so, much like books are easily authored by anyone who wants to put pen to paper. It takes a bit of training and practice of course, but the end result is the same: an abstract idea implemented inside of an abstract device (in the case of a novel, inside the pages of a book; in the case of a program, inside the processor of a computer).

As Ben Klemens at the Brookings Institute wrote,

When you use the Record Macro feature of your word processor, you are writing a computer implemented technology which may infringe a patent—and if you put that document online, you are distributing your infringing technology worldwide. There are over 170,000 software patents filed with the U.S. Patent and Trademark Office, many very broad and all written in unintelligible legalese. It is clearly absurd to propose that every person in the U.S.A. must do a full patent search every time they record a macro, but that is theoretically what one would need to do to avoid liability.

The result is a law which is only partly taken seriously. Ronald Mann, a scholar at the University of Texas, interviewed venture capitalists and programmers, and found a resignation toward software patents: programmers do not do a patent search on every line of code, but instead simply assumed "[...] that there would be something in IBM's [patent] portfolio that their product infringed." Testimony after testimony to the Federal Trade Commission by businessmen and programmers said the same thing: to stay within the law requires such an absurd, paralyzing amount of work that nobody bothers. Conversely, one would be hard pressed to find a pharmaceutical company which does not bother with regular patent searches.

To give a more ironic example, the Recording Industry Association of America is famous for its crackdown on infringement of the intellectual property of the artists it represents, but a company named Altnet claims that the RIAA's technological attempts to stop peer-to-peer networks are infringing its patents. Even the most vehement defenders of intellectual property are unable to ensure compliance with the vast array of software patent claims.

Simply put, it is effectively impossible for you or the people in your IT department to use a computer without infringing on a patent.