4 ms·

>You can't get rid of patents by arguing that they distort markets. You make a good point here that patents are intended to distort markets. However they are i

>You can't get rid of patents by arguing that they distort markets. You make a good point here that patents are intended to distort markets. However they are intended to distort markets in particular ways and the question is, do they distort markets in the ways in which they were intended. Specifically do they reward and encourage investments in innovation for the public good? Answering this question in general is very tricky, but within the cryptography space it appears that they do not. There are several reasons for this that I am aware of and probably the few that I am missing: 1. Companies and open source projects avoid using patented cryptography. This means that making money off of patented cryptography is hard and the public does not benefit from these innovations until the patent expires. 2. Researchers generally boycott patented algorithms, for instance submitting algorithms for government crypto contests like SHA-3/AES often require ceding all patent rights. Patented algorithms are less well studied and aren't standardized so don't benefit from the research or operational community. 3. Most cryptographic algorithms require at least two parties. Patents typically require both communicating parties to have a license. Thus network effects favor open standards unencumbered by patents. 4. The US government has a long history of not paying for patented cryptographic inventions due to the vagaries of US law. Thus, patents provide only minimal protection against historically one of the biggest buyers of cryptographic inventions, the US government. 5. The public good of having cryptographic algorithms be unencumbered by patents is so great that people will just figure out how to get around your patent. This is what happened with Schnorr signatures. The US government altered the Schnorr signature algorithm just enough to create an algorithm, DSA (Digital Signature Algorithm), such that the patent didn't apply. DSA isn't quite as good as Schnorr, but DSA is better than Schnorr+patent so nearly everyone used DSA until the patent expired. Patents in the context of cryptographic algorithms appear to not help much in terms of encouraging and rewarding innovation and they also appear to act counter to the public good.

mlthoughts2018 8y ago
Many of the same arguments apply to patented algorithms in image processing and computer vision as well. SIFT and SURF are probably the two most famous examples, e.g. [0] and [1]. Two quick anecdotes about this. The result of the message board thread of [0] once led to a compliance team in my former company to ban all use of OpenCV for computer vision applications (which, obviously, is insane). They took an exceedingly risk averse stance that if any of our software imported a third party OpenCV library that contained SIFT, this potentially opened us up to a lawsuit, even if we did not actually use SIFT (and might incur costs for us to prove we didn't use SIFT). Needless to say, this created an unbelievable amount of pain for being on computer vision and image processing teams there. Lots of re-implementing of basic algorithms, but without the benefit of years of community development, testing, etc. And [1] below indicates the sort of patent-avoiding contortions people will go through in research. If your algorithm is patented, typically it means people won't use it and will do whatever is needed to avoid it, even just abandoning a certain line of research all together and choosing something else without the possible patent hurdle. In the most innocuous cases, it just means the patented algorithm receives far less research attention and becomes irrelevant faster (giving the interpretation that the patent acts as a very short term royalty grab by the patent author, which seems antithetical to the stated public innovation goals of a patent system in the first place). In worse cases, the patent acts like a research roadblock that can thwart useful lines of research from others. When patents allow for uncompensated academic work to use the algorithm royalty-free, that can sometimes be an OK middle ground, but still doesn't seem to be in the spirit of the whole idea to spark and catalyze innovation in the first place. [0]: < http://opencv-users.1802565.n2.nabble.com/SURF-protected-by-patent-td3458734.html http://opencv-users.1802565.n2.nabble.com/SURF-protected-by-... > [1]: < https://dsp.stackexchange.com/questions/1288/what-are-some-free-alternatives-to-sift-surf-that-can-be-used-in-commercial-app https://dsp.stackexchange.com/questions/1288/what-are-some-f... >.
- EthanHeilman 8y ago
  I suspect that fields sufficiently similar to Mathematics are a poor fit for the patents and will inevitably adopt the norms of Mathematics research.
- AndrewKemendo 8y ago
  This is basically the story of why ORB-SLAM didn't become the de facto commercial slam implementation. They wanted iirc 30k per license, which is impossible for any startup.

mlthoughts2018 8y ago

Many of the same arguments apply to patented algorithms in image processing and computer vision as well. SIFT and SURF are probably the two most famous examples, e.g. [0] and [1]. Two quick anecdotes about this. The result of the message board thread of [0] once led to a compliance team in my former company to ban all use of OpenCV for computer vision applications (which, obviously, is insane). They took an exceedingly risk averse stance that if any of our software imported a third party OpenCV library that contained SIFT, this potentially opened us up to a lawsuit, even if we did not actually use SIFT (and might incur costs for us to prove we didn't use SIFT). Needless to say, this created an unbelievable amount of pain for being on computer vision and image processing teams there. Lots of re-implementing of basic algorithms, but without the benefit of years of community development, testing, etc. And [1] below indicates the sort of patent-avoiding contortions people will go through in research. If your algorithm is patented, typically it means people won't use it and will do whatever is needed to avoid it, even just abandoning a certain line of research all together and choosing something else without the possible patent hurdle. In the most innocuous cases, it just means the patented algorithm receives far less research attention and becomes irrelevant faster (giving the interpretation that the patent acts as a very short term royalty grab by the patent author, which seems antithetical to the stated public innovation goals of a patent system in the first place). In worse cases, the patent acts like a research roadblock that can thwart useful lines of research from others. When patents allow for uncompensated academic work to use the algorithm royalty-free, that can sometimes be an OK middle ground, but still doesn't seem to be in the spirit of the whole idea to spark and catalyze innovation in the first place. [0]: < http://opencv-users.1802565.n2.nabble.com/SURF-protected-by-patent-td3458734.html http://opencv-users.1802565.n2.nabble.com/SURF-protected-by-... > [1]: < https://dsp.stackexchange.com/questions/1288/what-are-some-free-alternatives-to-sift-surf-that-can-be-used-in-commercial-app https://dsp.stackexchange.com/questions/1288/what-are-some-f... >.

EthanHeilman 8y ago
I suspect that fields sufficiently similar to Mathematics are a poor fit for the patents and will inevitably adopt the norms of Mathematics research.
AndrewKemendo 8y ago
This is basically the story of why ORB-SLAM didn't become the de facto commercial slam implementation. They wanted iirc 30k per license, which is impossible for any startup.