[VanL]

(Disclosure: IAAPL, but this is my view and isn't reflective of any former clients or my current employer.)

The key problem with the patent system is the problem of disclosure. We always say that the patent system is a bargain between the inventor and the public. The public gets new knowledge and the inventor gets a period of exclusivity for the using/making etc that knowledge.

Where the system falls down is in the "disclosure to the public" part of that bargain. Patent language has become so abstracted that most of the time it fails to inform someone of skill in the art how to actually make the promised item. This has some corollary effects:

- The patent system becomes less useful as a source for knowledge (as argued by the article) - Novelty and obviousness (and later, infringement) all become less certain. This results in poor quality patents coming out of the patent office and high costs to litigate for infringement.

In the end, though, the result is that the patent system is (for most of the computer arts where I spend my time) not very useful as a store of knowledge. If there is a choice, I will almost always seek out a paper (or source code) rather than an accompanying patent to get an understanding of new technology.

This is due to the incentives associated with each type of publication. The source code is designed to convey certain algorithms to the computer in an unambiguous way. Papers are designed to get read and cited - and they don't do that unless they communicate what is new and meaningful to an audience of peers.

In contrast, patents are designed to cover a range of possible solution spaces, with as much room for reinterpretation later as is possible without being insoluably ambiguous.

This problem is hard to undo because the push to abstract the descriptions in patents was done with the best of intentions. - Congress wanted to allow new things to be described without constricting the form of the language used. - Prosecuting lawyers (and their clients) wanted to get the maximum coverage for their new ideas. - Courts want to reward inventors and disallow trivial workarounds (nail->screw) not addressing the core inventive concept.

All of these points are valid ones - but in combination, the result is that patents have become too vague, particularly in anything computer-related.

As an aside, the mechanical arts (and some others, such as many bio/chem arts) don't suffer from this problem to nearly the same extent, and so you usually don't see the same problems.

[rogerbinns]

I can substantiate this. I've done a few patents (all owned by employers) which take years to go through. When finally granted, I go and read them, and even though I am the "inventor" I can't make any sense of what the heck they are about! They are in no way disclosed in a form and language that makes it easy to fundamentally understand.

[Houshalter]

>If there is a choice, I will almost always seek out a paper (or source code) rather than an accompanying patent to get an understanding of new technology.

On the other hand patents do allow companies to publish their work without fear of losing their trade secret, and to invest in research in the first place. The papers might not be published at all if not for patents. For example Google, Facebook, Microsoft all fund and publish machine learning research now.

Of course this is probably only true for a small fraction of all patents, and probably doesn't benefit small players very much.

[ajb]

"the patent system is (for most of the computer arts where I spend my time) not very useful as a store of knowledge"..."This is due to the incentives"

Actually, the problem is more fundamental than that. For thing like medicines, the standard naming scheme for chemicals means you can reliably find patents covering a particular compound. Rice's theorem [1] means that that this cannot be done for software.

[1] https://en.wikipedia.org/wiki/Rice%27s_theorem