Empirical research reveals three big problems with how patents are vetted | Hacker News Reader

Empirical research reveals three big problems with how patents are vetted

arstechnica.com

87 pointsBerislavLopac26 comments

Great suggestions in this article:

* eliminate the issuance fee (that is paid when a patent is granted) and raise the application fee to compensate, to eliminate the perverse incentive for the PTO to grant patents

* use renewal fees to rebate application fees for individual and qualified small-business applicants (the amount of the rebate depending on the amount of renewal money coming in)

* only allow an application to be refiled once; the second rejection is final

* allow examiners somewhat more time to examine each patent

All of these would help the situation.

Finally, the article argues against funding the PTO through fees. I can put it no better than the author: "While a user-funded patent office sounds good in theory, the reality is that the social costs of bad patents are high — and they're borne by the general public, not patent holders."

Seems like it would be better if modeled after the criminal justice system. You have to prove beyond a reasonable doubt that the invention is novel and patent worthy. And duration of the patent should be dependent on its scope, and reviewable in the future. Of course this may be way more expensive, can't say I've calculated out the economics of it.

Except for variable length patent terms (duration), everything you suggest is true already.

The applicant has to prove to the examiner that their invention overcomes the prior art discovered by the examiner (or anyone). Also, inventors (and their attorneys) have to disclose to the USPTO all relevant public references (potential prior art) they know of.

Further, during patent prosecution, US examiners are allowed (required) to broadly interpret claims and prior art. Giving them more latitude than courts to declare something is prior art for a given invention.

As coinventor of a now-expired patent, I'm not very impressed by the bar for novelty.

There has been some truly new stuff patented, like the RSA cryptosystem. It's not clear to me whether protecting things like that has been a net benefit even before the cost of also protecting ordinary engineering and worse.

The novelty bar is low by design. An invention needs at most one difference from prior art to be novel. It can be a small difference. But if a patent relies on a small difference for novelty it is likely to be a narrow patent or otherwise easy to design around. Note, non-utility differences, like color or artistic design elements do not count in utility patents.

Non-obviousness is harder prove because multiple references can be combined and there is more room for subjective interpretation or argument. Overcoming obviousness rejections is where patent attorneys earn their money. If the filed specification is well-done, obviousness rejections can almost always be overcome at the expense of narrowing the patent claims.

Further, patents encourage innovation in at least two ways:

The classic view is that they reward inventors by granting them a limited term monopoly in their invention in exchange for public disclosure.

Though I think more innovation is spurred by people designing around existing patents. For example, Amazon's old one-click patents didn't stop the e-commerce industry from innovating. Neither did Priceline's patents on reverse on-line auctions.

Patents protect how someone does something, not what someone does. Two inventions that do the same thing can be patented, as long as how they do it is different. RSA patents only mattered because people wanted to use RSA's crypto. They didn't stop others from making or patenting different kinds of crypto or getting their own patents on novel extensions to RSA's crypto.

> But if a patent relies on a small difference for novelty it is likely to be a narrow patent or otherwise easy to design around.

Sure, if you're an omniscient being aware of every patent in existence and well-funded enough to defend against the people who have "patented the internet" (or what have you).

Another problem with this: engineers get instructed to for god's sake don't read or discuss patents, for fear of treble damages.

yeah, that kind of sucks. Though not only does such a policy reduce the likelihood of a bad outcome re willful infringement. It can save $$$ during litigation because it will not be a real issue.

Businesses deal with all kinds of inconvenient things. Patents are one of those things. Welcome to the big leagues so-to speak. Having a tight 10-50 patent portfolio around your key tech will go along way towards helping a small company play with the big boys. A quality patent portfolio can be a powerful equalizer if you are doing something innovative and it is protected by patents.

This is just one more anecdote, but I once read Sun had a cool new patent, the central idea of which turned out to be three lines of assembly code which had already been published in the well-known superoptimizer paper by Massalin 15 years before -- I only noticed this because I happened to have been reading it and had the paper lying around my bedroom. To Massalin it merited one paragraph in an appendix about interesting snippets discovered by the superoptimizer. Sun gave it pages of legalese broken out into multiple claims, ending with the usual boilerplate claiming any further variations not explicitly listed that would be obvious to one "skilled in the art".

I did nothing about this because the fight would be a drag with nothing in it for me.

Yes, the classic justification is to encourage invention, but there are publications arguing that patents aren't a net positive. It's not my field, I don't know, but I've started doubting.

Added: I don't think Sun did anything especially wrong -- they worked with the patent system as it is and missed the prior art. What I think is that this example shows some of what's wrong with that system.

That was just a mistake that slipped through the system. It would be caught later if the patent was ever asserted against anyone. Sun might have caught it themselves as part of their evaluation process before asserting the patent.

Now examiners have better access to all types of references. And, it is easier to search through them. Academic papers, foreign patents, foreign publications, books, website copy, etc, are all fair game. Though examiners seem to prefer to use US Patents/Applications if they can. US patents or published applications make great prior art references because they have a clearly defined priority date and patent prosecutors tend to draft specs to be as inclusive as possible.

It's encouraging to hear the situation may be getting better.

I brought up my example because it shows a friction-to-innovation ratio that seems way out of proportion and because the missed prior art makes the comparison unusually clear. The academic paper treated the idea as almost a throwaway -- I can imagine it as one of the first things to be dropped on tighter page limits. (By dim memory -- I ran into this over a decade ago.) More usually when a patent gets argued about, the interested parties make the matter harder to follow for the rest of us without the attention to spare. I think most of us, if we're fair-minded, when we read about one of these disputes should go "That sounds like ordinary junk but the law is complicated, the patent is complicated, my opinion about this case won't be worth much unless I spend way more effort than I can justify." Here was a patent that just by chance I could easily judge, which didn't otherwise look different from others I've heard about. (Maybe I could dig up the patent number since I've been harping on it. I'm not sure.)

Having lots of friction from many small inventions helps the big established companies more. My own patent, it'd never have occurred to me to patent it -- but it was work for a tiny startup that needed a defense in this system. (I don't think this patent was a bad one as patents go. It got referenced by a VPRI STEPS paper.)

A patent system could encourage growth, or it could get mired in a rent-seeking equilibrium. It seems hard to tell which we've got, but I'd bet on the latter if we knew how to settle this.

There's also a freedom-of-thought cost, in that independent invention happens all the time, unlike the case with copyright. That is, even an unarguable finding of welfare benefit wouldn't settle the question if patents are right. It'd have to be a substantial benefit to offset the cost to liberty.

Ah, but the applicant does not have to prove that the invention is nonobvious. I think that objective evidence of nonobviousness should be required, not at the time of application — the needed evidence might not exist yet at that point — but as the first step in an infringement proceeding. As for what would constitute objective evidence of nonobviousness, I think the Graham factors [0] are a great starting point:

* commercial success

* long felt but unsolved needs

* failure of others

[0] https://en.wikipedia.org/wiki/Graham_v._John_Deere_Co.

Graham factors will get looked at closely in litigation because post-grant challengers always allege that the patent claims they are challenging are invalid because they are obvious. Patent owners will defend charges of obviousness by: trying to distinguish their patent claims from the prior art; arguing that combining the prior art references is unreasonable; and by arguing that the Graham factors tilt in their favor.

If a claim is obvious, it is easier to prove it by finding written evidence (references) and then argue that it would be reasonable to combine those references. If a challenger has to rely on an absence of Graham factors they are in trouble. That means they could not find any published prior art references in the entire world that would be reasonable to combine to disclose all of the elements of the claims being challanged.

Note, litigators will raise every other reason they can think of to invalidate the claims they are challenging.

During examination, obviousness is often the sticking point. In crowded arts it can take several rounds of prosecution to overcome obviousness rejections -- as it should.

> If a challenger has to rely on an absence of Graham factors they are in trouble.

Right. I think the law should be changed.

The Amazon photography patent [0] is a great example of why. Although there were no prior art references specifying the distance ratio given in claim 3, there is also no evidence that nobody had used that ratio before. Nor, in all likelihood, is there any evidence that anyone had tried to achieve the same effect as this patent (eliminating the need for retouching product photographs) and failed.

Although this example is not from the software domain, from what I have seen, similar circumstances occur all the time in software: a patent application covers a technique that simply can't be known not to have already been discovered, because the problem it solves is simply not important enough that someone who solved it would necessarily have thought the solution worth patenting. Or, because they did so at a time when software was not generally thought to be patent-eligible, a situation that, as I recall, persisted among software engineers well into the 1990s.

In my view, and in the software domain, the inventions that deserve patent protection are primarily those that solve problems whose difficulty has been explicitly noted in the computer science literature. For those, we have very clear evidence that any solution that improves on previous attempts is nonobvious. Absent such evidence I think the public would be better served, on the whole, if software patents were not granted.

This is not quite the scorched-earth view that some here on HN take with regard to software patents, referring to those who think they should never be granted at all, but my impression is that it's pretty close to the EU policy.

[0] https://arstechnica.com/tech-policy/2014/06/how-amazon-got-a...

I want to measure novelty as the length of time since all of the components of the thing, including the market for the thing, have existed before the patent was filed. You then grant patent for some constant multiple of however long that period was.

If they wanted to get more extreme, they could take the initial filing fee as a deposit and only refund it with successful applications. That would help prevent the persistent applications and incentivize the USPTO to reject more. In the eyes of the examiner, the value of the patent has to be worth more to them through society than the individual filing fee.

Patent issuance is not based on their worth to society. Utility, Novelty, and Non-obviousness is all that's required.

Besides, how could an examiner evaluate the importance of a patent that has a 20 year term. How could anyone?

Part of me wonders whether a solution may be to do the following:

1. Subject all patent applications to a two month public appeal period, where any objections from the general public can be received for the patent 2. Then, these objections are reviewed by the patent examiners, with any patent applications that turn out to already exist in some form being invalidated 3. And all rejected patent applications being subjected to a Git style diff setup, where anything with the same wording is immediately rejected before it enters the system.

Add this to a requirement to 'have some plan for manufacturing a product based on the patent (ie, no way to register patents just for licensing or court battles), and that'd probably go a long way towards making the patent system more fair and stop bogus patents getting approved.

1 and 2. There are various public appeal periods but because they require examiner time and force applicants to pay attorneys to answer the objections they cost a lot of money.

3. That would be some awesome ML to pull that off. Since by law the patent applicant is his or her own lexicographer - within reason, they can define the meaning of the words in their application

Requiring plans for manufacturing will never fly. Universities like to patent stuff. Also, via R&D an inventor can identify more than one invention but may have just enough resources to produce one, or they may have "plans" to produce them later, or license to partners for production, etc.

edit: typo - missed your number 3

"objections they cost a lot of money"

The premise being that low cost is optimal.

The person raising the objection (the petitioner), pays the USPTO fees of $6000 to $18000+ depending on the type of post issuance review requested. The more expensive and complicated ones are more trial-like and require more back and forth between the petitioners, the patent owners, and the USPTO. This requires more attorney time which makes them very expensive to pursue.

edit: to add - still very "cheap" compared to real trials

They should up the patent application fee, say by 50%. Then spend this extra money on reviewing objections from the public (if any). Then the public or interest groups can actually help find prior art.

Subject to some timing requirements, the USPTO allows for anonymous/third-party submission of prior art before a patent issues.

I think there is a small per document fee.

https://www.uspto.gov/web/offices/pac/mpep/s1134.html#d0e121...

There should be a financial penalty to the office issuing a wrong patent is shown to be invalid later on. Right now the incentives are perverse, no reason to not let bogus or useless patents through.

Probably one would also need to have some of the patents reviewed by 3rd party within a year or two after approval, to ensure faulty patents are detected within a timeframe that makes the feedback relevant to the original examiners. Maybe pay the reviewer office based on number of parents invalidated.

This can't work. Patent law changes and is often retro-actively applied.

Though to your point, AFAIK examiners and their bosses hate to have patents they allow be overturned. It is noticed. But often, patents are invalidated based on changes to the law that occur after a patent was allowed.

One could exclude those cases? Of course cannot punish based on changes at a later point.

Reply on news.ycombinator.com