IAmA Patent Examiner on reddit
reddit.com
reddit.com
Consider the various compositions of linked lists that are patented (http://www.google.com/patents/US7028023), any of which would be casually "re-invented" by any competent developer. The USPTO should give the world more credit for "obviousness". I propose that the closure of established concepts under the algebra of established compositional operations shall be considered "obvious", in which case we can all go about applying those compositional operations without worrying about stumbling over patented combinations.
If you are looking for answers as to how some of the ridiculous patents manage to get through the patent office (including patents for software concepts that engineers are likely to consider "obvious" or for which there is likely "prior art"), you are unlikely to find the answers here.
Whenever someone defends software patents, I bring up Amazon's one-click patent. It's so obvious! I use the one-click-checkout feature on Amazon all the time and dearly miss it on every other website out there. Imagine if Pizza Delivery sites or Flower Delivery sites had that feature. People would order things a lot more if they didn't have to go through extra screens to confirm things every single time. The funnel gets smaller as the number of steps rise. And now one company has a patent on short funnels.
For instance, this question from your linked thread: http://www.reddit.com/r/Android/comments/ww982/iama_patent_e... was answered satisfactory. There is a difference between
"a plurality of heuristic modules, wherein: each heuristic module corresponds to a respective area of search and employs a different, predetermined heuristic algorithm"
and
"searching different sources of data wherein you search each with a method that makes sense for its contents"
But the replies from random commenters and vote totals would have you believe the opposite.
To me (and I know I'm simplifying just like the random commenters you speak of), the patent is _intended_ to cover 1. performing heuristic searches on multiple areas (modules) 2. using a different search on each area
However, it uses the word "predetermined" when defining the heuristic algorithm to use for each area. If my code randomly picks between two slightly different heuristics (even with a very low probability), would it still be covered under the patent?
Also it says "heuristic algorithm" meaning singular algorithm. If I ran 2 algorithms in parallel (even one that always returns no results) and then chose the "better" of the two, would that get around the patent too?
Are patents vulnerable to this sort of work-around due to subtle interpretations of wording, or is it really the intent of the patent that matters?
[0]: https://twitter.com/id_aa_carmack/status/137189212519792640
It seem every last adjective counts.
"In order to jump to an obvious argument, you'd have to prove this device reads on all the limitations minus the unlock image that is slid. Then you'd have to show something else that uses an image to unlock something. Then you'd have to provide rationale why the combination between the two things would have been obvious at the time of the invention, in this case 2005."
In other words, obviousness might seem, uh obvious as a defense for most this stuff. But actuality it's categorically dismissed with the "you can't prove it" argument. Plus the "but was it obvious in 2005, with phone!" (lovely to see those so-and-so's trying to wiggle out of that one but essentially coming back to it).
What I believe the commenters were getting at is that those aren't very meaningful limitations. One of the most straightforward implementations of a system where a user can search multiple orthogonal sources of information would be to define separate algorithms in a pluggable modular form. I also imagine most search algorithms could be described as having heuristic components, probably especially those tailored for a particular type of information.
I don't think this qualifies as one-click.
According to them this "presumption of patentability" (as I would call it) was a policy change that happened not very long ago, as before the policy was to reject as many patents as possible.
I would also make a wild guess: That this change was lobbied for by the IP law firms, because they are the only ones which really benefit from it.
I, like many others, have run across so many patents that should not have been issued at all. This isn't limited to software. There are examples across disciplines, like the guy who got a patent for running wires inside clothing to connect devices or the company that got several patents on using pulse-width-modulation to control the intensity of LEDs to mix colors or most software patents.
In hardware I have seen lots of patents that cover the exact same thing. I have seen patents for things that don't work. And, patents for things that I knew had been available or done prior to the filing date.
I call these "implementation" patents rather than true inventions. The term refers to the fact that a modern skilled engineer or team of engineers would be more than capable of coming up with that or a myriad of other implementations when given a problem to be solved. That's what engineering is all about. Many engineers across time and distance are able to solve similar problems and end-up with similar solutions. This is what we are trained to do. This is also the part of the human and other animals: the use of our brains to solve problems.
It seems that the patent system ought to be reformed in such a way that the rules recognize who we are, what we are taught, what we do, what we are capable of and, in general, the fact of knowledge acceleration.
Today it should be exceedingly hard to receive a patent on just about anything. Patents should be reserved for true invention and not implementation. Today's environment is such that there's so much distributed knowledge across so many disciplines that most "inventions" that are patented are really cases of someone choosing a solution to a problem (not inventing one) and then hiring an attorney to try and monopolize it. Case in point: The Apple "slide to unlock" patent. Someone chose that solution from available knowledge. No sane person would suggest that this qualifies as an invention by any possible imaginative stretch. This was a UI decision that the lawyers at Apple got a patent for. They got a patent for, effectively, the analog of a mechanical slide switch made to look nice, nothing less. Not an invention, but they have a patent to say otherwise.
I am not against patents. I am against the granting of patents for non-inventions. Today the granting of a patent should almost be so rare as to make national news. Just a few per year. In fact, they should be rarer and rarer with time, to the point that those granted the few patents that should make it through the system per year would be invited to the White House to receive a Presidential Innovation award or some such thing.
It only stands to reason that, as technology advances, there will be less and less true inventions per year. Part of the problem was alluded to on the reddit thread: the USPTO makes money and Washington takes it away. Fix that and you might just fix the system.
Maybe the patent review process should be open to anyone. Yes, your competitors, anywhere in the world, get to examine and weigh-in. If an invention passes such a test it is probably deserving of a monopoly for twenty years. This also means that someone will really think hard before spilling all the beans in front of the entire world. Only a true invention, where other engineers go "This guy deserves a patent for this!" should be granted patents. And rightly so.
I mean, consider Apple's bouncing lists patent. When you use it, it's very intuitive, and any moderately skilled software developer could implement it in an afternoon. So it seems obvious, but the question is then - why did nobody do it before? Touchscreen devices existed previously, so that seems to suggest that it is non-obvious. Synthesizing a drug is not hard or novel, but arriving at the conclusion that this is indeed the correct drug that is the reason that pharma patents exist. So the same could be said of bouncy lists.
So how do you determine the obviousness of this, objectively? The most straightforward ways are prior art and to to consider claims from previous patents. It's mostly objective and is hard to argue against, as opposed to one patent examiner's opinion on obviousness.
The problem with software patents is that the benefits of a 20 year monopoly far outweighs the benefit of having it released to the public - which is not true for all patents. If Apple couldn't patent the bouncing list, they'd still have bouncing lists, and it would still be an advantage to their devices, even if everyone else copied them. If you couldn't have a patent on a drug that was going to take $1 billion in costs to bring to market, yet would only cost pennies to synthesize, you wouldn't even bother trying. So the question is instead, how do you objectively determine which industries require patent protection, and which don't. For the rest of the industries, copyright and trademark laws are more than enough protection.
However, this is precisely what is happening in the software industry. In fact, typical legal advice for engineers is to not read patents at all. And still, it is impossible to write any sort of complicated program without running a foul of numerous software patents.
A truly innovative patent would be one which would be fairly unlikely to be rediscovered independently. How likely is it that a typical person skilled in the art would have come up with PageRank if the algorithm was not published? Another academic might have, but it would not be likely that an average programmer would.
On the other hand, that linked list patent? That is not innovation. The idea of searching multiple databases from a single search box? Not innovative. Searching for and identifying entities like phone numbers in text? Not innovative.
Usually, these obvious "innovations" are simply a combination of a technology that might or might not have existed before, plus a need or an opportunity that didn't exist before. Bouncing lists are one example; one-click ordering is another. For the list case in particular, Apple needed a touchscreen and a fast CPU, plus a reason to implement it. That combination didn't exist in general-purpose portable consumer devices before the iPhone.
We already have a mechanism for rewarding those who recognize opportunities and fill needs: it's called a "market." No additional monopoly protection is appropriate in such cases. When patents are granted on the basis of a foot race to the USPTO, everyone (except the lawyers) loses.
Basically, we're rewarding the first people who apply the easiest solutions to problems that few others have even encountered. Instead, we should be limiting the use of patents to cases where a difficult problem has been solved by means that might otherwise be kept locked away as a trade secret.
To me, the key problems with the notion of obviousness written into current patent law are that (a) it is mostly subjective and (b) the burden of proof is on the PTO, if the examiner feels the claimed invention is obvious.
I think the burden should be on the applicant to provide objective evidence that the solution is not obvious. While I suppose this could take various forms, it is hard to see how any such evidence could exist when the problem addressed by the claimed invention is itself novel. If the problem is not yet widely recognized as a problem, there has been little opportunity for anyone else to try to solve it and fail. To show nonobviousness under my proposal, the applicant would have to show that the problem has been known or at the very least that it has been a problem, even if not identified as such.
The applicant should talk more about the decisions made during the invention process, and tell us about why, in addition to the how. If there are more obvious alternatives which are all deadends, the working solution may be considered not so obvious.
An equally bad (or perhaps worse) problem with all of this is that if you are the target of patent litigation, even winning the case will likely bankrupt you if you are an independent developer or small firm. I'd like to see much more reform around how patent cases are litigated - limit forum shopping, require the plaintiff to pay legal fees if they are an NPE and they lose the case, make PTO reviews quicker and less costly, etc...
Side note: This issue is sort of related to all of the talk about smaller developers "selling out" - patent litigation is a huge, huge risk right now because our patent and court systems are so dysfunctional, and if I built something that was getting popular, selling out to a larger company is a pretty good defense against that.
Obviousness is not based on existing patents only. It is based on anything that is published before the relevant date. Thus, examiners should and sometimes do use all kinds of different publications to determine obviousness. But the problem is that the patent office is accustomed to do prior art searches based on existing patents. Thus, internally the PTO is set up to favor prior art searches based on patents rather than other publications. This generally makes sense for other slower moving fields of technology (e.g. mechanical devices, farming equipment, etc.) or for fields where everything gets immediately patented (biotech) but it makes no sense in software.
Another problem is that software was considered generally unpatentable until a landmark supreme court case in 1995. At that time people started filing for software patents and the patent office was generally set up to mostly use patents as prior art searches. Well, there were no software patents then, because software was not patentable until then. The unsurprising result was that a lot of very obvious software patents got through.
Thus, my pet theory on software patents is that there is nothing inherently wrong with them, they were just not correctly administered by the PTO. If the USPTO hires proper competent computer scientists as examiners and encourages the searching of all available prior art, then we would get much fewer and much higher quality software patents.
And I do agree with your idea of opening the review process. The pto is trying to make tepid steps in that direction but they are obviously scared of a flood of comments that may happen if they open up the gates and them having to pay examiners to evaluate every single comment from every crazy person on the net.
The amount of ground-breaking computer science done in the 1950s continues to amaze me. They were inventing things that only today are seeing many concrete applications.
There are so few things that are unique enough to even merit a patent, and fewer of those things that should be patented.
If every computer science paper on every new technique were patented you would be unable to do anything useful with a computer without paying millions in licensing fees.
I think for the greater good of the industry software patents should be eliminated entirely. If you need to protect something, keep it a trade secret.
Agree, to the point where I could let this summarize my view on patent related problems. There is obviously something wrong with the USPTOs cultural definition of obviousness. Part of it is as you say that they see patents as the only prior art (in practice - I know it's not so in theory). But another part is that they seem to think that the slightest difference in comparison to prior art suddenly makes it non-obvious. That's ridiculous. The obviousness of a patent claim has to be judged in light of prior art. The test should be that the difference is non-obvious "to a person skilled in the art".