Categories
Uncategorized

An Open Letter to the USPTO

Following the Supreme Court’s decision in Bilski v. Kappos, the United States Patent and Trademark Office (USPTO) plans to release new guidance as to which patent applications will be accepted, and which will not. As part of this process, they are seeking input from the public about how that guidance should be structured. The following is an open letter than I have sent to the USPTO at Bilski_Guidance@uspto.gov. More information is available at http://en.swpat.org/wiki/USPTO_2010_consultation_-_deadline_27_sept and http://www.fsf.org/news/uspto-bilski-guidance. As with all of my posts, the following represents my personal opinion and is not the opinion or policy of my employer.

To whom it may concern at the United States Patent Office:

Since completing my undergraduate studies in mathematics and computer science at the Massachusetts Institute of Technology (MIT) and my doctorate in computer science at Carnegie Mellon University (CMU), I have spent my entire professional life in software research and development. I have worked at large software companies, such as IBM, AT&T, and Google, and I also was a founding employee at Endeca, an enterprise software company where I served as Chief Scientist. I am a named inventor on eight United States patents, as well as on eighteen pending United States patent applications. I played an active role in drafting and prosecuting most of these patents. I have also been involved in defensive patent litigation, which in one case resulted in the re-examination of a patent and a final rejection of most of its claims.

As such, I believe my experience gives me a balanced perspective on the pros and cons of software patents.

As someone who has developed innovative technology, I appreciate the desire of innovators to reap the benefits of their investments. As a founding employee of  a venture-backed startup, I understand how venture capitalists and other investors value companies whose innovations are hard to copy. And I recognize how, in theory, software patents address both of these concerns.

But I have also seen how, in practice, software patents are at best a nuisance and innovation tax and at worst a threat to the survival of early-stage companies. In particular, I have witnessed the proliferation of software patents of dubious validity that has given rise to a “vulture capitalist” industry of non-practicing entities (NPEs), colloquially known as patent trolls, who aggressively enforce these patents in order to obtain extortionary settlements. Meanwhile, the software companies where I have worked follow a practice of accumulating patent portfolios primarily in order to use them as deterrents against infringement suits by companies that follow the same strategy.

My experience leads me to conclude that the only beneficiaries of the current regime are patent attorneys and NPEs. All other parties would be benefit if software were excluded from patent eligibility. In particular, I don’t believe that software patents achieve either of the two outcomes intended by the patent system: incenting inventors to disclose (i.e., teach) trade secrets, and encouraging investment in innovation.

First, let us consider the incentive to disclose trade secrets. In my experience, software patents fall into two categories. The first category focuses on interfaces or processes, avoiding narrowing the scope to any non-obvious system implementation details. Perhaps the most famous example of a patent in this category is Amazon’s “one-click” patent. The second category focuses on algorithm or infrastructure innovations that typically implemented as inside of proprietary closed-source software. An example in this category is the patent on latent semantic indexing, an algorithmic approach used in search and data mining applications. For the first category, patents are hardly necessary to incent disclosure, as the invention must be disclosed to realize its value. Disclosure is meaningful for patents in the second category, but in my experience most companies do not file such patents because they are difficult to enforce. Without access to a company’s proprietary source code, it is difficult to prove that said source code is infringing on a patent. For this reason, software companies typically focus on the first category of patents, rather than the second. And, as noted, this category of innovation requires no incentive for disclosure.

Second, let us ask whether software patents encourage investment in innovation. Specifically, do patents influence decisions by companies, individual entrepreneurs, or investors to invest time, effort, or money in innovation?

My experience suggests that they do not. Companies and entrepreneurs innovate in order to further their business goals and then file patents as an afterthought. Investors expect companies to file patents, but only because everyone else is doing it, and thus patents offer a limited deterrent value as cited above. In fact, venture capitalists investing in software companies are some of the strongest voices in favor of abolishing software patents. Here are some examples:

Chris Dixon, co-founder of software companies SiteAdvisor and Hunch and of seed-stage venture capital fund Founder Collective, says:

Perhaps patents are necessary in the pharmaceutical industry. I know very little about that industry but it would seem that some sort of temporary grants of monopoly are necessary to compel companies to spend billions of dollars of upfront R&D.

What I do know about is the software/internet/hardware industry. And I am absolutely sure that if we got rid of patents tomorrow innovation wouldn’t be reduced at all, and the only losers would be lawyers and patent trolls.

Ask any experienced software/internet/hardware entrepreneur if she wouldn’t have started her company if patent law didn’t exist. Ask any experienced venture investor if the non-existence of patent law would have changed their views on investments they made. The answer will invariably be no (unless their company was a patent troll or something related).

http://cdixon.org/2009/09/24/software-patents-should-be-abolished/

Brad Feld, co-founder of early-stage venture capital firms Foundry GroupMobius Venture Capital and TechStars, says:

I personally think software patents are an abomination. My simple suggestion on the panel was to simply abolish them entirely. There was a lot of discussion around patent reform and whether we should consider having different patent rules for different industries. We all agreed this was impossible – it was already hard enough to manage a single standard in the US – even if we could get all the various lobbyists to shut up for a while and let the government figure out a set of rules. However, everyone agreed that the fundamental notion of a patent – that the invention needed to be novel and non-obvious – was at the root of the problem in software.

I’ve skimmed hundreds of software patents in the last decade (and have read a number of them in detail.) I’ve been involved in four patent lawsuits and a number of “threats” by other parties. I’ve had many patents granted to companies I’ve been an investor in. I’ve been involved in patent discussions in every M&A transaction I’ve ever been involved in. I’ve spent more time than I care to on conference calls with lawyers talking about patent issues. I’ve always wanted to take a shower after I finished thinking about, discussing, or deciding how to deal with something with regard to a software patent.

I’ll pause for a second, take a deep breath, and remind you that I’m only talking about software patents. I don’t feel qualified to talk about non-software patents. However, we you consider the thought that a patent has to be both novel AND non-obvious (e.g. “the claimed subject matter cannot be obvious to someone else skilled in the technical field of invention”), 99% of all software patents should be denied immediately. I’ve been in several situations where either I or my business partner at the time (Dave Jilk) had created prior art a decade earlier that – if the patent that I was defending against ever went anywhere – would have been used to invalidate the patent.

http://www.feld.com/wp/archives/2006/04/abolish-software-patents.html

Fred Wilson, managing partner of venture-capital firm Union Square Ventures:

Even the average reader of the Harvard Business Review has a gut appreciation for the fundamental unfairness of software patents. Software is not the same as a drug compound. It is not a variable speed windshield wiper. It does not cost millions of dollars to develop or require an expensive approval process to get into the market. When it is patented, the “invention” is abstracted in the hope of covering the largest possible swath of the market. When software patents are prosecuted, it is very often against young companies that independently invented their technology with no prior knowledge of the patent.

http://www.unionsquareventures.com/2010/02/software-patents-are-the-problem-not-the-answer.php

In summary, software patents act as an innovation tax rather than a catalyst for innovation. Perhaps it is possible to resolve the problems of software patents through aggressive reform. But it would be better to abolish software patents than to maintain the status quo.

Sincerely,

Daniel Tunkelang

By Daniel Tunkelang

High-Class Consultant.

112 replies on “An Open Letter to the USPTO”

I like the argument that software is algorithms, algorithms are mathematics, and mathematics is not patentable. I am a named inventor on a software patent, US 7366790, and it’s very much a mathematical algorithm.

Like

Here is what the USPTO says about the patentability of “mathematical algorithms”:


If the “acts” of a claimed process manipulate only numbers, abstract concepts or ideas, or signals representing any of the foregoing, the acts are not being applied to appropriate subject matter. Gottschalk v. Benson, 409 U.S. 63, 71 – 72, 175 USPQ 673, 676 (1972). Thus, a process consisting solely of mathematical operations, i.e., converting one set of numbers into another set of numbers, does not manipulate appropriate subject matter and thus cannot constitute a statutory process.

In practical terms, claims define nonstatutory processes if they:

– consist solely of mathematical operations without some claimed practical application (i.e., executing a “mathematical algorithm”); or

– simply manipulate abstract ideas, e.g., a bid (Schrader, 22 F.3d at 293-94, 30 USPQ2d at 1458-59) or a bubble hierarchy (Warmerdam, 33 F.3d at 1360, 31 USPQ2d at 1759), without some claimed practical application.

Cf. Alappat, 33 F.3d at 1543 n.19, 31 USPQ2d at 1556 n.19 in which the Federal Circuit recognized the confusion:

The Supreme Court has not been clear . . . as to whether such subject matter is excluded from the scope of 101 because it represents laws of nature, natural phenomena, or abstract ideas. See Diehr, 450 U.S. at 186 (viewed mathematical algorithm as a law of nature); Gottschalk v. Benson, 409 U.S. 63, 71-72 (1972) (treated mathematical algorithm as an “idea”). The Supreme Court also has not been clear as to exactly what kind of mathematical subject matter may not be patented. The Supreme Court has used, among others, the terms “mathematical algorithm,” “mathematical formula,” and “mathematical equation” to describe types of mathematical subject matter not entitled to patent protection standing alone. The Supreme Court has not set forth, however, any consistent or clear explanation of what it intended by such terms or how these terms are related, if at all.

Certain mathematical algorithms have been held to be nonstatutory because they represent a mathematical definition of a law of nature or a natural phenomenon. For example, a mathematical algorithm representing the formula E = mc2 is a “law of nature” – it defines a “fundamental scientific truth” (i.e., the relationship between energy and mass). To comprehend how the law of nature relates to any object, one invariably has to perform certain steps (e.g., multiplying a number representing the mass of an object by the square of a number representing the speed of light). In such a case, a claimed process which consists solely of the steps that one must follow to solve the mathematical representation of E = mc2 is indistinguishable from the law of nature and would “preempt” the law of nature. A patent cannot be granted on such a process.

http://www.uspto.gov/web/offices/pac/mpep/documents/2100_2106_02.htm

For example, that rules out patenting singular value decomposition of a matrix but not its practical application to indexing and retrieval. That doesn’t do much to restrict the patentability of software. And it’s even more of a stretch to argue that user interfaces are “mathematical algorithms”.

So I think we need a different line of attack.

Like

Good points. What is your view of the LZW patent, should algorithms of that nature be patentable? I think LZW became infamous because it was published and then widely adopted without anyone realizing the existence of a patent. If it had been originally invented at CompuServe, and quietly incorporated into the GIF file format, with everyone knowing from the start that it was proprietary, that file format would have gone nowhere, while inspiring all the free formats that did arise without the use of the patent. The monopoly benefits of GIF could not have been realized without the stealth/trap tactic. Early enforcement of the monopoly would have rendered it worthless.

Other famous algorithms that come to mind are FFT algorithms (none patented as far as I know, but very important), and CDMA (patented and used for profit and public benefit)

Like

I haven’t read the LZW patents, so I can only comment in general. I certainly don’t like submarine patents in general. I understand when patent holders practice delayed or selective enforcement because of the high costs of litigation, but I don’t think it’s ethical to stealthily let innocent parties wander into the trap of unknowing infringement and then ambush them.

But to answer your question more generally, I’m glad the USPTO doesn’t think that “mathematical algorithms” are patentable subject matter, and I wish it would go further and not consider software to be patentable.

I feel that patents make sense for subject areas where, without the prospect of patent protection:

1) The state of the art would advance more slowly because advances would be kept as trade secrets.

2) People would avoid investing in innovation because they’d feel copiers would siphon off the return on that investment.

As I argued in the post, I don’t think the above are true for the material associated with software patents.

Like

I think a lot of the problems with software patents could be rectified, and their benefits (such as they are) preserved if we recognized that software development is just _faster_ than hardware and reduced the term of software patents to (say) a year or two. This would incentivize innovation without stifling it.

Like

A shorter term might help, but I feel it’s a hack. Plus, it takes 3 to 6 years to prosecute a patent, and it seems that a shorter term only makes sense if the prosecution process can also be accelerated.

Still, I maintain that software patents are not necessary to incent innovation. Indeed, I believe that today they discourage innovation. So I’m not sure there’s much baby to throw out with the bath water.

Like

A shorter term would clearly need to be paired with a shorter prosecution process. I’m agnostic regarding the benefits, but clearly there are a lot of people who believe they exist, who might be more likely to accept a shortened term rather than abolition.

Like

Fair enough: I certainly agree a shorter term would be better than a longer one, and a lesser evil would be an improvement. But, as a practical matter, I think it may be more productive to push for restricting the definition of patentable subject matter than to try to get differential term lengths based on subject matter.

Like

For this reason, software companies typically focus on the first category of patents, rather than the second. And, as noted, this category of innovation requires no incentive for disclosure.

I’m mostly concerned with the second type. And perhaps the reason that most software companies focus on the first type and not the second type is that investment in basic research is drying up. Software companies today are doing less basic research now than they were 20 years ago, or 40 years ago. Were they doing more, I think you’d see an increase in the need for type 2 patents.

I mean, suppose I were the first person to invent face recognition. Not a particular, secret, embedded algorithm for face recognition. But the whole concept of using features extracted from digitally-captured images to tell whether the faces in two images are the same or not.

You don’t think that is an invention worthy of being protected? In the history of mankind, no one has come up with a device to recognize faces before. Someone thinks of it, and creates it. And you don’t want to let them patent it?

And at the same time, you’re fine with someone being able to patent Post-It notes?

I agree with you about the silliness of patenting one-click shopping. But there are people working on fundamental game changers, putting together ideas and concepts that no one has ever put together before.

And at the time face recognition was invented, I bet there was also not a big market for it. Not a lot of digital consumer cameras. Not a lot of digital security cameras. So unless there is a patent protecting the invention, the inventor has no ability at that point to go straight to the market and immediately cash in on their invention.

To me, that would be pretty stifling. I’d want to not work on inventing completely new categories of inventions, like face recognition, if that were the case. Would be a serious de-motivator.

Like

I don’t think the issue is whether inventions are “worthy” or “deserving” of being protected. This isn’t a moral issue, it’s a question of establishing rules that establish the right incentives.

Going with your example, I’m not aware of any broad patents on face recognition that everyone in the space is licensing. It wouldn’t surprise me if such patents exist, but I doubt that the first person who developed a system for face recognition did so with patent protection in mind. Or that said person was necessarily the first person to think of developing such a system by using features extracted from digitally-captured images to tell whether the faces in two images are the same or not.

I’m not taking a stance on patents in general. Outside of software, there’s at least a lot more room for debate. But, as someone who works and has worked on game-changing technology, I wish I could do so in a world without software patents. It would not inhibit me in the least.

Like

The other examples that I talked about in the limited twitter conversions were photo stitching, speech recognition, and music analysis/extraction (converting audio into notes).

I think you’ve framed type 2 patents as specific algorithms for doing things like photo stitching. But back up for a moment, and remember that there was once a day in which the whole concept of automatically, digitally, seamlessly stitching two photos together had not yet been invented. Not only had no one come up with a good algorithm for doing it, no one had come up with any algorithm for doing it, because no one had done it.

The first person to then do it, pretty much invented it, right? And it was significant step. The world went from not having the concept of photo stitching, to having the concept of photo stitching.

I again maintain that this is at least as significant as inventing Post-It notes (if not much more so). And deserves to be protected in some way. Whether that way is through the current patent system or not, I don’t know. But the idea that there would be no protection of any kind is seriously demotivating.

Like

I’m not debating significance. Or what “deserves” protection. What I am debating is your assertion that not having patent protection is demotivating. I have seen no evidence to that effect. In fact, my personal experience and that of my colleagues suggests exactly that opposite — that developing big software-related ideas is actually easier in a world without software patents.

Like

I’m not aware of any broad patents on face recognition that everyone in the space is licensing.

Face recognition was an example. It was an example that was neither a trivial example like one-click shopping, nor an embedded algorithm, like LSA. Rather, it was a whole new category of product.

So I’m talking about establishing protection for the “next” face recognition, whatever that is.

But you feel you don’t need that. And you’ve worked on game-changing technology that you didn’t worry about patenting? Could you give me a few examples of some of the specific technology that you’ve created, that you didn’t worry about patenting? Pick something from 10 years ago or whatever, I don’t care. But something where the invention wasn’t just a one-click shopping, but also wasn’t an embedded, secret algorithm. Something where you created a fundamentally new way of doing something, of putting together concepts into a working invention, which concepts hadn’t ever been put together before.

Like

Look at all the work I did at Endeca. I think some of it was game-changing. And sure, we filed patents — it would have been irresponsible of us not to do so.
I can’t go into specifics for reason I’m sure you understand. But you can go to the USPTO site and look at my patents yourself.

Point is: I would have pursued the same work in a world without software patents. Which is why I want to live in such a world.

Like

But you can go to the USPTO site and look at my patents yourself.

No, work you did that you felt you didn’t need to patent 🙂

I can’t go into specifics for reason I’m sure you understand.

Those are for Type 2 patents, the secret, embedded LSA patents. In my comments above, I think I introduced a third category, Type 3 patents. The new-category introducing patents. The idea of face recognition, when such a thing has never been conceived of before. The idea of photo stitching, when such a thing had never been conceived of before.

Those are the things that are public-facing, not embedded in a secret algorithm that you can’t talk about. I’m not asking about your Type 2 patents. I’m asking about your Type 3 work. That’s something you should easily be able to point to, because it’s visible, is it not? People see when their software is doing face recognition. iPhoto does it, people see it. Point me to something at Endeca that was a Type 3 invention. Something that took 5 years and heavy research time to create. The first photo stitching algorithm probably wasn’t knocked out in a couple of months. It took time to figure out how to really do it. Invention time.

Look, I hear where you’re coming from.. as researchers and technologists, a lot of us would be working, on a personal level, on a lot of this stuff anyway. But we also work for companies that want to protect the work that we do. Companies that invest research salaries, and operate at a loss while we’re inventing. Believe me, I dislike patent trolls as much as the next person. But I do think there are type of inventions that do need to be protected. Let’s, again, call these Type 3 patents, building on your Type 1 and Type 2 categories above.

Like

I guess what I also don’t understand is where you draw the line at “software”. Pretty much all inventions, if you really reduce them to their key, components parts, are software.

I was thinking for example of a patent that Corning had on the type of glass that gets used in your smartphone. That glass has certain properties that make it ideal for that particular application.

But what is it that they’re really protecting, when they patent that glass? It seems to me that they’re patenting a particular chemical arrangement, right? A certain “algorithmic” structure of molecules. Attach W to X. Then loop in a Y. Surround with Z. It is an arrangement of logical units, which have certain rules about how they interact with each other, which when assembled together in a certain way produce an overall property that is greater than the sum of the parts. That is what the patent is for, ultimately. Is it not?

So how is that really different from much of the software that gets patented?

Like

The line I’m drawing is this: would an inventor’s willingness to invest the effort in developing an invention be contingent on the ability to obtain the temporary monopoly of patent protection. If the answer is no for a significant class of inventions, then patents serve a useful purpose. But if the prospect of patent protection is not a necessary condition for the investment, then it’s pure downside. That is, in essence, how I see the world of software patents.

I am not very familiar with R&D outside of software, but my understanding is that investment in materials, manufacturing processes, pharma, and other non-software fields *is* contingent on the prospect of patent protection. If that’s not true, then my objection to software patents would extend to other domains.

Like

Point me to something at Endeca that was a Type 3 invention. Something that took 5 years and heavy research time to create.

I don’t work that way, nor do any of my friends in the software industry. I don’t even know that this is how game-changing inventions happen in the software industry, e.g., search engines, social networks, cloud computing, etc.

Let me throw this back at you: point me to a game-changing software invention which was created by someone that consciously decided to invest 5 years and heavy research to develop it.

Like

Since you’re still discussing this you might enjoy the following nice analysis software patents: https://thisismynext.com/2011/08/11/broken-patent-system/ .

Besides thinking about the patent system it might be nice to think about fixing the patent _litigation_ system. One thing I think we could really fix is patent trolling. Patent trolls use bad patents to extort small amounts of money from many businesses—a divide and conquer approach since no one business is willing to spend the money necessary to litigate, even if they are likely to succeed.

So suppose we enacted a law that said, if a patent holder litigates and the patent is overturend, then all their _previous_ “winnings” from settlements on that patent must be disgorged to the current defendant? This would give the second defendant a strong incentive to defend the litigation if they thought they could win—they could make a profit instead of a loss on the defense.

From the standpoint of trolls, it becomes riskier and riskier to assert a patent, because more and more money is at stake. And being able to assert each patent only a few times would, I think, badly cut into the economics of the trolls, and hopefully make them unprofitable.

Even for non-trolls, there’d be a nice self-regulating effect. Initially, people who believe in the validity of their patent would be eager to litigate infringement. But once someone has made $100M off their patent, they’re going to become a bit reluctant to put that money at risk But the more clearly innovative the patent is, the more they’ll be willing to keep going. In other words, greater innovation brings larger profits. Just what we’d like.

This wouldn’t fix the current patent wars between Google and Microsoft, but I bet it would take the pressure off a lot of small startups.

Like

David, I love the idea of making past ill-gotten winnings subject to reclamation. That could indeed be a game-changer with respect to the incentives for NPEs and even practicing patent holders.

Do you know if there any legal precedent for this approach? I thought our legal system was like Wheel of Fortune in that once you win a prize, it’s yours to keep.

Like

would an inventor’s willingness to invest the effort in developing an invention be contingent on the ability to obtain the temporary monopoly of patent protection. If the answer is no for a significant class of inventions, then patents serve a useful purpose.

An investor? What about the company itself, investing its own money in R&D? Would the investment in R&D be contingent on the ability to secure patent protection?

I think you/we answer your own question when you write:

Point me to something at Endeca that was a Type 3 invention. Something that took 5 years and heavy research time to create.

I don’t work that way, nor do any of my friends in the software industry

So now we’re opening up a whole can of worms about whether industry invests in basic research anymore. And patents or not, I don’t believe that, on a whole, it does. People are too busy harvesting ideas (investors, engineers, etc.) to bother storing up some seed corn for the next generation.

What I am saying is that I think that there are fewer and fewer valid things to be patented, in the first place, because fewer and fewer people are working on the sort of projects that have that time scale. I mean, even in your own work experience, have you ever worked for a large corporate research lab? Something like the Bell Labs of old? Even at Google, where they are heavy into “applied” research.. they don’t do much investing in basic research, right?

So I can see that it is your experience in the industry that you’ve only seen patents that come out of more short term research. But my experience is different.

Let me throw this back at you: point me to a game-changing software invention which was created by someone that consciously decided to invest 5 years and heavy research to develop it.

Gladly. Take music information retrieval, for example. If you want to based your recommendations and playlists off of user behavior only, then you can release a product in a couple of months. However, if you want to add music extraction features (rhythm, harmony, notes, lyrics) into your recommendation, then it is going to take longer, because converting polyphonic audio into notes it a *very* hard, long term problem.

But working on the audio features is exactly what MIT startup company “The Echonest” did. In fact, here is one of the co-founders, Brian Whitman, talking about how they literally took 5 years (2005 to 2010) to get their systems to the point where they felt comfortable releasing a product:

http://musicmachinery.com/2011/05/14/how-good-is-googles-instant-mix/#comment-10534

I’d say that’s pretty slow to market, wouldn’t you?

Here’s another example: IBM’s Jeopardy-playing Watson. I don’t know if that literally took 5 years, but I do know people at IBM that started publicly talking about the project 3 years ago. I have to assume that there was work going on before that. Especially given IBM’s history of longer, term, larger-reaching projects.. like Deep Blue in 1997. Oh, and Wired magazine claims that IBM owns more technology (of which I assume a good portion is software) patents than any other technology company in the U.S.? Maybe this is too big of an inferential leap (correlation is not causation), but I have to assume that the reason IBMers are allowed to work on something like Deep Blue or Watson is because they’ll be able to patent the technologies that arise out of that work. That the only reason IBM invests money in basic, long-term R&D is because it thinks it can get that money back, through protection from its patents.

I think audio-based retrieval and recommendation is game-changing. I think Watson is game changing. And both were the result of conscious investment of many years.. 5 for sure in one case.. of a conscious decision to invest in heavy research to develop it.

Like

Daniel, IANAL. All I know is that punitive damages seem to be possible in many different legal proceedings. Perhaps the whistleblower law (which allows whistleblowers to keep a share of the ill-gotten gains they reveal) could be model.

Like

Here is the line that I am drawing: The things that you don’t like about patents, the things that you’re seeing as broken about the system, have (in my mind) not so much to do with the patents themselves, as they have to do with what is being invented. Basic research in the technology sector has been declining for decades. So there are fewer true/deep/substantial things being patented (no offense to your own inventions, of course — I’m talking broadly here).

As a result of this more shallow, more applied approach to research, more shallow, more applied patents are being created. These sorts of inventions have a high likelihood of being duplicated in the startup world, because they by definition *don’t* take 5 years to invent. And so you’re seeing a higher incidence of overlap between patents and startups, leading to the trolling behavior that shuts down startups done by your friends.

Don’t get me wrong.. I very much sympathize with that.

But what I’m trying to say is that the problem is not the patents, the problem is with our collective willingness to do basic research. Maybe it’s that U.S. industry had more cold war government funding before, and that has relatively dried up now. I don’t know why exactly things have changed. But changed they have, and that’s not the fault of patents, but the fault of our own values.

Like

I think this is becoming more a debate over what is basic research. I don’t see Echonest or Watson as being any more basic research than project I’ve worked on. Releasing early and incrementally vs. waiting 5 years to release a first version is a development strategy rather than a reflection on how fundamental the innovation is.

I have worked in traditional research labs — namely IBM Research and the old Bell Labs. That’s part of how I learned to appreciate iterative delivery.

Also, I don’t believe that “the only reason IBM invests money in basic, long-term R&D is because it thinks it can get that money back, through protection from its patents.” IBM is not especially litigious. And, while IBM does license its patents, it makes most of its money by other means.

I’m not convinced either than the current patentability of software is motivating basic research not that abolishing software patents would demotivate such research. It’s an interesting hypothesis, but I see no evidence to support it.

Like

Releasing early and incrementally vs. waiting 5 years to release a first version is a development strategy rather than a reflection on how fundamental the innovation is.

I think that’s where we simply have to disagree. If you’re building an airplane, you can’t just release it with one wing and half and engine. You have to wait and finish building it, or it’s never going to fly.

Oh, and you were at Bell? My bad.. sorry 🙂

Like

You have to wait and finish building it, or it’s never going to fly.

The point being that there is a qualitative, not just a quantitative, difference between research that takes 5 years, vs. research that you can release early and iterate on.

Like with Cornell’s fancy smartphone glass. You can’t just release an early version of the glass, something that’s brittle and weak, and then start iterating on it. You basically have to work and work and work until you get it right.

But if you don’t see a qualitative difference, but only a quantitative difference, then I can see how you wouldn’t like patents. If you think all of software is iterative, and that there is no such thing as a need to get something right, completely working, first, before you release it, then again, I can see how you wouldn’t like patents.

If what you’re saying about software is true, I would agree with you about patents, too.

Like

Oh, and for the record: Also, I don’t believe that “the only reason IBM invests money in basic, long-term R&D is because it thinks it can get that money back

I don’t think that it’s the only reason, either. But I think that patents are enough of a motivating reason that if you took them away, you would see less of this sort of research from IBM. And you don’t see evidence of this, but I see a high degree of correlation between a company’s willingness to do 5-year long research, and the number of patents they file. Again, I admit correlation is not causation. But correlation is better than no correlation.

I suppose the only way to really test this, though, would be to abolish patents, and then see what happens.

Like

I think we’ve converged.

If and only if what you’re saying about software’s iterativeness (incremental vs. 5-year-all-at-once) being a development strategy, rather than a fundamental necessity for certain classes of problems, is true. Of that I remain unconvinced and divergent.

Like

Now that’s an experiment I’d like to try for at least 5 years!

Oh, I know you do. The other experiment, of course, would be for more software companies to start to invest in basic research again. I’d like to see that experiment go on for 5 years. And then at the end of that experiment, I’d like to see how they felt about patents, and what type of work they set about patenting after that period, vs. what type of work they patent now.

So let’s do it. 5 years of your experiment followed by 5 years of my experiment.

Do we need to create one of those “Long Now” bets?

Like

The other experiment, of course, would be for more software companies to start to invest in basic research again.

But if they are not already doing so under the current patent regime, then why do you think software patents are important? Clearly they are not sufficient — perhaps they are not even necessary. I’d like to see software pioneers step up and describe work they would not have done (or would not have been paid to do) if it weren’t for the protection of software patents. Until then I will maintain my absolutist stance against software patents.

Like

But if they are not already doing so under the current patent regime, then why do you think software patents are important?

Because those two still do invest under the current regime deserve protection. Because without protection, yet another incentive for investing in basic research goes away, and it becomes less likely to happen.

And finally, because there is a delayed penalty for basic research going away. Again, it’s the seed corn analogy. If you’re busy eating the corn that you’ve just harvested, then you already have a full belly (are making money). And you have two choices.. you can eat every single ear of corn, or you can save some for planting next year. If you make the decision right now to eat all your corn, and not save any, it doesn’t hurt you right now. Your belly will be just as full, if not fuller. However, next year you’ll be paying the penalty. I think that’s what we’re seeing in software now. We’re eating this year’s corn, and next year’s penalty is just an abstract concern.

Patents are like a grain silo, to protect your grain. If you’re not going to eat all of it now, but want to save some for the future, you need to protect the grain from rats, weather, etc.

But if you don’t want to save for the future, then you don’t have need of a grain silo.

The existence of a grain silo doesn’t force us to save our corn. We can still eat it all, even if the grain silo exists. That’s what I think is happening now. But if we do want to change our ways, and save for the future, we will need to have that silo.

Like

Here is another analogy: Global vs. local maxima. I think our disagreement comes down to whether we believe the technologies that express themselves in software have an evolutionary path that is globally convex, or only locally convex.

If the path is globally convex, then I agree with you: We can iterate and increment our way to the best solution. And patents only get in the way of that hill climbing.

But if the path is only locally convex, then there will be the need for *someone* to leave the current hill that everyone is climbing, and strike out on their own, to some forelorn, empty, risky new hill. And that has a cost associated with it. So given that one is going to pay that cost, one wants to be able to claim that new territory, in the hopes that once you climb the hill it will be even higher than what came before. If someone else wanted to, they could have also started climbing that hill. But because they stuck to the other hill, they should have no expectation of being able to claim property on the new hill that you/one started climbing.

Really, I think that analogy explains the key differences in our understandings. If there is only one hill, and everyone is on that same hill, then there is a commons aspect to that global convexity that patents would only get in the way of — because everyone has to be on that one and only one hill. But if there are multiple hills, and risk involved with leaving one hill to climb another, then there isn’t a commons, and patents don’t get in the way of global optimality the same way that they would in a globally convex landscape. You just have to find some other hill to climb, find your own local maximum.

And I don’t care if we’re talking about software, hardware, glass, petrochemicals, or whatever else is inventable and patentable. I see more evidence that the real world is locally convex, rather than globally convex. So patents don’t concern me as much.

Like

..and I think what you’re seeing today is that the majority of software companies think that there is only a single, global hill: The Web. And that iterating is the only way to climb that single hill.

So I can see why you and many others see things the way that you do. But I neither believe that the web is the be-all and end-all of computer science and technology, nor do I think that the landscape has but a single hill.

And I’m sure that heavily informs my position.

Like

Earlier today you twittered/wondered whether the Google acquisition of Motorola might be due to patent issues. Rather than debate that with you right now, I think an interesting follow-on question to ask would be, assuming that patents were the primary driver for this move, whether the benefits in question were software patents.. or hardware patents.

Given that Motorola is a hardware maker, and Google already has its own software and will probably not be using Motorola’s, it seems just as likely that hardware was the issue.

And if hardware was the issue, then that isn’t something that you object to anyway, right? I mean, you’re just against software patents. If someone comes up with better tangible engineering, better hardware patents, then that’s something you’re fine with being able to protect, and not get in the way of innovation, right?

Like

My understanding is that the patents being exerted in the mobile space have at least as much to do with software — and with interfaces in particular — as they have to do with hardware.

And let me make my objection clear: I don’t like patents in areas where they are a net negative for innovation. For software, I feel that my sampling of the space is sufficient to reach a judgment that software patents are a strong net negative. I’m not a hardware guy, so I’m not making a judgment call there.

Like

Do you really mean a net negative for innovation? Or a net negative for competition? Is innovation alive, but fewer players (at least for a limited time) entering certain markets than you’d like to see? Is that what you’re measuring when you measure the net?

Stated more simply: How do you measure innovation?

Like

Comments are closed.