A few points. First, the point of the paper, as many have correctly noted, is to begin moving the discussion of the desirability of IP away from the model of monopoly profit versus incentives to create and towards and understanding of its effects on industry structure.

I wanted to emphasize that the choice of rules and exceptions in IP will centralize and/or decentralize decision-making. For example, as I say in the paper, from a Hayekian viewpoint the derivative work right probably gives the copyright owner power to decide the future of a given work.

In response to Richard Bennett and a few others, the use of Hayek is of course in good faith. I believe the that centralization of decision-making in industries like the software and, say, music industries is aided by the accumulation of large stores of intellectual property.

However, on re-reading my paper, I also think that it would have been useful to do more to emphasize some of the ways that IP rights can decentralize decision-making. For example, patents on drugs make it easier to decide to research a drug even if the NIH thinks its a bad idea. That point deserves more exposure.

Finally, to Noel Le, there are plenty of current efforts to both weaken or strengthen IP protection. For example, a treaty under consideration by the WIPO proposes to create a new 50-year IP right in broadcast signals, for what reasons I have not completely understood.

Anyhow thanks to those who read the paper,

Tim Wu

PS I had understood from the preponderance of articles I read and computer scientists I talked to that Japan’s Fifth Generation project hadn’t been a great success. However, I suppose dissent from that position is to be expected.

You really can’t draw a bright line between “software” patents and “hardware” patents.

This is true, but can you draw a bright line between patents on software and patents on pure mathematics? No doubt some folks who frequent this board would have no problem with pure mathematics patents, but courts have not generally agreed with this position. Yet there are a great many patents which boil down to pure mathematical algorithm once you strip away details of implementation or embodiment. Which could lead once to believe that patentability has more to do with embodiment than with the underlying core algorithm, method or process.

It’s a tough, somewhat philosophical issue, and I don’t pretend to have anything like a bright-line answer. I am convinced, however, that there are far too many bad patents being awarded by the PTO, and many of these bad patents cover software-related “advances.”

You really can’t draw a bright line between “software” patents and “hardware” patents.

You really can’t draw a bright line between “software” patents and “hardware” patents.

“aspects of technology that involve concrete laws of physics rather than abstract laws of thought.”

There’s the difference. A method or procedure should be patentable (according to Knuth) only as applied to a physical system, not as a self-contained piece of mathematics.

Anything we can do with software we can do with hardware, and more easily now than ever before.

This is probably true, but if something can be implemented in software or hardware, then it’s fundamentally nothing more than an algorithm. Should algorithms be patentable? Donald Knuth doesn’t think so:

http://lpf.ai.mit.edu/Patents/knuth-to-pto.txt

The very notion of a “software patent” is ridiculous. Anything we can do with software we can do with hardware, and more easily now than ever before. With such things as FPGAs and ASICs, hardware design resembles software design to such a degree that can’t tell them apart. We even have hardware design languages, such as System C and Concurrent Pascal, derived from pure software languages.

The issue is the lack of incentives to do the job properly, as the planning bureaucracy would ultimately be run by people, and we all have self-interested motives.

The patent system actually facilitates the flow of information about inventions, as the bargain you make when you apply for a patent is full disclosure of your methods in return for a limited-time monopoly. You have every incentive to develop your idea in a rational and optimal way because your profits depend on that, and you have access to capital.

So it seems to me that Wu emphasizes an unimportant aspect of centralization-by-patent and ignores the counteracting influences.

* He (Professor Wu) isn’t claiming … that this is a knock-out argument (decentralization) against (IP) … he’s pointing out that this is one of the factors that needs to be weighed when evaluating IP policies.* If current IP policies perpetuated the dominance of big tech firms, prevented the viability and entry of new firms, and if the industry thus became stagnant with little innovating activity, Professor Wu’s theory may be more important, or at least immediately relevant to tech policy, but this isn’t the case.

***Noel, centralization and decentralization are on a spectrum. *** Despite our disagreement over exactly “what is (de)centralization” lets make sure we call the industry more/less centralized, rather than either centralized or decentralized. Personally, I don’t like where this discussion is going. We’re heading full speed into a debate on “what is the optimal level of (de)centralization” which will just draw us into a discussion on “whats the optimal level of innovation.” Uhh. As long as there is a non-concentrated software industry (see below) that’s healthy and productive, abstract debates that don’t consider the reality in front of us, are senseless (you don’t believe that the tech industry is doing fine, look at the stock market).

Stronger IP protections push things in the direction of central planning, in the sense that more economic decisions are made by large, hierarchical bureaucracies like Microsoft and the DVD CCA, rather than by individual entrepreneurs. Obviously, it’s a lot less central planning than having, say, the entire software industry run by a single government-protected monopoly. But on the margin, more expansive (IP) rules mean more central planning. In contrast, the software industry is very NON-CONCENTRATED. That means a significant amount of innovation, and patenting, occurs with small companies; as illustrated by the amount of turn-over in leading firms, and the rapid rate of entry of new industry players. (see some work from Professor Merges http://weblog.ipcentral.info/archives/2006/09/merges_on_paten.html). And yes, companies like MSFT and IBM are big, and they hold many patents, but there’s a significant software industry beyond them that holds important IP assets, and commands influence in important markets.

… liberal licensing mitigates this effect to a large extent. However, licensing still requires some foresight on the part of the licensor. The licensor might become wedded to a particular licensing model that works well with certain uses of a patent but not others. And some companies, for whatever reason, might opt not to license at all. I keep bringing up the concept of substitutions in my discourse with you. If a company has a patent, even if its “weded to a particular licensing mode,” competitors can STILL invent around, or licensing essential technologies from other companies. If these competitors are actually innovative, they might invent something new or find a new approach. These are activities that we consider “market activities.”

Richard, I don’t read him as taking any particular position on the desirability of intellectual property in general, for or against. Actually, Professor Wu advocates, as you do, the complete elimination of software patents. I hope he takes this idea to some patent reform testimony, as its neither a serious nor viable policy option, and his argument for it is even worse.