A century ago the U.S. Patent & Trademark Office grew so wary of claims for mythical perpetual-motion machines that it refused to review any patent application without a working model. Thus did it establish the principle of ignoring claims that defied the First Law of Thermodynamics or any other physical law.
Would that it were so nowadays. The overworked, underpaid and often inexpert clerks at the patent office let a lot of crazy ideas--and some painfully obvious ones--get past them. Patent No. 6,025,810, issued in February, is for an antenna that sends signals faster than the speed of light--impossible, if you believe the decades of science based on Einstein's theory of relativity.
Patent 5,533,051, issued in 1996, covers a technique that purports to compress any data set by at least one bit without loss of information--a process that, if done recursively, could shrink the Encyclopaedia Britannica to a single word from which the original could be flawlessly reconstructed. The very idea is preposterous. Yet the patent office agonized for three years over the application--and in the end approved it.
"I don't really mind patents on impossible things, which I just laugh about," says Jean-loup Gailly, a French computer scientist who has written several well-known compression programs. "The real problem comes when patents are granted for processes that were already known at the time."
The U.S. Patent Office is withering under the fiercest criticism it has endured in its 200 years of existence. It has simply been overwhelmed by the rapid pace of change on the Internet and an explosion of patent applications by companies seeking to string barbed wire around innovation. One patent for a Compton's CD-ROM covered any method for retrieving data from a disk; another company's patent protected the very notion of letting buyers use the Internet to bid on things like airline tickets. Gailly is right: When patents are issued for sweeping claims or obvious applications, they don't foster innovation, they crush it.
Given that patent examiners start with salaries as low as $28,000, we can scarcely expect a lot of Einsteins to apply for the job. Their workload has swelled in recent years, turnover is high and greenhorns end up deciding issues that would stump seasoned experts.
The weaknesses are especially evident in the arena of patents for business models. Software business-process patents go back to 1982, when Merrill Lynch won one for the cash management account. The field got an even bigger boost two years ago when Signature Financial Group overturned a judge's order invalidating its patent on a way to pool mutual fund assets for better efficiency and tax treatment.
The web has taken these process patents to their natural extreme. The most notorious example: the "one-click" patent, No. 5,960,411, awarded to Amazon.com in September. Simple and obvious, it lets registered customers buy something, charge it and have it shipped to them, all in a single mouse click. Amazon sued Barnes & Noble for infringement and won an injunction in early December. The rival had to add a second "verification" click to keep selling during the Christmas season.
Amazon won yet another method patent in February for an equally obvious item: Patent No. 6,029,141 protects a system that lets websites collect a cut of sales when they refer a customer to another site. What's next--a toll for the first guy to design a way to click on an ad to order a freebie?
Less infamous awards have been no less dubious. One patent recently was granted for a "Y2K" fix that was old hat to all but the examiner who approved it; it has since been recalled for a second look. Other bogus patents live on in limbo, unchallenged and unasserted, impressing no one except, perhaps, investors who hope for a fat portfolio of intellectual property.
The patent office has simply been overwhelmed by change.
The patent office's gaffes stem in part from cruel mathematics. In just eight years the number of patent applications is up 38%, to 290,000 a year. That's owing to companies swamping the office with applications for multiple narrow patents to gain a competitive edge.
The patent bureaucracy's budget hasn't kept pace with the exploding workload. From fiscal 1998 to 1999, patent applications rose 11%, but the office endured a 16% decline in how much fee revenue it was allowed to spend on operations. The funding fell to $673 million, and the U.S. government simply siphoned off the other $214 million collected in patent fees.
No wonder the office says that it lacks the funds to hire more examiners. "We would have liked to hire another 750 examiners this year and next, but budget constraints will probably preclude that," says Brigid Quinn, the patent office's spokeswoman. A bill introduced in Congress last year would empower the patent office to retain all fees for running the system, but its chances are uncertain.
Last year its 3,200 examiners awarded 161,000 patents, about 50 apiece. Even in software, where the issues can be far more complex, some 375 examiners pored over 29,000 applications and awarded 13,900 new patents. One software reviewer approved 200 last year, four a week, a frequency that defies doing exhaustive research of "prior art" to weigh a claim's uniqueness.
Q. Todd Dickinson, who heads the patent office, acknowledged the problems recently when he announced that web-oriented patents would have to clear both a regular examiner and a new layer of scrutiny by a more senior official. After that, the number of spot checks for quality control would be stepped up. But even those fixes fall short.
In software, patent examiners often have far less experience than those in other areas. "In mechanical arts, a lot of examiners have been around for decades and know the technology like the backs of their hands. Chemistry has great databases for prior art," says Gregory Aharonian, a patent expert who helps clients invalidate bogus claims.
Worse, the booming software business has sucked up all the programming talent. "If you know squat about computers, you can make a lot more than the patent office pays," he says. A high-tech veteran in the patent office can earn only as much as $110,000--that's starting pay for some rookie programmers.
The European Patent Office, by contrast, gives its employees privileged tax status and a higher wage scale. It is derived from Switzerland's patent office--the most generous on the Continent (it's where Einstein got his start; he surely would never have approved that faster-than-light antenna).The U.S. patent bureaucracy's working conditions aren't up to snuff, either. Computer systems have been maddeningly slow, so much so that examiners last year had to work in shifts to lessen the burden on their database search machines.
Finding remedies may be as befuddling as the patent examiner's job itself. The best reform may be to introduce the adversarial setup of the court system--which already decides patent disputes at an exorbitant cost--at the front end of the process, before approval. That way rivals could pay for a thorough prior-art search and block nettlesome patents.
Another reform is offered by none other than Jeffrey Bezos, the Amazon founder and master of the all-embracing software claim. He suggests limiting software patents to three to five years, enough time to grant a head start without stifling competition.
That was the driving idea behind the patent system in the first place. There's one snag: The patent office rejected the Bezos plan because it would violate international agreements that all patents must run 20 years.