In the past 40 years, many trading-related patents were granted. In retrospect, they probably should not have been; though a select few were based on reason and sound application of the law. In an effort to expand our understanding of the current state of technology-based patents, let’s explore two that I helped develop.
In 1988, I worked for a company called Travelsoft. The owner of that company co-invented patent No. 4,931,932 (“Computerized System with Means to Automatically Clear and Sell Wait-Listed Customer Reservations”). We then co-founded Promise Land Technologies. At this time, I was granted patent No. 5,241,620 (“Embedding Neural Networks into Spreadsheet Applications”).
Today, both of these patents are expired, but offer two good examples of how interpretation of patent rules has changed today. In 1988-1993, many patent lawyers advocated that software patents are process patents: If the computer was not used for the process before, and you figured out how to do it on a computer, then what you established could be patented. Now, many of these patents have been overturned.
One famous patent decision was the Alice Decision, in which the Supreme Court nullified a process patent. This precedent has allowed for the overturning of numerous patents since then. On Sept. 21, 2014, The Wall Street Journal reported: “The high court’s unanimous opinion, written by Justice Clarence Thomas, said that for a software patent to be valid, it must describe more than an old idea, such as escrow, simply applied to a computer.”
Nevertheless, too many bad patents have been granted by the examiners, and early patents were granted under the false belief that a computer performing a public-domain task was patentable. The court’s view is now correct, but it will take a long time to fix years of bad policy.
Tale of two patents
The TravelSoft patent, which was licensed under the name FlightClear to American Express during the late 1980s, is an example of a patent that may not be granted today. It required domain expertise in booking flights, but not advanced thought processes, which may include natural language processing and face recognition. Advanced thought processes include certain data transformations or things that a computer can’t inherently do on its own, but a human can.
In the case of FlightClear, the patent booked a flight within a given time frame from various airlines. It then searched for a cheaper fare for that flight. If one was found, the computer would cancel the old flight and book the cheaper one. This was done over dial-up connections for airline reservations. In these days, this was a novel concept: passengers almost always booked through travel agents, who were paid 10% of the fare. The FlightClear process mimicked an agent using his or her terminal and searching for cheaper fares, over and over again.
This patent cited many different travel-related patents, which were granted then because the innovators had to invent new computer technology. Many of these airline reservation patents were good patents because they required new concepts, algorithms and technology to make them work; they weren’t simply re-purposing existing processes.
In the case of FlightClear, there was nothing unique about the way the technology searched. It just worked harder than a single agent could. A staff of agents working 24/7 could have done the same thing. Clearly, there was value in the product from a business resources standpoint, but that alone does not make it patentable.
Now consider the neural network spreadsheet patent, granted in 1993. From
“The method comprises providing an application program in which information is stored in rows and columns or a database containing fields and records and embedding a neural network in the application program or database using the stored information. The embedding step includes allocating unused memory in the application program and creating both a neural network engine and an application interface structure from the unused memory. Once the neural network engine and an application interface structure have been created, the neural network may be trained using variable numerical and symbolic data stored within the application program.”