The digitalization of information and the implementation of computer-based processes in society has fundamentally changed how humans interact with the world and one another. Now, almost everyone in the US has access to an enormous pool of information in some form or another. While this is touted as the necessary step towards information equality, it is not without its problems. In a society where we demand efficient and accurate results, what happens when the pool of information becomes too vast for humans to navigate? How do we assure that a user is getting the relevant information they are seeking? The answer is simple, use more computers.
Algorithms are humanities answer to the information problem. Algorithms not only navigate vast swaths of information in a way that humans could never hope to achieve, but they can be taught to learn and grow while doing so. Machine learning is the process by which computers use large amounts of data to “learn” how to recognize patterns and make decisions without having to be programmed to do so.[i] Machine learning is used everywhere, from Google’s web search algorithm to your phones autocorrect function. When you click the second link in a Google search result, the algorithm notes the “input” and “learns” which result you were looking for so the next person who inputs the same search gets that option as the first choice.[ii] Simply put, a computer using machine learning can “teach” itself how to more accurately accomplish a goal by building complex algorithms, all to achieve what humans could never hope to accomplish on our own.
It comes as no surprise then that the rapid emergence of machine learning has put various long-standing, human run practices on the chopping block. In the world of patents, digitalization has not only expanded the subject matter by which patents can be granted, but it has also expanded the scope of prior art to be considered by the USPTO. As time moves forward, the pool of prior art an examiner at the USPTO has to navigate through continues to grow. Therefore, the USPTO has been implementing a new wave of AI tools for their examiners to utilize while conducting various parts of the patent process:
“USPTO has established an advanced analytics program that combines big data/ big data reservoir (BDR), machine learning, and artificial intelligence (AI) to enhance understanding of USPTO policies, processes, and workflows. AI is basically defined as cognitive assistance using feedback from human users where the AI is capable machine learning (deep/neural) to provide the most useful and relevant information to determine patentability by an examiner during prosecution.”[iii]
While this is an important step forward in bringing the US patent system up to date with the digital world, is there a future where all patent searching is automated by machines? Is that something we ought to strive for, or is the job of a patent examiner something we should protect from automation?
Erich Spangenberg, founder of IPwe, believes that the future of the patent system is going to be found in the blockchain. Blockchain technology “is an incorruptible digital ledger of economic transactions that can be programmed to record not just financial transactions but virtually everything of value.”[iv] Using this technology, IPwe is attempting to create a blockchain-based registry to tackle the lack of access to good information on patent transactions. And it is using machine-learning algorithms to better evaluate patent validity and worth.[v] This ambitious goal is not only lucrative to IPwe, but also to the entire US economy and patent industry. In an economy like the US that centers around intellectual property and the right to exclude others, clarity in the patent system translates to real world implications.
The World Intellectual Property Organization has estimated that roughly $180 billion in annual patent value derived from 2 percent of all patents.[vi] Spangenberg believes that by implementing machine learning and blockchain technology into the USPTO could result in an explosive growth in patent value. As previously mentioned, the growing digitalization of prior art has created an enormous issue for the USPTO, how do you assure that the patents that are granted are “quality patents”? Mark Schankerman, a professor at the London School of Economics, believes that as much as 75 to 80 percent of patents are “junk”.[vii] If machine learning, coupled with the blockchain, could be used to filter out “junk” patents, should the USPTO be turning its attention to automation rather than education of its patent examiners?
While the USPTO has the Office of Process Improvement, a department that “provides the methods, resources, and training to optimize United States Patent and Trademark Office business processes and strengthen the agency’s strategic goals of timeliness and quality”, one could ask whether that is sufficient.[viii] It is my opinion that the future of the patent system will turn on how committed the USPTO is to automation of the prosecution process. Unfortunately, the future is not bright as the new administration of the USPTO has “put on hold” an ambitious partnership between the USPTO and artificial intelligence start-up AI patents to implement its prior art search technology. The latest word from the USPTO was that “[t]he Office continues to investigate uses of emerging technology for various applications.”[ix]
It is unclear what the future of patent prosecution will look like in the US, but if recent history has taught us anything it is that the patent system can undergo rapid change and come out the other end for the better, it is now only a matter of time.
[iv] Don & Alex Tapscott, authors Blockchain Revolution (2016)
Eric Loverro is a J.D. candidate, 2019, at NYU School of Law.