Rise of the Machines: The Ever-Evolving Impact of Artificial Intelligence on IP Law
Oct 24th, 2017 by Michael Hinrichsen | News | Recent News & Articles |
Long heralded as a technology just around the corner, artificial intelligence is finally making an impact in today’s world. As the technology progresses and becomes more widely adopted, routine mental tasks will become increasingly automated, so much so that nearly 50% of the jobs currently performed by humans are predicted to be automated in the future1. In this article, we examine the impact artificial intelligence is having on intellectual property law today, and how this technology will likely affect the field in the future.
Stated simply, artificial intelligence is the ability of machines to perform tasks that would typically be considered to require human intelligence. Many strategies have been devised to reach the goal of automation, but what has driven most of the recent progress in the field has been advances in a particular kind of strategy called machine learning. With machine learning, algorithms “learn” by examining huge datasets and extracting patterns or trends in data that are correlated with a certain result. This approach is powerful for automating tasks that are formulaic and repetitive, but hard to reduce into a strict set of universal rules. Machine learning is also a powerful aid for extracting insights from large and unorganized data sets that would otherwise be too massive and complex for human comprehension. Within patent law, there are several areas for which artificial intelligence can offer helpful tools, and services that use this technology are already being used to assist attorneys and patent agents today.
Improving database searches
One of the areas of most active research in machine learning is in improving natural language processing (NLP) software programs, which can analyze and understand natural human language. Whereas traditional computer programs interpret language through extremely strict and literal rules, NLP programs can “understand” information that is encoded in the relatively imprecise and ambiguous language we often use on a daily basis. Machine learning algorithms are not required for NLP programs, but their incorporation has greatly increased the performance of NLP software in recent years. While not yet approaching humans in terms of comprehension, the real power of NLP programs lies in their ability to sift through and extract information from enormous document libraries that would otherwise be impenetrable to humans.
NLP programs can improve the efficiency of database searches by both better “understanding” the search queries being posed by human users as well as the documents being searched through. Some legal service companies, such as ROSS intelligence and Innography, further improve database search performance by using a machine learning algorithm to sort and curate hits. While not quite the “robot attorney” the search capabilities are sometimes billed as2, ROSS intelligence has been shown to be superior to search engines that use NLP programs alone, in its ability to identify documents that are most relevant to a given search query3. Programs have also been developed that generate short summaries of input documents, allowing users to more quickly skim through research results. While not revolutionary, improving the efficiency of database searches can significantly reduce the amount of time required to research a topic and increase lawyer productivity.
In addition to reducing the amount of time spent on research, NLP programs can also uncover new insights that would be otherwise buried in massive document libraries. Legal analytics companies such as Lex Machina, Innography, and Ravel Law provide litigators with useful data for creating litigation strategies with the best chance for success. By combing through millions of pages of court documents, these services can track statistics such as the decision-making tendencies of judges, tactical approaches of opposing litigators and companies, the average time to key events, and the litigation history of the underlying patent. Search engines can also tie together data from disparate sources, revealing otherwise hidden connections. Innography offers a particularly interesting search database that connects company information with the litigation history of the underlying patents. The value of these learnings extends beyond patent law, offering companies a useful resource for competitive intelligence, investments, R&D decisions, and identifying licensing and other business opportunities. The influence of these databases on business strategy will continue to grow, as databases become more powerful and widely used for applications outside of patent law.
Patent Application Drafting
In addition to natural language processing programs, much effort has also been spent in developing natural language generating (NLG) programs, or programs capable of generating their own language. On the whole, however, NLG programs are a technology still very much in their early stages, and are capable of generating work that is acceptable for only the most formulaic writing tasks. Applying these tools to less rigidly defined tasks results in nonsensical, and sometimes even humorous results4. Still, in IP law there are some uses for programs that can automate repetitive and formulaic writing tasks. The legal services company, Specifio, has a product that claims to automate the task of writing patent applications, requiring only the patent claims as the initial input. While the current version of Specifio generates applications that still require human proofreading, founders Ian Schick and Kevin Knight estimate their program can complete 90% of the work, significantly reducing the work burden on the patent professional5. Even for writing tasks that will likely always require high levels of human input, NLG programs can still provide a useful resource by offering more intelligent editor programs. While current digital editors can only assist by highlighting spelling or grammar errors, editors with AI abilities would be able to perform more complicated services such as making word choice suggestions and highlighting poorly written portions of text. Given enough data, AI editors could even learn the preferred writing style of users and offer tailored suggestions to fit that style.
Conclusions and Future Implications
In its current state, artificial intelligence is already having an impact on the IP world, primarily by enabling the development of increasingly powerful research tools. Automated writing programs, while still in their infancy, are beginning to have an impact as well, and will become increasingly prevalent in the coming years as the technology progresses. Automation in IP law is serving to increase human efficiency rather than replacing attorneys and agents entirely, allowing these professionals to perform the same tasks in less time and with better results. Future practitioners will be able to increasingly spend their time on tasks that require the greatest creativity, while delegating repetitive and tedious tasks to computers. Increased automation in IP law has downstream implications for the broader society as well. It will be interesting to see whether the increased productivity and efficiency promised by automation will translate into lower patent prosecution costs and increase the number of inventions patented in the future.
1Frey, C. B., and Osborne, M. A. “The future of employment: how susceptible are jobs to Computerisation?” Technological Forecasting and Social Change. 114, 254-280.
2Turner, K. “Meet ‘Ross,’ the newly hired legal robot.” The Wall Street Journal. May 16, 2016.
3Blue Hill Research. ROSS Intelligence: Artificial Intelligence in Legal Research. 2017. Web.
4Ars Technica. “Sunspring | A Sci-Fi Short Film Starring Thomas Middleditch.” Online video clip. Youtube, June 9, 2016. Web.
5“Meet Specifio the AI Start-Up Automating Patent Drafting.” July 28, 2017. Web.
-Michael Hinrichsen and Anthony Sabatelli, PhD, JD
Mike Hinrichsen is a PhD Candidate in the Molecular Biophysics and Biochemistry department at Yale University. His thesis research is focused on using protein design to develop novel methods for imaging proteins and genomic loci in living cells. Prior to attending Yale, Mike graduated with a B. S. in Chemistry from the College of New Jersey.
This article is for informational purposes, is not intended to constitute legal advice, and may be considered advertising under applicable state laws. The opinions expressed in this article are those of the author only and are not necessarily shared by Dilworth IP, its other attorneys, agents, or staff, or its clients.