Artificial Intelligence, Machine Learning And The Patent Process: A More Cost-Effective Tomorrow?
The notion of “time is money” was born in the factories of the Industrial Revolution. Since then, people have always strived towards greater efficiency, trying to achieve more in less time. As Artificial Intelligence (“AI”) and machine learning (“ML”) become more conventional, stakeholders to the global intellectual property (“IP”) system are gradually adopting these tools. WIPO Director General, Francis Gurry, remarked that, “AI is a new digital frontier that will have a profound impact on the world, transforming the way we live and work.”
This article explores the effects of incorporation of AI and ML into the IP ecosystem, with special focus on reduction of costs in the patent process.
What are AI and ML?
While the term “AI” has been around since 1956, the technology picked up pace only around 1993-2011, when AI started to become data driven and computers became more powerful. Post 2012, more data, increased connectedness and growing computer capabilities led to breakthroughs. In 2020, increasing numbers of people use AI every day in the form of music recommender apps, Google maps, and many other services that run on AI.
Machine learning is the dominant AI technique. It is found in 40% of all AI-related patents studied, and it grew at an average rate of 28% every year from 2013-2016. Deep learning and neural networks are techniques within machine learning that are currently revolutionizing AI. These are central to processes like automatic translation.
AI and ML in the IP System
AI and ML are making a slow but steady entry into various aspects of the IP system, such as through the increasingly popular concept of smart contracts. A smart contract is a computer coded contract which makes use of blockchain technology and is self-executing. It can be used for the assignment of IP, which would occur automatically upon receipt of the consideration amount. By removing the need for human intervention post execution of the contract, administrative costs are reduced, and hence efficiency is improved. 
The overarching goal is to reduce the amount of manual labour involved at various stages of the IP life cycle. IBM is now offering an IP advisor named Watson that uses AI for fast patent processing, better insights, and analytics. IBM’s Watson is capable of analysing massive amounts of unstructured data input in several languages and from varied sources.
AI and ML are also finding use in detection and consequent prevention of infringements. For instance, Red Points, a brand protection company based in New York and Barcelona, developed a client-centric platform that can identify unauthorised sellers of a company’s products and track pirated content such as video and films.
How Can AI and ML Reduce Costs in the Patent Process?
Patent software companies are now relying on AI to produce advanced patent tools. They are adopting a hybrid approach to give conventional patent tools an AI upgrade.  AI and ML have the following applications in the patent process.
- Prior art search
AI-based tools can review vast amounts of data in short amounts of time and hence are apt for prior art search, which requires the examination of millions of documents in various databases. A survey conducted on 60 senior in-house patent lawyers across industries found that prior art search was the most popular choice of the best use for AI in the patenting process. Previously, prior art search could take between two and four days, but AI can do the job in just a few hours, giving the searcher a head start when it comes to manual analysis.
More advanced solutions include natural language search tools that allow users to input natural language terms. The AI tool understands these terms and uses them to retrieve the requested documents and give suitable search results. Apart from prior art searches, such tools can be used to identify potential patent infringements.
AI is already in use in some countries due to its time-saving benefits. The European Patent Office’s master database contains 108 million documents. It processes about 30 million search reports from patent authorities and 79 million full-text records in various languages. Staff at the EPO say its technology can scan all of these with ease. In 2018, Austria’s patent office conducted trials with several commercial providers of AI to help with the classification of patent filings. The Norwegian Industrial Property Office has begun using an AI-based tool to conduct trademark image searches.
AI and ML also analyse the behaviour of the searcher, compare and include competitor and market information with patent data and thus, incorporate market and business information in the analysis. 
Prior art search, which would take teams of searchers a few days to complete, can now be done in just a few hours with a fraction of the manual labour, thus saving tremendously on costs.
- Language translation
One major hurdle typically encountered by patent applicants during the patent filing process is the language barrier. It not only restricts them while filing for a patent in different jurisdictions as the procedural instructions are unintelligible, but also comes in the way of understanding what has already been patented in that jurisdiction. Applicants then turn to professional translation service providers, which increases costs involved in obtaining the patent.
Semantic-based algorithms focus on performing document retrieval by identifying the meaning of the words used in accordance with an appropriate context. They can hence help a searcher to find the most relevant documents irrespective of the terminology or language used.
The use of AI to translate documents filed in different languages can hence make a country’s patent filing processes as well as document database more widely accessible and understandable. Translation costs are a significant component of the process of filing a patent in different jurisdictions, and could be dramatically reduced with the usage of ML. This could pave the way for larger numbers of patent applications, thus contributing to economic development.
- Assessment of patentability
Patent lawyers can use the new wave of technology to improve their own efficiency. AI can process large volumes of data and identify patterns that would be impossible or too time-consuming for humans to detect. For example, it can analyse descriptions of similar products and compare them using image recognition.
AI can detect immediately whether or not applications meet criteria such as novelty, inventive step and industrial application to make them eligible for patents. AI can also pre-classify incoming applications to assign the file to the correct unit and automate the annotation of patent literature. This makes the process of assessing whether an invention is patentable faster and easier.
- Benefits for inventors and patent applicants
Patent attorneys and intellectual property offices across the world are not the only organisations turning to AI. Companies and innovators can use it to keep track of their own inventions and intellectual property.
This technology can create value for client companies by unearthing patents that were forgotten because they have moved out of a particular product line. The companies can then generate revenue by selling these patents.
- Detection and prevention of infringements
AI is also becoming an essential tool for patent attorneys working to protect their clients’ intellectual property. For example, US-based legal analytics start-up DataNovo uses big data to conduct prior art searches and to identify patent infringements. Such technology helps companies engaged in designing new products to avoid infringing someone else’s IP. After conducting a rapid assessment of similar existing products, they can either tweak their own design accordingly or obtain a licence to avoid causing a breach of IP rights.
- Prediction of outcome of litigation
AI can even support patent actions in court by predicting the result of a case. For instance, on the basis of detailed historical data of cases, AI can determine situations in which a particular judge has decided favourably. Companies and patent attorneys can use this information to build their litigation strategy and decide whether to settle a certain case, or to fight it out in court. This aspect is controversial as it detracts from the human discretion which plays a vital role in judicial decisions and could result in judges losing value.
Based on prior cases, it is even possible to analyse the way an adversary conducts litigation and gives an idea as to whether they tend to settle at a certain point or fight to the bitter end. However, this would require feeding AI the data of millions of court cases, all of which must be formatted in the exact way to make them accessible to the relevant AI tools.
Better prior art searches curated by AI can also help firms plan litigation strategy better by knowing which patents can be challenged.
Challenges and the Way Ahead
Several concerns arise with respect to the inclusion of AI and ML into the IP system. What ownership and regulatory models should apply to data, which is essential to the development of AI? Is an invention generated by AI eligible for IP protection and, if so, who owns those rights? Free access to data can provide highly personalized experiences, but how can this be balanced with personal privacy concerns?  Questions like these must be addressed at a global level.
All jurisdictions currently require a human individual to be registered as the inventor in a patent application. The doubt arises when a discovery or invention is created entirely or in part by AI: who is the “inventor”? This could lead to disputes over whom the credit should go to in case of an AI-aided discovery. Issues of inventorship, obviousness and plausibility would then arise while patenting AI-discovered inventions.  Once AI and ML begin to play a larger role in identifying solutions and inventions, laws surrounding the disclosure of the role of AI in an invention may need to change.
Patent Offices today evaluate whether patent applications pass the test of “inventive step” based on the limitations of human imagination. In the case of machine-invented products and processes, patent offices may choose to raise this bar for inventive step, thus making it more difficult to obtain a patent. They may even find it difficult to understand and accept how a machine could come up with a particular idea.  The economic and societal impact of non-human inventions would receive sharp focus in this case.
One can agree with Andrew NG, CEO of Landing AI and deeplearning.ai, who said that, “AI is the new electricity. I can hardly imagine an industry which is not going to be transformed by AI.” AI and ML are making services and processes easier across the globe by reducing human error and human effort. It is time for the stakeholders of the IP system to embrace this technology at a measured pace.
It is evident that the use of AI and ML can dramatically reduce costs incurred at various points of the patent process, making it cheaper and easier to apply for patents and to process applications. However, while the benefits are alluring, one must be wary of the unaddressed concerns that accompany the incorporation of this technology. Yes, AI and ML can reduce costs. But at what cost?
- The Story of AI in Patents, World Intellectual Property Organisation (Accessed 27 May 2020) https://www.wipo.int/tech_trends/en/artificial_intelligence/story.html.
- Sarah Neil, Mike Jennings and Nikesh Patel, The benefits of Artificial Intelligence in the field of IP, Lexology (Accessed 27 May 2020) https://www.lexology.com/library/detail.aspx?g=43b8660b-8e61-42ff-9d2c-d5ed5d9f730c.
- Amit Goel, Prashant Singhal and Raj Kishore, Why AI is crucial for patent searching and mining, Lexology (Accessed 27 May 2020) https://www.lexology.com/library/detail.aspx?g=ece7e6a5-7a3c-499d-9c00-4cb37f031e78.
- Sarah Murray, AI drives down cost and drudgery of routine patent work, Financial Times (Accessed 27 May 2020) https://www.ft.com/content/80ae35aa-7711-11e9-b0ec-7dff87b9a4a2.
- Patent Strategy, Machine yearning: AI and patents, ManagingIP (Accessed 27 May 2020) https://www.managingip.com/pdfsmip/Machine-yearning-AI-and-patents.pdf
About the author :
Tanvi Chaturvedi is a fifth-year student of the B.A.LL.B (Hons.) course at Symbiosis Law School, Pune. An environmentalist at heart, she finds music and art meditative.