AI patents on a growth curve

Inventions involving CII, and AI in particular, challenge the current IP system in several ways. The issues centre on patentability, ‘inventorship’, and ownership. Evidence of this can be found in the World Economic Forum 2018 white paper; Hu 2019;³ and WIPO’s decision in 2019 to launch a public consultation on AI and intellectual property.⁴

Most current patent laws and conventions require a patent application to designate a natural person as the inventor, thereby excluding AI. In a recent case, both the EPO and the UK IPO have rejected two patent applications, which explicitly declare an AI system as the inventor (see AI Inventorship Project in 2019 WIPO Magazine⁵ and EPO and UKIPO Refuse AI-Invented Patent Applications (article in IP Watchdog 2020⁶).

AI enhances the ability to collect, categorize, and analyse large datasets, to simulate complex processes, and/or to predict their outcomes. Thus, the classical topics of biotechnological patents offer attractive applications for AI: compound design/discovery, diagnostics, prognostics, or control of complex biological processes.

Biology was among the first areas of application for AI, and it continues to be an important topic (see oliveira7, Goh8, Shah9(2019) and Fuji (2017)10). A recent article on drug discovery proposes five levels of AI involvement, ranging from analytical assistance to entirely autonomous activity without human input.

In order to identify the patents disclosing AI applications in biotechnology, we created the intersection of two independent and comprehensive patent collections; one relating to AI/machine learning and the other to biotechnology. Each of them comprises more than 300,000 active patent families.

The overlap yielded 3,136 patent families, accounting for less than 1% of each parent technology field, and including 52 patents from Swiss inventors. Despite this small number, the contribution of Swiss inventors in this set is larger in proportion (1.7%) than in biotechnology (1.2%) or in AI (0.5%).

With regard to patent quality, an impressive 40% of all AI patents in biotechnology worldwide qualify as world-class. This is substantially more than in biotechnology or AI alone, and close to the high values observed for patents invented in Switzerland (Figure 2), which have been presented in the Swiss Biotech Report 2018.

Notably, the majority of the patents of Swiss origin in all three sets are in fact international inventions, i.e. they designate inventors from Switzerland as well as from other countries. For 2019, the proportion of international inventions within the Swiss-invented portfolios are 77% in biotechnology, 62% in AI, and 74% for AI in biotechnology.

The overlap found between AI patents and biotechnology patents seems surprisingly small. However, this selection is restricted to patents in which AI represents a core feature of the invention, as illustrated by the examples in Table 1. These patents correspond to levels 4 and 5 according to the classification proposed by Green.¹¹

Beyond these AI-centred biotech inventions, AI may have served as an assisting tool in many more biotechnology invention. Accordingly, if the patents had not prominently disclosed the use of AI, the classifications and keyword concepts applied would have failed to capture them.

In conclusion, AI patents with biotechnology overlap for which AI is a core component are just the visible tip of a much larger iceberg of biotechnology-AI patents!

The article also features in the Swiss Biotech Report 2020.