1. Artificial intelligence – what is it?
Even the question of what artificial intelligence (AI) means will give many people quite a headache. In today’s hype, anything remotely to do with non-manual data entry is thrown together, advertised and sold under the name AI. In reality, many of the currently existing products are probably “only” more or less useful algorithms, or software programmed by human beings, which learns from data and commands entered and later uses the lessons learned within its narrow programming boundaries. The data does not necessarily have to be input by a person but can also be entered “automatically”, for example from a camera programmed to recognise images or text. However, this is not an autonomous – intelligent – action of the computer, the machine or the robot.
The European Patent Office (EPO) defines AI as “the ability of systems to learn”, although that definition does not seem particularly helpful and is extremely vague.
Using that definition, however, in order to possess actual AI a machine would at least have to perform an evaluation based on the data input to then make an autonomous decision.
2. Who is required to deal with artificial intelligence and related IP rights?
Anyone working in the service, manufacturing or distribution sectors should (and must) address the issue of AI, that is to say, virtually all legal entities and individuals who engage in contact with third parties in a business context. AI concerns nearly every private individual as well.
In fact, the ongoing development of computer and internet-based technologies will affect all areas of commercial and private activities, be it automated factories, autonomous driving, banking that is increasingly internet-based, or chat-bots answering questions about a diverse range of topics immediately and without any time lag.
3. How can artificial intelligence be protected?
All Patent Offices worldwide are currently seeking to define what AI actually is and how AI inventions can be properly protected in order to promote investment in AI on the one hand whilst not obstructing innovation by blocking markets too fast on the other.
In 2018 the EPO hosted numerous events and conferences exclusively dedicated to the patentability of developments in, for example, machine learning, block chain technologies or artificial intelligence as such.
According to the Guidelines for Examination the EPO considers that the technologies in question are presently based on computational models and algorithms for classification, clustering, regression and dimensionality reduction, such as neural networks, genetic algorithms, support vector machines, k-means, kernel regression and discriminant analysis and that such computational models and algorithms are per se of an abstract mathematical nature. Hence, the guidance provided in G-II, 3.3 generally also applies to such computational models and algorithms. This is an express indication that these computational models and mathematical methods as such are still excluded from patentability but can normally be patented by applying the principles of patentable computer implemented inventions, for example by their formulation as a process, use or device.
4. How can artificial intelligence be commercialised?
Depending on the field of operation and whether it is basic technology or specific areas of application that are concerned, it stands to reason that competitors will also be interested in commercialisation or are even required to implement the particular technology to take part in the market.
Therefore, the already very broad field of AI is likely to lead to inventions that are at least situated in the conflict area between individual and standard essential use.
For those acquiring IP rights relating to AI, it will become increasingly important to also develop knowhow for licensing purposes. Naturally, this also concerns all those that incorporate AI in their products and therefore may have to rely on the IP rights of others.
Notably, new players will appear on the market who did not previously have a presence or significant presence in the field of patents. For example, a trend that is already visible is that in the area of blockchain technologies, an increasing number of –companies (including well-established ones) from the banking sector enter the scene as patent applicants and are also apparently making considerable investments in businesses that commercialise the relevant patents, for example patents for transmission technologies, IoT, etc.
5. Can developments achieved by artificial intelligence be protected?
Another highly important question will be whether developments accomplished through AI are in fact patentable. German law at least consistently refers to achievements made by human, not artificial, intelligence. Applied strictly, this principle will probably preclude things created by AI from patentability in many cases, which in turn could stunt or even stop incentives to invest in these technologies.
In contexts of copyright law, software or databases, this is probably causing problems already but should still be some way off for technical inventions (patents). However, there can be no denying that these issues will become relevant sooner or later. Companies that are active in these fields and are using or planning to use AI – in any form whatsoever – in their R&D departments are well advised to prepare early and build up knowhow, including on the legal side. Once an AI “invention” has been conceived of, it may be too late to obtain solid protection.