Patenting Artificial Intelligence
On 30th May 2018, Anna was privileged to attend the first conference on “Patenting Artificial Intelligence” held at the European Patent Office (EPO) in Munich. The conference brought together a range of interesting and enlightening speakers and below is her summary of the presentations.
We were told that Artificial Intelligence (AI) is the fourth Industrial Revolution. The first industrial revolution occurred as a result of innovations in steam energy, coal and transport in the late 18th century; the second occurred as a result of innovations in electricity, oil and mass production in the late 19th century; the third was brought about through innovations in electronics and IT, flight and nuclear energy in the mid twentieth century; and now we are in the mists of the fourth Industrial Revolution – AI.
AI will dramatically change the way that we work and live. However, AI is not a new idea, the concept has been around since the 1950’s and is featured in many science fiction films. However, AI has never previously being possible. The three components required for AI are:
- Powerful algorithms to perform computations;
- Large quantities of data in order to train the algorithm – Large quantities of data are being generated every day, for example, autonomous cars are fitted with a plurality of sensors all collecting data; and
- Computing power – with the advent of cloud computing, large computing capacity and high speeds is now available to all.
AI spans all the different technology sectors, such as vehicles, robotics, 3D printing, materials, genomic diagnostics, medical treatments, engineering, IoT, and blockchain. This is because AI is so diverse. AI can be used to:
- increase productivity – anything which is repetitive and definable can be performed by AI – consequently some jobs may be at risk, but all jobs will be impacted by AI;
- augment existing capabilities – increasing productivity and growth, which is predicted to create more jobs than it will eliminate;
- solve problems; and
- create something new – AI can write poetry and create art.
AI is the number one growing trend in technology industries.
During the first and second industrial revolutions patents predominantly protected hardware technology, during the third industrial revolution patents predominantly protected hardware and software technology. However, during this fourth industrial revolution how will patents protect AI?
The number of AI application filings at the EPO, is increasing year on year, as illustrated below, in all areas of technology:
In order to tackle this broad spectrum of technologies to which AI may be applied, the EPO has created interdisciplinary technical divisions of three examiners to examine each application. In addition, it was stressed that the EPO intends to continue examining AI using the same approach as that currently applied to computer implemented inventions (CII). The EPO aims to maintain a consistent predictable approach to patenting CII and AI. Although we were informed that a new section of the Guidelines for Examination in the European Patent Office regarding “Mathematical Methods, Artificial Intelligence (AI) and Machine learning (ML)” will be going live in October 2018 and coming into force in November 2018, we were told that this section will provide numerous examples consistent with the current approach and will not be changing the EPO approach to examining AI.
When examining AI applications, regardless of the whether the application relates to core AI or using AI, the EPO will use the current two-step approach. It will first determine whether the claim as a whole has a technical character, even when the claims contains a mix of technical and non-technical features. The EPO presenter considered this hurdle to be relatively easy to overcome – for example by claiming a technical means, such as “a computer implemented method”, rather than just “a method”.
The EPO will then determine the second step, whether the claim is novel and inventive over the prior art based on the technical features only. This is when it will be determined if the AI algorithm (mathematical method) contributes to the technical character of the invention. The EPO presenter indicated that for a mathematical method to contribute to the technical character of the invention, the claims should relate to a specific technical implementation, for example the AI algorithm specifically adapted for that implementation or the AI design being motivated by technical consideration of the internal functioning of the computer.
Consequently, when drafting claims relating to AI inventions, it is prudent to:
- provide technical character in the claims (i.e. “a computer implemented method”); and
- provide technical features in the claims which contribute to the technical solution.
Furthermore, the EPO presenter also indicated that AI specific terminology is not always clear in the field and therefore, extra consideration should be given to the clarity of these terms, particularly when the terms are used in the claims. It is best practice to define all terms in the specification in order to avoid a clarity objection. Moreover, the specification should emphasise the technicality/technical effect of each feature, include at least one technical implementation and a specific use case. Another issue discussed is that method claims may be performed by multiple different actors, so it was recommended to include multiple different independent claims, even if in the same category.
The USPTO’s stance on AI cases is currently that AI is considered to be a mathematical method and therefore not patent eligible. However, as indicated by McRo, Inc. v. Bandai Namco Games America, when it can be proved that the processing performed by the system/method is not the same as that performed by a human, a case may be considered to be patent eligible. Consequently, when preparing a patent application for US filing, it may be prudent to describe the differences between a human operator and the computer implemented method, indicate the technical effect of the invention and described the advantages in the specification. Although this will not guarantee the invention being found to be patent eligible, it should help.
The afternoon session went on to discuss the wider implications of AI on the patenting process. For example, when determining whether an AI invention is obvious, who is the skilled person? Is an invention obvious if a skilled person, aided by AI arrives at the same result?
Also, what happens when AI invents, is the invention patentable? Article 60 EPC states that the right to a patent belongs to the inventor or the inventors’ employee. In addition, an inventor is a natural or legal person (i.e. a human). Therefore, when AI invents, the inventor is no longer a natural or legal person so should a patent be granted? The opinion provided by Dr. Noam Shemtov (Queen Mary University of London) was that when AI invents, the inventor would be considered the person who selected the AI algorithm and input the data and thus patent could still be granted. However, this is just one interpretation of the law, what would happen if the user’s contribution is much less than that of the AI?
Finally, the discussion moved on to whether the current patenting system is even suitable for fast moving AI technology, for example, is it still appropriate to not publish applications for 18 months, or should this be reduced to increase certainty in such as fast-moving technology? Would a new expediated grant system by advantageous?
It is apparent that there is plenty still to discuss with regards to the fast-paced evolution of AI technology and the protection that can be obtained via the patent system. It is a topic which will be monitored closely by the EPO. However, currently the EPO intends to continue examining AI inventions using the two-step approach defined for CII.
The EPO’s own report can be found at https://www.epo.org/news-issues/news/2018/20180530.html