AI as a judge and patent-destroying tool: panel discussion
Artificial intelligence (AI) technology is beginning to touch every single industry, and the IP space is certainly no exception. But uncertainty remains as to how the rapidly advancing technology can benefit the IP system and whether the system needs to and can adapt to take full advantage of these advancements.
Whether or not machine learning can power decisions on the likelihood of confusion in the trademark arena was one of the first questions raised by Ilanah Fhima, reader in IP law at University College London, yesterday, June 19.
Fhima was speaking at the “AI: decoding IP conference”, hosted in London by the UK Intellectual Property Office (IPO), in cooperation with the World Intellectual Property Organization (WIPO).
During the panel session, “Regulation & incentives breakout sessions–registered rights”, Fhima looked at the practicalities of an AI making likelihood of confusion decisions.
The EU Intellectual Property Office (EUIPO) has launched an AI-powered image comparison search, where you add your image in and trademarks that are similar to your image appear. But this is just the first stage, since the AI isn’t considering any likelihood of confusion and the system itself can be patchy, according to Fhima.
“I took a picture of a Lego brick, plugged it into the EUIPO’s [comparison search system] and it gave me a box of biscuits. It’s understandable as it’s looking at blocks, 3D rectangular things,” she cautioned.
Fhima had significant reservations about having AI as a decision maker on whether there has been infringement of a trademark.
Gwilym Roberts, chairman of Kilburn & Strode in London, posited that it might be attractive for businesses to get rid of judges in favour of an AI solution.
He said: “At the moment, when we advise clients we give them a percentage [of their likelihood of winning, based on] what the judge is likely to say when you get to court. If you had a system where, even if slightly flawed, you knew what the decision is going to be at the outset, it would give you 100% certainty about your business decisions. Wouldn’t that be attractive to business?”
Moving down a level, AI could be used to augment decision makers, for example, providing a preliminary kind of decision before a human makes the ultimate decision.
Fhima warned: “One of the fears I have is that even if we don’t use AI as a proper judge, if you have a preliminary system and a formal decision to be made afterwards [by a human], you end up with a two-tier justice system. Small and medium-sized enterprises will perhaps stick with the preliminary decision, but better funded companies will go on to bring their case in front of a person.”
That’s not all—another problem may arise where, with an AI as a judge, parties will receive an outcome but not necessarily a reasoned decision.
In the UK and before the EUIPO, parties have the right to ultimately appeal against the decision of a hearing officer. But, if you have an AI decision and there is no reasoning apparent, how are you going to make that appeal, asked Fhima.
As discussed in many sessions, AI is only as good as its training data. And previous EUIPO trademark decisions sometimes have “interpretations of languages which suggest a lack of familiarity with the language” of the applicant, according to Fhima. Ultimately, the AI will be replicating these biases.
She summarised: “Are we going to replace judges with computers? No. Can we augment decision making? Perhaps.”
The destroyer of patents?
Heli Pihlajamaa, director of patent law at the European Patent Office (EPO), said: “New and emerging technologies put us in a situation where we have to reflect and think about the very basics of IP rights and the patent system.”
Pihlajamaa added that there are three main categories of AI: inventions on AI such as machine learning, inventions which use AI, and inventions created by AI.
In the first two categories, the EPO has seen a gradual increase in patent applications since the beginning of the millennium, and now receives applications for about 1,200 patent families each year. The EPO hasn’t yet received any applications for the third category.
“Anything in the area of patent law cannot be done in isolation. It’s very important to have everybody globally onboard. Having consensus on any such fundamental change is already a huge task,” Pihlajamaa said.
The EPO’s discussions on AI with member states have focused on four aspects: patentability, disclosure, persons skilled in the art, and inventorship.
With respect to disclosure, Pihlajamaa said there is concern over ‘black box’ patenting, with inventions in the AI space not being sufficiently disclosed in patent applications.
Pihlajamaa added: “I have seen a lot of new technologies and often, when we talk about new technology, there is a new language created…a new [set of] terminology. It takes some time before there is a general agreement on how we express these different things.
“But, as I hear from examiner colleagues, we are already there. There’s already a lot of common understanding about how to draft patent applications [in the AI technology area].”
Sir Robin Jacob, the Sir Hugh Laddie chair of IP law at University College London, added that patent law will survive the fourth industrial revolution of AI, just as it has survived the first three.
“The very function and purpose of the patent system has always been a little confused. On one hand, there is the idea of the heroic lone inventor who has a flashbulb eureka moment. There’s quite a different view of how the patent system really works, not so much to have ideas in the first place but to have the ideas and develop them,” he added.
We’ve moved away from seeing a patent as blueprint of invention to using patents to protect the development of the idea itself, Jacob said. In turn, AI is expected to affect this system.
“Firstly, you may find the idea itself is much more difficult to protect. AI will undoubtedly be able to do better searches than current searches ... that will be a patent destroying tool. It will also affect obviousness because the invention has to be obvious to the person skilled in the art and this person now has a new tool. This may also knock out some patents,” said Jacob, who is also a former judge in the Court of Appeal of England and Wales.
The idea that a person skilled in the art is actually a team of people was developed in the 1960s, said Jacob. He added that the AI computer is just another part of the team of the person skilled in the art, and it will make some things more obvious.
“Will everything be obvious? No. I’m reminded of the fact that Albert Einstein was an examiner in the Swiss Patent Office. I always wondered what happened, [did he always say] ‘obvious, obvious, obvious’ [when looking at applications?]. The law will cope with AI just as it has coped with putting teams together,” he concluded.
Birgit Clark, lead knowledge lawyer, IP and technology at Baker & McKenzie, moderated the panel.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories like this sent straight to your inbox.
Today's top stories:
TPN Europe 2019: USPTO and EPO aligning on AI patentability
USITC to investigate Amazon, Microsoft, and Samsung
