Artificial intelligence is leading a patent renaissance. In recent years, patent applications for inventions developed using AI have increased exponentially, not just in tech but in every field. In 2018, one inventor took a bold step that was at once revolutionary, yet probably inevitable. Dr. Stephen Thaler filed patent applications all around the world for two simple inventions — a food container and a flashing light emergency alert device. The big shock was that Thaler listed an AI machine as an inventor, raising questions about the legal personhood of AI and the criteria for awarding patents.

In the United States, patent examiners apply standards to determine if an invention is patentable: It must be subject-matter eligible, new, useful, and non-obvious. The obviousness question has been the subject of legal debate and litigation for a century, well before the dawn of AI. Now artificial intelligence, machine learning, and neural networks have raised new questions about the obviousness standard.

Thaler is not a food container designer or emergency light emitter engineer, so he never could have come up with his inventions without AI. To qualify for a patent, examiners ask if the invention in question would be obvious to a hypothetical person having ordinary skill in the field. What is going to happen to the obviousness standard when everyone is using AI? How will patent examiners determine an industry standard?

As AI becomes a standard tool of nearly every trade, anyone who doesn’t use it is going to be left behind because the obviousness bar is going to get higher. This article discusses how AI is already changing the game, and how this new tool is going to make it even harder to win patents.

AI’s First Patent Application

Thaler created an AI system called the Device for Autonomous Bootstrapping of Unified Sentience, or DABUS. In his patent applications, Thaler said that DABUS applied its own creative functionality to invent the food container and the light device. So far, the United Kingdom and the U.S. Patent and Trademark Office (USPTO) have rejected the applications and dismissed appeals, and other rejections have come from the EU with appeals still in process. South Africa did accept the applications, but it accepts any application that complies with formalities. And upon appeal Australia decided AI can be listed as an inventor, but cannot own a patent.

Thus, the United States still holds that an inventor must be a natural person, but the debate around AI’s impact on patent eligibility is far from over. AI is creating waves around the obviousness standard as its use becomes more widespread.

The Obviousness Test

Patents are intertwined with the fabric of our culture; in fact, the symbol for inspiration itself, the lightbulb, is one of the most famous and widely used patented inventions in history. Most inventors who have obtained patents know, however, that coming up with a patentable idea rarely happens in a flash of lightning. Inventing is a blend of persistence, inspiration, science, and numerous other ingredients. After a long process, a product is born. It takes a lot of knowledge, skill, time and work to come up with something truly innovative, something novel enough to win the prize of a patent.

The legal system created the incentive of a patent to encourage the innovation that comes from all this hard work. Most innovations don’t come in quantum leaps, so the legal system has decided that trivial improvements in existing technology do not qualify for patents. But if an inventor comes up with something truly new and unique, a bona fide evolutionary step forward, the legal system awards the innovation with a monopoly for two decades.

There are certain legal standards to getting a patent. An invention must be new, useful, and non-obvious. The obviousness test is where the AI question gets very interesting. The legal system created a character called a “person having ordinary skill in the art” (PHOSITA). To obtain a patent, an invention must not be obvious to the PHOSITA. That means the PHOSITA would be an average person in the given field, having access to the same tools, skills, and knowledge base.

For example, if an inventor decides to improve on a mousetrap, to obtain a patent she needs to engage in an unusual inventive effort beyond what a typical mousetrap inventor would have known what to do. If an inventor files a patent application for a new mousetrap spring, would an ordinary mousetrap inventor have known to come up with that spring?

Perhaps the mousetrap inventor decided to use neural networks to help come up with new spring designs. The AI might help her calculate torque, pressure, the way different metals would impact the spring’s operations, etc. The new spring may mark an improvement on the technology, but if an ordinary mousetrap inventor could have arrived at the same invention, then the spring wouldn’t qualify for a patent. However, if the AI assists in developing a novel spring design that would have been beyond the skill of an ordinary mousetrap inventor to design, that invention may qualify for a patent (assuming the mousetrap inventor has hired a good patent attorney to draft the application).

Crucially, the U.S. patent office has decided that an inventor’s education level should not impact a patent application. The important consideration is the standard education level of the PHOSITA. Likewise, inventors do not need to disclose the use of AI on their patent applications, and patent examiners should not consider the use of AI in the invention process until it becomes a standard tool of the given trade.  Yet the example above illustrates something strange about the current situation, namely that an inventor may use AI to develop an invention that would have been beyond the skill of the inventor to create on his or her own, and obtain a patent on the result, even if other inventors could have used the same AI to create the same invention.

AI in the Toolkit

As more companies and inventors use AI to create new inventions, the legal system will have to shift. At some point patent examiners will have to start assuming that a PHOSITA has access to AI, which will raise the bar for obviousness in the patent process. That’s not the case now. Currently, AI confers a major competitive advantage, and companies developing new AI processes or using it to come up with inventions may have a golden goose on their hands. In fact, many inventors in the technology space may be overlooking a patentable golden goose.

Software engineers and other inventors in the tech space may become so accustomed to their inventions that it becomes obvious to them, but patent examiners look at applications from the standpoint of a hypothetical PHOSITA. For this reason, the attorneys at Blueshift IP regularly review the invention portfolios of their clients to determine if anything is patentable. It’s helpful to have an experienced eye and an objective perspective from a third party to review what an inventor has developed. At the same time, inventors may prefer to keep their AI processes as trade secrets, hide away the golden goose, and just keep obtaining patents on the goose’s eggs — the inventions developed with the assistance of AI.

Changing All the Games

Artificial intelligence isn’t only raising the bar on obviousness in software and technology. Across the field in other industries, AI is becoming a major edge in innovation that will likely raise the bar on obviousness. Neural networks can help with inventing by combining disparate ideas to create new consumer products or improve processes, but more importantly, AI speeds up the innovation cycle.

AI helps companies come up with more prototypes easily and quickly. Or it can help researchers rapidly sift through and analyze data. For example, by analyzing chemical and biological reactions, AI identified potential treatments for COVID-19 in just three days. Researchers and normal computers could have done that work, but it would have taken much, much longer. In the same way, AI is boosting the skill level of inventors across the board. (You can read more about AI in healthcare here.)

AI isn’t replacing human ingenuity and creativity when it comes to patentable inventions — yet. AI is useful in processing data and cutting out a lot of drudgery that makes inventing such a slog. But humans must feed the information to AI, assigning value to datasets, and guiding the machines to focus on what’s most important. The partnership between humans and AI is where the magic happens.

We are on the threshold of a revolution. Artificial intelligence will go from being an edge to being a standard when it comes to patentable inventions, and anyone not using AI will be left behind. Once the legal system decides PHOSITA has access to AI, it’s going to be much harder to win patents.

The post AI Raises the Obviousness Bar for Patents first appeared on Blueshift IP.

Photo of Robert Plotkin Robert Plotkin
Founding Partner @Blueshift IP
Attorney Robert Plotkin has been a leader in obtaining software patents for two decades, and consistently obtains software patents for clients even after the Alice Supreme court decision stopped most companies from obtaining software patents. He uses his decades
Founding Partner @Blueshift IP
Attorney Robert Plotkin has been a leader in obtaining software patents for two decades, and consistently obtains software patents for clients even after the Alice Supreme court decision stopped most companies from obtaining software patents. He uses his decades of legal and engineering experience to maximize the value of his clients’ patent portfolios – allowing them to realize the largest return on investment even in the post-Alice world. His clients have profitably sold and licensed the software patents he has obtained for them to major corporations worldwide.