Can a machine patent an invention?

robot on the piano

With AI patents increasing 400% over the past decade, businesses, regulators and philosophers ask where AI inventions will go next and how best to answer ethical questions of ownership and liability

In his opening remarks at the AI: Decoding IP conference in London this week, UK Supreme Court Justice Lord David Kitchen addressed one of the burning questions of IP and AI: can a non-human file a patent for an invention?

As stated in Article 52 of the European Patent Convention (EPC), a patent can be awarded for all inventions provided they are new and include an inventive step. Excluded from the provisions are algorithms and mathematical methods.

Across the pond in a post-Alice US, Section 101 limits patentability to any new and useful process, machine or composition of matter. Laws of nature and mathematical algorithms are excluded from the US Code, echoing the regulations in Europe.

According to Lord Kitchin, the exclusion indicates that a human being must be the source of invention. “So there is once again a presumption AI is not patentable,” he said.

“But there will be cases, and potentially more and more of them, where the artificial intelligence (AI) has carried out all of the predominant part of the creative and innovative work.”

In some domains the science fiction has become a reality. Anat Elhalal, head of technology at London-based innovation centre Digital Catapult, told the audience that AI has already written books and composed music.

For the moment the copyright for the created works belongs to the coders and inventors of the AI. But what happens when a cognitively capable machine creates a new and useful invention without the input or direction of a human being?

And of equal importance, does a robot need the incentive provided by a patent to innovate?

“Patent systems are intended to encourage and reward innovation. Inventors disclose their inventions to the world in return for a monopoly of a certain duration. Do the same considerations apply to inventions generated by AI?

“I am not sure that they do. Computers do not respond to rewards or incentives, although maybe one day they will,” said Lord Kitchin.

He added: “If we allow the notion of inventorship to escape from its human boundaries, where will it end and will the system become swamped by computer-generated inventions and end up stifling human innovation?”

Perhaps even more fundamental to the question of ownership is the question of rights. Can a robot be given the same rights to a patent as a human being? At its essence, a patent is a contract between the state and an inventor. A robot cannot yet give consent to sign a contract.

New Applications of AI

While inventions from autonomous robots remain within the realms of science fiction for the most part, AI is rapidly speeding up the pace of innovation today.

Andrew Burgess, a strategic advisor on AI, explained at the conference what exactly is meant when one talks about artificial intelligence, pointing to different applications for AI and machine learning, and indicated potential paths for future growth.

Citing examples of image and speech recognition, Burgess explained how machines help businesses take unstructured data and make it structured in a way that helps them understand the information more coherently. He also cited prediction and the clustering of data as ways AI have helped businesses grow.

Addressing the audience he said: “The final objective of AI is to help understand why something is happening and here we fall into the realm of science fiction and labs. No machine has cognition, yet.”

With or without the capacity to recognise themselves, machines are doing a remarkable job recognising others with facial and speech recognition.

Speaking about the ethics around data protection, Lord Tim Clement-Jones, chair of the UK House of Lords AI select committee, said: “Treat what I am about to say as provocation. A number of key questions to raise is how do we recognise the need to protect data sets and the need to protect people?”

As machines continue to accumulate more information about the details of citizens’ lives, Lord Clement-Jones asked his audience if governments should consider the regulation of access to some data sets. Secondly he posed the question of how should regulators resolve the ethics regarding the distribution of benefits of any autonomous AI created invention. Should the invention belong to the company owning the AI or to society at large?

When asked how parliament is looking to regulate future questions of ethics and IP, Lord Clement Jones cracked a half-smile, saying: “Some people in parliament are beginning to understand the ethical issues, but most politicians need to be lead extremely gently by the hand.”

Stepping into the future

Where AI will take innovation next is anyone’s guess. With all the potential for growth and progress, businesses are filing patents for AI at an unprecedented rate. The UKIPO and WIPO issued a report this year showing a 400% increase in AI related patents with 160,000 patents filed in 2017 alone.

Francis Gurry, director general of WIPO, said: “Artificial intelligence is fundamentally altering the human experience, with profound implications for the legal and regulatory regimes that underpin our communities. These include the IP systems that promote human innovation and creativity.”

At the heart of the IP debate around AI is the central question of how to incentivise innovation. When asked by an audience member whether a robot should be able to file a patent for its invention, Lord Kitchin reflected on the central contract of the patent.

“The question of whether or not a robot can file a patent comes back to the question of what we are trying to incentivise. If we want people who make the investment to reap the rewards there needs to be some sort of protection,” he mused.

The question of extending IP rights to a robot may well be answered far in the future as robots develop more human like capacities. While at the moment AI is limited to suggesting a playlist of break up songs based on a coded algorithm of specific words, it cannot yet understand the poetic genius that sets Leonard Cohen apart from the generic and permanent adolescence of say, Taylor Swift. Without cognition and the capacity to feel emotion the ethical questions of rights and ownership could be answered by future generations.