| | News

WIPO: AI is disrupting the intellectual property status quo

View of the WIPO Main Building from the Place des Nations in spring, 10 May 2016. (Credit: WIPO/Emmanuel Berrod)

Artificial intelligence is developing fast and is raising questions about who, or what, can own intellectual property.

Ulrike Till, director of the intellectual property and frontier technologies division at the World Intellectual Property Organization (WIPO), spoke with Geneva Solutions ahead of the Geneva Science and Diplomacy Anticipation summit, where the organisation will be taking part. Experts from across the science and diplomacy communities will discuss AI’s rapid advance and how to make it inclusive.

Till’s team looks at the implications of AI and other frontier technologies for intellectual property. Central to WIPO’s work is “horizon scanning”, says Till. WIPO anticipates developments on the technological frontier so that it can support and protect innovators and creators.

Among other puzzles, she says the team also spends time investigating how AI and its growing contributions to creative products is putting the flame under fundamental assumptions about intellectual property.

AI creations. One of the main questions the intellectual property community faces is whether AI can count as a creator, capable of owning what it creates.

“Certainly one of the challenges is the IP system is designed to foster human innovation and creation. What do you do with something that is created by AI, if that ever happens?” says Till.

Deep learning technologies, also known as artificial neural networks, are the most autonomous forms of artificial intelligence. Artificial neural networks are first trained to associate certain data with certain outputs.

For example, some “natural language” technologies are fed words, which they are programmed to associate with labels like “noun”, “verb”, and “adjective”. They can learn sentence structure by tracking label configurations in input sentences. This is called the training phase.

Once trained, these technologies can generate sentences when fed a new word batch. Based on associations it has learned between words and their role in a sentence, the AI can churn out comprehensible literature using new word inputs.

The programmer writes the code that gives a system the tools to associate a word with its role in a sentence. But the programmer has little idea how the AI will arrange new words it has been given. Who owns these sentences?

AI in court. Some argue that the AI is the owner. Rapid developments in deep learning technology mean that courts are already having to rule on whether AI systems get to hold patents – documents that certify ownership of intellectual property. Earlier this summer, UK and US courts ruled that AI cannot be patent holders. “Only a person can have rights,” one of the UK judges wrote.

In July, a South African court ruled otherwise. Their explanation is still pending.

The subject is still hotly debated in Australia, where the federal patent office is appealing a court decision that favoured AI. The initial decision read, “an inventor is an agent noun; an agent can be a person or thing that invents”. The judge added that the invention was patentable, but that the nature of that deep learning machine means “it cannot sensibly be said that a human is the inventor”.

AI on the move. AI’s stage of development figures heavily in the debate. Till explains that at the moment we have “narrow” AI, “AI that is designed specifically to solve a problem or to look at a particular issue”. Many experts say that AI is not advanced enough to be considered a patent holder.

This could change within the next 25 years. GESDA anticipates that within five years we will have a second wave of AI, a third wave in ten, and a fourth in twenty-five. In that time we may see “general” AI. Till explains that “general” AI would be more human-like, able to tackle a broader range of problems and need much less data to produce sophisticated results.

“We are certainly seeing a decrease in the human input and more of a contribution of the AI,” says Till.

Small stakeholders rising. The GESDA summit will ask how to make such a transition to more advanced AI inclusive, in response to growing concerns that big tech is dominating AI development at the exclusion of smaller stakeholders.

In organising the WIPO Conversation on intellectual property and frontier technology – meetings that inform intellectual property policy – her team engages with a range of stakeholders. Till is hopeful about the growing role of smaller companies and individuals and says their presence should not be underestimated.

“Sometimes I think there can be a perception of AI being limited to a number of companies or a number of sectors. But we’ve really had contributors that were showing us what they’re doing and their businesses from every single geography around the planet; from the creative industries to the innovative technology industries; and from very, very small companies and entrepreneurs, on to much larger enterprises.”