WIPR webinar: Patenting artificial intelligence
AI-powered machines and software are still reliant on human beings for their existence and development because, for now, we still need to write the code and manufacture the parts.
But, with AI increasingly becoming a part of everyone’s lives ( Apple’s Siri, for example, uses AI-based natural language processing, while Google’s DeepMind is applying AI to process raw environmental data to emulate human learning), what are the patenting implications?
Susan Keston and Jeffrey Clarke, both partners at IP specialists HGF, discussed these issues and more during a webinar jointly hosted by WIPR and HGF yesterday, February 12.
“AI is pushing forward many new business models, such as driverless cars, and will move society to new places,” said Keston. “We already use secure banking systems, automated diagnosis technology, and voice recognition technology, all of which features an AI component. It’s already becoming part of everyday life.”
It was thought that case law and existing legislation would be enough to deal with AI patenting issues, Keston continued, as the AI is fundamentally based upon mathematical calculations, and hence not patentable.
However, issues start to appear when AI starts to interact with the real world, moving from the theoretical to the practical.
Both participants agreed that simulations are a market with huge growth potential, with AI at the heart of that. You can effectively simulate, for example, numerous different chemical compounds or different building and environmental designs, assess the results, and use that as the basis for investment plans.
However, they commented, the issues arise when we need to determine whether using AI to model pedestrian crowd movement, for example, and designing a building structure accordingly, is technical or technology—and this has implications for its patentability.
The aftermath of the US Supreme Court’s Alice decision can lead to some AI innovations being categorised as patent-ineligible “abstract ideas”. However, the European Patent Office ( EPO) has a stable and long-established body of case law in the assessment of computer-implemented inventions via the problem-solution approach.
“Classifying data is not enough to prove you have a valid patent,” added Clarke. “You need to indicate that you have a use for that data.”
Keston continued that there’s “simply more awareness of AI now, and the importance of the technology is evident in the huge amounts of money being spent on its development, as well as the dramatic increase in the number of AI-related patents filed over the last few years”.
Deep learning
Another issue raised was that of ‘deep learning’ technology, started to appear in 2012, and differs from other forms of AI because it’s designed to rely less on humans to quantify and label data, and can seemingly teach itself new rules and processes.
For example, UK-based AI company DeepMind Technologies (now a subsidiary of Google) has created a neural network that can learn how to play the board game ‘Go’, one of the most complicated games in the world. It’s technology subsequently beat Lee Sedol, a Go world champion.
“AI has the potential to be all persuasive, and it’s still being invented, which is so exciting,” said Keston. “If we just look at AI on 5G devices, that offers massive possibilities.”
Listen to the webinar here.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox.
Today's top stories:
Buffett’s Brooks sues Brooks Brothers over TM registration
