Patenting Artificial Intelligence Might Hamper Progress, EFF Says

The Electronic Frontier Foundation launched a project last year to measure progress in artificial intelligence innovations and understand the legal, political, and technical issues potentially raised by those inventions. Some eight months later, the project has tracked rapid progress of those technologies, in particular in machine learning. According to the foundation, patents might be hampering the progress of artificial intelligence, and with the risk of patent trolls claiming rights on patents on machine learning systems.

According to a September 2017 post by the Electronic Frontier Foundation, titled “Stupid Patent of the Month: Will Patents Slow Artificial Intelligence?,” some “very broad patents” have been issued in the artificial intelligence (AI) space. The post cited a Google patent on a machine learning technique called “dropout.” “This means that Google could insist that no one else use this technique until 2032,” EFF said. Microsoft also as a pending patent application on active machine learning, they said.

“Patents on fundamental machine learning techniques have the potential to fragment development and hold up advances in AI,” the post said, adding that AI patents, as a subset of software development, “are likely to raise many of the same problems as software patents generally.”

Last June, EFF launched the EFF AI Progress Measurement experiment, monitoring advances in AI and machine learning.

The project seeks to understand what types of AI EFF needs to start engaging with on legal, political, and technical safety fronts. Machine learning researchers were encouraged to give feedback and contribute to the experiment.

As it pursues the project, EFF will monitor patenting in AI and try to gauge its impact on progress, EFF said in its September post.

According to Peter Eckersley, chief computer scientist at EFF, the project “has tracked rapid progress on tasks ranging from understanding and translating written language, recognizing images and answering questions about them, to playing numerous kinds of games.”

“It’s incredible how fast machine learning is advancing on all these fronts; we’re really living in the early robot future,” he told Intellectual Property Watch.

“Patents don’t seem to be assisting this cascade of invention in the slightest, and there’s good reason to fear that intellectual property can only hamper the progress of artificial intelligence as a field,” he said.

“We can be sure than in five to ten years, patent trolls will be popping up everywhere, claiming to hold patents on all of the machine learning techniques that are currently being invented in universities and commercial research labs, and generally published openly on arXiv and GitHub,” he added.

 

Image Credits: EFF