It was a hugely disruptive communications technology that broke the grip of elites and allowed the masses access to information that had been tightly controlled.
That’s a reference, of course, to the printing press, the original information revolution. The arrival of the printing press swept away the old world of careful copying by religious orders. In the decades that followed, the new technology also created chaos, where anyone creating an original work could see it reproduced without compensation.
The result, eventually, was the birth of modern copyright law, which struck a balance between the rights of content creators and those wanting to reproduce that content. Now, nearly six centuries after the invention of the printing press, artificial intelligence is threatening to upend that balance.
Technology and the law
This is part of a series on the intersection of technology and law.
Cellphones: Time to disconnect students from their devices
The raw resource of artificial intelligence is text and data mining, which scrapes vast amounts of information from across the internet. That information is in turn the foundation of generative artificial intelligence, including large language models that create a human-like response to queries.
The promise of AI is vast, but the potential drawbacks for content creators is also huge. The purpose of copyright is to foster the creation of original works by preserving the economic benefits from doing so. To the degree that AI simply bypasses those content creators, it undermines the benefits – and imperils the act of content creation. Why pay an artist, a writer, a journalist or any other content creator if AI can simply deliver a processed amalgam of dozens, hundreds or thousands of versions?
Court battles are already being waged. Last year, The New York Times sued OpenAI, which owns ChatGPT, and Microsoft, alleging that millions of its articles were used to create chatbots that now compete with its efforts. Eight other U.S. newspapers launched similar litigation in April.
On the other hand, overly onerous Canadian restrictions could smother AI innovation, and allow other jurisdictions to capture the benefits of this revolutionary technology.
Is there a way (or a need) to constrain the power of AI to protect copyright, and content creators? Ottawa is in the midst of a consultation process on generative AI and copyright that examines those fundamental concerns.
One of the biggest questions is whether there should be copyright carve-outs for text and data mining. Supporters of such exceptions argue that large language models depend on scale; data fortresses would undermine their purpose.
There is a solid case for exemptions for non-commercial operators conducting basic research on AI. The United Kingdom, for instance, has already enacted legislation that allows for such limited exemptions, ideally with acknowledgment to those whose data has been ingested. Canada could follow suit with similar limited exemptions for non-commercial efforts to experiment and innovate.
Broader exemptions that would allow a free-for-all for text and data mining would be foolhardy.
Supporters of a broad exemption argue that the resulting output bears little resemblance to the information scraped. The original sources can always be accessed, they say.
But to argue for a broad exemption is to ignore the clear danger that generative AI, and its ability to amalgamate and recombine vast amounts of data and serve that up easily to end users, will simply push content creators out of public view.
And such a radical policy is unnecessary: the companies that operate AI technology can license the content they need to feed their large language models. University of Ottawa law professor Michael Geist makes an excellent point that licensing could end up entrenching the power of big incumbents such as Google and Microsoft, since startups would not be able to afford the cost of content licences. That problem can be addressed by setting a revenue threshold, below which licensing fees are nominal.
But it’s clear that big technology companies are creating enormous economic value using the work of (a vast number of) content creators. Canadian policy-makers need to act now to head off the obvious harms that will result if copyright law is not updated to meet the challenges of the 21st century.
Machines Like Us: More from The Globe and Mail
AI utopians and doomsayers have one thing in common: They’ve bought into too much hype about what the technology can do. Researcher Kate Crawford spoke with The Globe’s Machines Like Us podcast about how to reach a more realistic view. Subscribe for more episodes.
Editor’s note: A previous version of this article incorrectly stated that the printing press was invented eight centuries ago. It was invented six centuries ago. This version has been updated.