Was 2023 the Wild West era of AI development? Will lawsuits address types of abuse in 2024, or will there be more of it?
The 12 months 2023 will go down in history because the 12 months that generative AI became ubiquitous, but 2024 might be marked by a special sort of change attributable to copyright challenges.
The breathtaking growth of generative AI, spearheaded by corporations like OpenAI, Meta Platforms and Microsoft's Midjourney, has sparked numerous copyright disputes.
Creators, particularly writers and artists, claim that AI's success relies on the unauthorized use of their works, resulting in high-profile litigation.
These disputes largely revolve around web scraping, which involves extracting massive amounts of knowledge from the web to coach AI systems, a practice that’s currently under intense legal scrutiny.
While publishing work online implies that it is essentially in the general public domain, does this now also include the expectation that it should be used to coach AI systems?
Some well-known plaintiffs, authors like John Grisham, George RR MartinSarah Silverman and Mike Huckabee and other copyright holders, including Getty Images and the New York Timesare demanding damages and court orders to stop unauthorized use of their works.
To date, courts have shown some skepticism about claims of copyright infringement related to AI One case was dismissed nonetheless, it was later refiled and more artists were added to the grievance.
There are also quite a few separate legal debates. The UK Supreme Court recently ruled that AI can’t be registered as an inventor for patentswhile a Chinese court spoke of AI-generated art was protected by copyright.
Now even lawyers themselves have grow to be involved in legal cases, with at the least two cases of pretend AI-generated legal references getting used in real legal cases, most recently in cases involving litigation Donald Trump's former lawyer, Michael Cohen.
The defense largely holds up
Technology corporations fear that these lawsuits could significantly hinder AI development. They argue that their AI training methods fall under “fair use” of copyrighted material.
As Meta explained in a press release to the US Copyright Office, “Just as a baby learns language (words, grammar, syntax, sentence structure) by listening to on a regular basis language, bedtime stories, songs on the radio, etc., a model learns.” Language by being exposed – through training – to very large amounts of text.”
This analogy underscores the industry's stance that AI training mimics human learning processes and must be treated as such under copyright law. Essentially, AI corporations argue it's a bit like using other people's work to create and sell a textbook or encyclopedia.
Others are unconvinced and suspect that AI corporations still find themselves in a dangerous environment.
Andreessen Horowitz, a Silicon Valley enterprise capital firm, expressed grave concern concerning the potential impact of those lawsuits on AI development.
They stated: “Putting the prices of actual or potential copyright liability on the creators of AI models will either kill or significantly hinder their development.”
Copyright holders also argue the financial success of AI programs like OpenAI's ChatGPT, declaring that tech corporations must afford royalties and can’t proceed to justify their mental property infringement.
The Authors Guild recently stated: “Licensing copyrighted materials for the training of their LLMs could be expensive – and indeed, consideration must be given to the large proportion of an LLM’s value that comes from professionally produced texts.”
As these legal battles unfold, the AI industry finds itself at a crossroads where the long run of AI innovation is closely tied to evolving interpretations of copyright law.
The decisions made in these cases could significantly influence the course of AI development this 12 months. The extent of this could grow to be clearer – or at the least a bit of clearer.
