The tech world is watching an unfolding legal issue around an application of generative AI tools used to create computer code. It’s a huge timesaver for programmers, but it turns out that the code is not “written” by an AI, but rather based on samples of code written by humans. In some cases, the samples are outright copies of specific pieces of code, and not everyone is happy about that. I fear similar fallout for generative AI tools used for writing, music, and other creative fields, and for similar reasons. Here’s the money quote from the article on GitHub Copilot:
“But Butterick and other Copilot critics charge that many of Copilot’s suggestions aren’t ‘boilerplate’ at all; they bear the unmistakable fingerprints of their original authors, because Copilot was ‘trained’ on GitHub’s repositories of public and open source code.”
Lawsuits have been filed in the Copilot case. Substitute “GitHub” with Google Books, news websites, online images, WordPress blogs, YouTube, Spotify, and other sources of samples that these AI systems will be trained on, and you can begin to understand the legal risk in other creative fields, too. It’s already happening with low content book scams.
Example: Publisher X uses the AcmeBlogAI engine to “write” a blog post about a specific topic that’s the subject of an upcoming nonfiction release, only to discover that supposed boilerplate language in the blogpost is actually based on an article published on a BigNewsCo website 2 years before that touched on the same topic.
It worries me as a creator, and also as a publisher who needs assurance that shiny new generative AI tools aren’t going to lead to legal trouble. It’s happening in big tech already:
On the other hand, creators have learned to adapt to sophisticated production tools in the past. There is an opportunity to leverage AI tools – but that requires changing habits, learning new skills, and taking other risks that may not be comfortable.