The cycle of breakneck AI news is just spinning faster and faster. Everyone just close your eyes, hold your breath, and dive in. Because this week is no exception. As usual, OpenAI is at the center of a lot of the attention. From a ChatGPT feature that can search the web and cite sources, or maybe a Google search competitor, or maybe… none of those things? It’s unclear.
But one thing that was clear is that OpenAI’s relationship with data and copyright really spilled out into the open this week. First up, Stack Overflow and OpenAI announced a partnership for OpenAI to get access to the company’s data. We’ve seen similar partnerships between the likes of Reddit and Google after companies with data pushed to get compensated for their platforms being used as training data for LLMs.
Stack Overflows users didn’t react positively to the news, with many of them deleting their contributions to the site en masse. And Stack Overflow responded by banning those users. As Gergely Orosz points out:
“The reality is that now StackOverflow and Reddit are both sites where they are free to use to provide answers because you are providing training data for AI models with every one of your answers (and past ones). The only way to opt out is to stop contributing to them.”
Stack Overflow wasn’t the only data set in OpenAI’s crosshairs. In a class action lawsuit brought against OpenAI by the Author’s Guild, it was alleged that OpenAI had used 100K+ copyrighted books in training its models, and then deleted the datasets. That’s just one of the many lawsuits from artists, writers, and publishers that are pushing back on their work being used in training data.
OpenAI is attempting to respond by announcing a tool called “Media Manager,” which would “allow ‘creators and content owners to tell [OpenAI] what they own’ and specify ‘how they want their works to be included or excluded from machine learning research and training.’” But details are sparse, and the execution seems complicated. Even companies like Reddit, who have made access deals, are attempting to introduce public content policies to better regulate how the data on its platform is being used.
Ironically, OpenAI is the pot calling the kettle black on the other side of copyright law. The company recently made a copyright complaint against the ChatGPT subreddit for using the OpenAI logo.
The unfortunate reality is that the same copyright battle will continue to play out across writing, art, music, and more, because what we thought was the unstoppable force of the open internet is now running into an immovable object of the exorbitant demand for data that AI companies have. And that demand is only going up as we see ever more competition in the space. Even Microsoft, the sugar daddy of OpenAI, is working on its own in-house LLM called MAI-1, which is explicitly meant to compete with OpenAI.
