You’re not the only one who turns to Wikipedia for quick facts. Lately, a deluge of AI bots training on Wikipedia articles has put enormous strain on the organization’s servers.
To curb the influx of “non-human traffic” scraping the site for training data, Wikipedia is taking a proactive approach: serving up its data directly to AI developers.
On Wednesday, the Wikimedia Foundation announced a partnership with Google-owned company Kaggle to release a beta dataset “featuring structured Wikipedia content in English and French.” Uploaded on April 15, the company said the dataset “simplifies access to clean, pre-parsed article data that’s immediately usable for modeling, benchmarking, alignment, fine-tuning, and exploratory analysis.”
According to Ars Technica, bots that scrape Wikipedia and Wikimedia Commons pages have consumed 50 percent of its bandwidth, putting a massive strain on the nonprofit’s entire operation. Wikimedia hopes that serving up data to developers will dissuade them from deploying bots all over its pages.
The rise of generative AI has let loose a flood of scraping bots hungrily crawling all corners of the internet for more data. To compete against rivals, AI companies have a seemingly insatiable appetite for data. This has included copyrighted works, a contentious issue with artists. Authors, artists, and musicians are arguing in court that this training violates copyright law when it’s done without credit, compensation, or consent.
That’s why companies like Meta and OpenAI are currently embroiled in legal battles over copyright infringement from plaintiffs like the Authors Guild and The New York Times, who argue this practice is not protected by the fair use doctrine.
But the difference here is that all Wikipedia content is licensed under the Creative Commons Attribution-ShareAlike license, which means its content is free to use as long as it’s properly attributed and distributed under the same license. The Wikimedia Foundation told Gizmodo that Kaggle paid for the data through the Wikimedia Enterprise, and AI companies “are still expected to respect Wikipedia’s attribution and licensing terms.”
The partnership between Wikimedia and Kaggle represents a more nuanced way forward, allowing AI companies to train models on internet data that’s been legally and, at least more ethically, obtained.
