Kirill Kudryavtsev/AFP via Getty Images/Futurism
wikipedia have signed training deals As with many major AI companies, this helped offset some of the exorbitant costs arising from being constantly plundered by data scrapers.
The companies include Amazon, Microsoft, Meta, Perplexity and France’s Mistral AI. As part of a series of licensing deals announced Thursday by the Wikimedia Foundation, Wikipedia’s operator, AI developers are joining Wikimedia’s Enterprise program, which provides direct access to its collection of more than 65 million articles.
Using Wikipedia and its content is free, but the official program allows instant access at a “specially designed volume and speed” for “mass re-users and distributors”, such as AI chatbots.
Wikipedia had already agreed to a licensing deal with Google in 2022, and had also signed deals with smaller AI firms like search engine Ecosia. Now, with the new series of deals, Wikipedia has partnered with every big name in the city, and it sees commercial partnerships as a way to keep the lights on, though it hasn’t provided specifics on the financials of the deals.
“Wikipedia is a critical component of the work these technology companies do, which they have to figure out how to support financially,” said Len Baker, president of Wikimedia Enterprise. told reuters.
“It took us a while to figure out the right set of features and functionality to offer if we were going to move these companies from our free platform to a commercial platform… but all of our Big Tech partners really see the need to commit to maintaining the work of Wikipedia.”
It’s an interesting direction for the nonprofit, which has struggled with AI’s growing hold on the internet. AI companies rely heavily on free sources of information such as Wikipedia to train their AI models, and frequently refer to the encyclopedia for their responses. But as more people turn to chatbots for answers, fewer turn to Wikipedia, which relies on small donations to survive and unpaid volunteers to maintain its articles.
Furthermore, Wikipedia’s large-scale data scraping by AI web crawlers has taken a heavy toll on its servers, which are becoming more expensive to maintain. In fact, Wikipedia was paying for the AI industry’s intensive training – Wikipedia founder Jimmy Wales says that its readers were not donating money for it.
Wells, “They’re not donating to subsidize these giant AI companies.” told associated Press. Donors are saying, “You know what, you can’t actually destroy our website. You have to come the right way.”
There’s another reason Wales engages in partnerships: we are better off than less honest sources.
“I’m personally very happy that AI models are training on Wikipedia data because it is human-generated,” he said. AP. “I really wouldn’t want to use an AI that’s only trained on (Wells is referring to the website owned by Elon Musk. Musk, by the way, retaliated against Wikipedia by launching his own AI-generated “Growwikipedia” as an anti-Semitic alternative, which promptly turned out to be racist.)
All told, it will be interesting to see how this deal is received by Wikipedia editors and authors, who have campaigned against AI content used on the platform, and Rebelled against site owners When they tried to deploy AI-generated summaries on articles.
More on AI: Google is now filling its AI products with ads
