Wikipedia pauses AI summary experiment after editors say it 'would do immediate and irreversible harm to our readers and to our reputation as a decently trustworthy and serious source'

AI summaries are all up in our search engines these days, and I think it's fair to say the response has been varied to date. The Wikipedia editor community, however, appears to have taken a strong stance on a recently proposed experiment by the Wikimedia Foundation to add AI-generated summaries at the top of Wiki articles, causing the test to be paused for now.

“This would do immediate and irreversible harm to our readers and to our reputation as a decently trustworthy and serious source" said Wikipedia editor Cremastura (via 404 Media). "Wikipedia has in some ways become a byword for sober boringness, which is excellent."

"Let's not insult our readers' intelligence and join the stampede to roll out flashy AI summaries. Which is what these are, although here the word ‘machine-generated’ is used instead."

The comments came in response to a Wikipedia village pump announcement from the WMF web team, informing editors that a discussion was underway regarding the presentation of "machine-generated, but editor-moderated, simple summaries for readers."

Among the proposals was a planned two-week experiment on the mobile website, wherein 10% of users would be given the opportunity to opt in to pre-generated summaries on a set of articles, before the experiment would be turned off and used to collect data on the response.

To be fair, that seems like a pretty tentative step into AI summary testing on the Wikimedia Foundation's part, but the response was mostly negative. Several editors simply commented "yuck" to the proposal, with one calling it a "truly ghastly idea." However, some seemed to take a more positive view:

"I'm glad that WMF is thinking about a solution [to] a key problem on Wikipedia: Most of our technical articles are way too difficult", writes user Femke. "Maybe we can use it as inspiration for writing articles appropriate for our broader audience."

The announcement appears to be a continuation of the WMF's proposal earlier this year to integrate AI into Wikipedia's complex ecosystem. In May the WMF announced it was implementing a strategy to develop, host, and use AI in product infrastructure and research "at the direct service of the editors", which seems to have been received by the community with a similar degree of trepidation.

Speaking to 404 Media, a Wikimedia Foundation manager said: "Reading through the comments, it's clear we could have done a better job introducing this idea", before confirming that the test has been pulled while it evaluates the feedback.

For those familiar with the intricate debates that go on behind the scenes of even the seemingly most straightforward Wikipedia articles, all this internal discussion probably comes as no surprise. Wikipedia has thrived on lively discourse and editor debate regarding even the smallest of details of its pages, and over the years has become one of the internet's primary sources of reference—with the English version said to receive more than 4,000 page views every second.

Yep, I just sourced Wikipedia statistics from a Wikipedia page discussing Wikipedia. How's that for a fractal of dubiosity?

Back when I were a lad, I was discouraged from using Wikipedia as a reference in my studies, as in the dark days of the early 2000s it was regarded as untrustworthy due to its reliance on crowdsourced information.

However, although it's probably still best not to cite it as a reliable source in your college essays, over the years that perception has changed—with strict moderation (and rigorous guidelines) helping to cement its reputation as one of the more reliable information outlets on the internet.

That being said, in a world where the very sources it relies on are increasingly affected by the rise of AI-generated content, I do wonder how long it will be before even this stalwart of internet reference succumbs to some of the downsides of modern AI—in ways that many of its editors (and those of us that use it on the daily) would rather it didn't.