[pradn]

I'm unsure if this will happen. There's plenty of checks-and-balances for Wikipedia edits. There's automated spam detection, editors manually looking over edits for articles on their watchlist, editors who look over subtopics, and even editors that take a look at the general stream of edits. It's already possible to flag mass edits. As for whether ChatGPT will inflect the subtle tone and bias of edits made using it, that's the same as bias from human users. And the same mechanisms for dealing with human bias apply here.

In terms of practical utility, for the vast majority of humanity, access to translated articles in their local language is the biggest problem, I think. There is no Yoruba-language Wiki article on General Relativity, for example. Second comes entire biased communities - like some of the smaller Wikis are full of far-right editors, and most editors (like 90%) are men.

[worrycue]

I can see AI bots submitting convincing edits at random times in no particular pattern. Eventually they will overwhelm Wikipedia checks and balances.

[tivert]

>> I wish I had the time or facility to take a snapshot of wikipedia now before the imminent deluge of Chat-GPT based updates that start materially modifying wikipedia is some weird and unpredictable manner.

> I'm unsure if this will happen. There's plenty of checks-and-balances for Wikipedia edits.

I think it will. It's so tedious to edit Wikipedia (due to bureaucracy and internal politics) that their editorial population is in a long-term decline, which means their oversight ability is declining too.

Probably what will happen is LLM generated content will creep into long-tail articles, then work its way into more "medium-profile" articles as editors get exhausted. The extremely high-profile stuff (e.g. New York City), political battleground articles (e.g. Donald Trump), and areas patrolled by obsessives (railroads, Pokemon) will probably remain unaffected by the corruption the longest. At some point, the only way to resist will to become much more hostile to new editors, but that's also long-term suicide for the project.

I think they're painted into a corner.

[pradn]

I mean, maybe. AI on the "good side" will also improve. It should be possible to check a sentence against its reference with LLMs. And anything not sourced is suspect, just as it is now.

I also don't like the attitude of Wikipedia being "them", as in "their editorial population". It's our public good, like our air, and everyone should care to ensure its high quality. If you see a problem in the world, you have to try to fix it, instead of sitting on the sidelines, looking from the outside in.