AI Is Tearing Wikipedia Apart

[suckadmin]

The irony being how much WP has been used to train LLMs

Volunteers who maintain the digital encyclopedia are divided on how to deal with the rise of AI-generated content and misinformation.

https://www.vice.com/en/article/v7bdba/ ... edia-apart

[boredbird]

The irony being how much WP has been used to train LLMs

Volunteers who maintain the digital encyclopedia are divided on how to deal with the rise of AI-generated content and misinformation.

https://www.vice.com/en/article/v7bdba/ ... edia-apart

The funniest part is that by vetting the AI content they are training it again. For tech billionaires. For free.

[suckadmin]

The irony being how much WP has been used to train LLMs

Volunteers who maintain the digital encyclopedia are divided on how to deal with the rise of AI-generated content and misinformation.

https://www.vice.com/en/article/v7bdba/ ... edia-apart

The funniest part is that by vetting the AI content they are training it again. For tech billionaires. For free.

making the dead internet theory a reality

Another example that bugs me is that I used to rely on Google image search which was an earlier example of ML but now the discover button mostly returns pinterest hits which usually have no info about the image. Fortunately there's still tineye.com.

[journo]

For a few years Wikipedians on talk pages have been showing resignation that a sufficiently funded Google-Gemini-powered encyclopedia would, by merely existing, outperform them. Specifically outperform them in their nominal goal of "taking Google material and arranging it into encyclopedia articles without original research".

There's a chance Google might even fund such a thing and make their own Wikipedia. They are already replacing Wikipedia's featured snippets on Google search with their own AI results that is, in theory, much less biased than Wikipedia editors.

Also, doesn't Wikipedia already rely on AI to keep out vandals and bad sources? I thought that was their primary technology for that. Next up, writing articles. Then they could use that first technology, with a function to web search sources, to clean up the second generative AI one.