It doesn’t matter how bad a reputation it may have had. Wikipedia, even though many insist on dismissing it as a place of reference due to the inaccuracies and falsehoods that can be found among its paragraphs, continues to be one of the main sources of information on the Internet for many people. Its vast catalogue, which includes entries on any historical event, discipline or public personality, is too extensive not to come across a link to Wikipedia every time we search for something online.
That is why it was not surprising that, with the invasion of artificial intelligence chatbots, junk content generated by AI (or AI slop, as some call it), also reached this platform.
Wikipedia is possibly one of the most famous knowledge and free software projects in the world, and that is why it has many defenders who, despite its inevitable flaws and gaps, are willing to collaborate voluntarily with their unpaid work. to maintain and care for one of the greatest exponents of solidarity collaboration on the Internet.
It is in this context that a group of volunteers has decided to create the «WikiProject AI Cleanup«, a team that reviews and eliminates false or misleading content created by artificial intelligence that is also beginning to reach the free encyclopedia. The AI Cleanup project is “a collaboration to combat the growing problem of poorly written and unsourced AI-generated content on Wikipedia.”
Anyone is invited to join or participate in the project, in which at the time of writing this there are about 78 people registered according to the Wikipedia page, a very small number taking into account the dimensions of Wikipedia.
As they explain in their entry, the project has three main objectives:
- Identify text written by AI and verify that it follows Wikipedia policies. Any unsourced and probably inaccurate statements should be removed.
- Identify AI-generated images and ensure appropriate use.
- Help and track editors who use AI and may not realize its shortcomings as a writing tool.
Furthermore, they emphasize that their perspective is not one of absolute rejection of artificial intelligence, but of the integration in a responsible way: «The purpose of this project is not to restrict or prohibit the use of AI in articles, but to verify that its result is acceptable and constructive and, if not, correct or eliminate it,” they indicate.
Easily recognizable (for now)
As some members of the project have said, it is usually quite easy to recognize when a text has been written by AI. This is due to the well-known writing style of bots like ChatGPT, which tend to always use the same structure and tone.
What is more difficult is identifying false references. Wikipedia articles, similar to academic articles, strictly use references to be transparent about the sources of information. ChatGPT is perfectly capable of inventing names of fake journals or scientific articles, so to check the veracity of a reference, reviewers have to click or manually enter the link provided.
ChatGPT may link to a real post, but it may not include exactly the information the bot claims. If we also talk about magazines or sources written in other languages, the reviewers’ work becomes more complicated.
They have also created a page in which you can see all the fake photographs or images, created by AI, that have been located on Wikipedia, and removed from the articles in which they were included. Also included are those that, after being reviewed, have been accepted because they are judged to offer a correct representation of the object or topic to be discussed, even though they have been generated by AI.
With all this, the defenders of Wikipedia are not discouraged by the possible boom of false information entering AI, and they are organizing to review little by little the content that has been published. labeled as possibly manipulated.
