Wikipedia has banned using AI to write or rewrite articles in English. Human knowledge begins to raise barriers
The English version of Wikipedia has just banned articles made with AI. In the last update of their guidelines are clear: content generated with language models violates content policies. The largest encyclopedia on the internet positions itself as a refuge for content created by humans. AI no thanks. The ‘AI yes or AI no’ debate has been going on for a while generating tension on Wikipedia and they have finally opted to support human content with an overwhelming majority 40 to 2. The new restriction imposed reads as follows: “Text generated by large language models (…) often violates several of Wikipedia’s fundamental content policies.” Those fundamental policies What it refers to are the neutrality of the content, verifiability and that the content cannot be original research, but must be attributed to reliable sources. With this change, editors are prohibited from using LLM “to generate or rewrite article content.” Two exceptions. Wikipedia contemplates two scenarios in which the use of AI is allowed: Basic style suggestions and corrections, as long as the LLM does not introduce its own content. They warn that it must be used with caution since LLMs tend to “go beyond what is asked of them and alter the meaning of the text.” Translation of articles into other languages, as long as it is reviewed by a person competent in the two languages involved. Here it is important to note that Wikipedia has already had dramas in the past because of AI translations. Why is it important. Wikipedia has positioned itself as a repository of genuinely human content in an internet that is flooded with artificial content. At a time when distinguish the authentic from the synthetic is increasingly difficult, the largest encyclopedia in the world chooses to rely on human authorship as a guarantee of reliability. There is certainly something ironic and that is that Wikipedia rejects AI, but AI continues to draw on Wikipedia to provide answerscausing them to lose clicks and saturating your servers. AI generated vs human made. Until recently we thought that the solution was flag artificial content on platforms with the classic ‘AI’ label, but we are already at a point where it is more valuable and useful to highlight the opposite: that it is made by humans. The advancement of image generation tools and the amount of texts made with AI are overwhelming, to the point that an anti-AI current is emerging; Some artists are starting to designing “badly” to differentiate itself from AI homogenizationthey have created extensions to return to the internet before ChatGPTthere is browsers that filter AI results and even ‘Not by AI’ badge has been created. The point is that it is a David against Goliath. The Etsy case. It is perhaps one of the most bloody cases of the flood of low-quality AI content. The platform that It was presented as a refuge for the authentic, today it is an AI market which also tries to pass itself off as artisanal. Ghibli-style portraits for 20 euros, profiles managed entirely by AI that say things like “I can’t wait to draw you”… Etsy allows content made with AI, but says you have to label it as such. Nobody does it. Proof that the label is no longer useful. A key detail. The last paragraph of Wikipedia’s guidelines is especially striking because it talks about possible sanctions for those who violate the rule, the problem is how they plan to detect who uses AI. Wikipedia admits that “some editors may have writing styles similar to those of large language models” and that “more evidence than mere stylistic or linguistic clues is needed to justify the imposition of sanctions.” We have no idea how they are going to do it, what we do know is that AI text detectors fail more than a fairground shotgun. Image | Wikipedia, edited In Xataka | The last barrier against AI is good taste. The problem is that an entire generation is growing up without developing it
