I really have to disagree with the broad brush statement.
Wikipedia is an incredible resource and yes given its model it is maintained and moderated by volunteers, which has meant during particular global events the Wikipedia articles associated are subject to potential changes representing misinformation that requires the volunteers to moderate through.
If systems are scraping this information and they don't have processes to correct itself then that's not the fault of Wikipedia, but the fault of those tools blindly scraping the data and serving it.
Wikipedia is an incredible resource. Wikipedia doesn't poison global knowledge, it is a knowledge source that changes over time and is subject to bias like anything. If you scrape this information and use it as gospel without assessing its bias or updating it based on changes then the LLM's are knowingly serving information that they know might be incorrect.
I should add that the information you get out of an AI tool is also based on your prompt, if you ask it for information relating to Gaza, you should ask it for an independent view that covers multiple knowledge sources to help mitigate these issues.
I agree that clearly there are areas with factual inaccuracies on Wikipedia and I'm not sure on what the solution is to resolve it, I also don't see this article coming up with a solution either. Okay move to a different platform, but as that platform grows it will become a similar target for the exact issues mentioned.
if wikipedia was opaque to ai bots, there would be less motivation to use wikipedia as a poison waterhole.
this isnt wikipedia poisoning global knowledge, this is ai being attacked BWO a convienient vector, with a "who cares" blast radius
I really have to disagree with the broad brush statement.
Wikipedia is an incredible resource and yes given its model it is maintained and moderated by volunteers, which has meant during particular global events the Wikipedia articles associated are subject to potential changes representing misinformation that requires the volunteers to moderate through.
If systems are scraping this information and they don't have processes to correct itself then that's not the fault of Wikipedia, but the fault of those tools blindly scraping the data and serving it.
Wikipedia is an incredible resource. Wikipedia doesn't poison global knowledge, it is a knowledge source that changes over time and is subject to bias like anything. If you scrape this information and use it as gospel without assessing its bias or updating it based on changes then the LLM's are knowingly serving information that they know might be incorrect.
I should add that the information you get out of an AI tool is also based on your prompt, if you ask it for information relating to Gaza, you should ask it for an independent view that covers multiple knowledge sources to help mitigate these issues.
I agree that clearly there are areas with factual inaccuracies on Wikipedia and I'm not sure on what the solution is to resolve it, I also don't see this article coming up with a solution either. Okay move to a different platform, but as that platform grows it will become a similar target for the exact issues mentioned.