Post #3534484

Gonna need a source for that, boss– Wikipedia’s intense scrutiny of sources and requirements for reliable citations are actually one of the reasons that Sanger started his malformed crusade. If there’s actual, provable lies about a notable person in the encyclopedia, then there should be actual, provable truths to combat it; and any Wikipedia editor can update the article in question to correct the record. If an edit war emerges, a community discussion can take place wherein the person in question can have their say. Wikipedia isn’t the wild west, and any reasonable argument that it is died twenty years ago. –but even if those two things weren’t true–it’s literally impossible to remove misinformation from AI models. I’m not saying that to be dramatic or overstate the problem. When a model is trained with misinformation, the misinformation becomes a part of the model; the entire corpus of everything it was trained on is baked into the neural network on a fundamental level, and humans can’t manipulate it manually. Which means you can’t remove any datapoint from the model without excising it from the training data and then retraining a whole new model. So now not only are you drinking your dog’s urine, you’re claiming that the tap water is too yellow. Even if your assertion’s true, your alternative is demonstrably worse.