AI models conflict on history
Different AI models can provide conflicting accounts of the same historical event. Privacy advocate nevscape attributes this possibility to private training data that is inaccessible for external verification. Users are increasingly replacing Google with chatbots, which means the entity controlling model training effectively determines how history is recorded.
Transcript (en)
even AI, the models, like, to me, it's concerning that it's private companies that develop them, not from like a, again, it's the same thing with the algorithm. Like, there's a big belief in decentralization, right? Like, if you've privately trained this model, we don't know how it's been trained. We don't know the things that make the thoughts. And that's how you can rewrite history. Like, information sources is how history is written. Who writes the books? Like, history is written by the victors. So it's like how people get their information. Instead of going to Google, you go to either Grok or Claude or Chatea Petit. In a weird, dystopic future, what if they all say a different thing happened? Which one is true? How was it trained? If you enjoyed this, be sure to hit that like button and follow to stay up to date with what's going on in the world and everything else memes and markets. Peace.
