> In simple terms you can never really “trust” chatgpt’s answers because it’s just guessing the answer that looks right, not applying structured reasoning like

> In simple terms you can never really “trust” chatgpt’s answers because it’s just guessing the answer that looks right, not applying structured reasoning like humans do. Here's an argument otherwise, that they really are developing some sort of model: https://the-decoder.com/stochastic-parrot-or-world-model-how-large-language-models-learn/ https://the-decoder.com/stochastic-parrot-or-world-model-how...