>> LLMs have no knowledge of the underlying reality > They have no common sense & they can't plan their answer If his arguments are entirely based on this, the

>> LLMs have no knowledge of the underlying reality > They have no common sense & they can't plan their answer If his arguments are entirely based on this, then it's not fully correct: - GPT style language models try to build a model of the world: https://arxiv.org/abs/2210.13382 https://arxiv.org/abs/2210.13382 - GPT style language models end up internally implementing a mini "neural network training algorithm" (gradient descent fine-tuning for given examples): https://arxiv.org/abs/2212.10559 https://arxiv.org/abs/2212.10559