.@OpenAI 的 Lukasz Kaiser 解释为何“泛化能力”仍然是核心关键： “我对泛化着迷。我一直认为，这是整个机器学习领域，以及理解智能的关键主题。预训练有点不一样……它不一定能提升泛化能力，它只是使用了更多知识。” “我们不断推动模型。模型会学习我们教它周边的东西。但它们依然有局限，因为它们不生活在物理世界里，因为它们的多模态能力还不够好，因为推理仍然非常早期，我们的做法里还有很多 bug。” “有人说，这就像在雾中高速驾驶——你永远不知道自己离目标有多远。” . @lukaszkaiser of @OpenAI on why generalization is still the crux: “I am fascinated by generalization. I always thought this was the key topic in machine learning in general and in understanding intelligence. Pre-training is a little different… it doesn't necessarily increase generalization. It just uses more knowledge.” “We push the models. They learn things that are around what we teach them. They still have limitations because they don't live in the physical world, because they're not very good at multimodal, because reasoning is very young and there's a lot of bugs in how we do it yet.” “Someone said, it's like driving fast in a fog. You never know very clearly how far or close you are.”