Omg https://arxiv.org/pdf/2304.11062.pdf this machine learning paper is absolutely crazy. It provides a way to have a context length of 1M+ tokens. Just to put it into context, GPT-4 with 32k tokens looks child play against the 1M+ tokensize...