Large Language Models (LLMs): Representation Matters, Low-Resource Languages and Multi-Modal Architecture

Large Language Models (LLMs) have recently burst into popularity via contemporary chatbots and are being considered for information retrieval as well as reasoning tasks by ordinary citizens. Examples from ongoing research are presented in an attempt to highlight the chasm between high and low-resource languages. The importance of representation as well as multi-modality, well known in the research community for decades, is highlighted and brought to focus. We argue that those pillars are likely key to making the new generation of systems more powerful, usable, accessible and utile for all.