The Enigma of LLMs: Creativity, Compositionality, Pluralism, and Paradoxes

Explore the complexities and paradoxes of Large Language Models (LLMs) in this thought-provoking lecture by Yejin Choi from the University of Washington and NVIDIA. Delve into the ongoing debates and divisive reactions within the research community regarding LLMs' capabilities. Examine key concepts such as creativity, compositionality, and pluralism in the context of these powerful AI systems. Investigate the emerging field of "Science of LLMs" and its importance in complementing the engineering aspects. Gain insights into the current limitations of our understanding of LLMs and the open research questions that continue to challenge experts in the field. Engage with potentially confusing findings and paradoxes surrounding generative AI, encouraging a more nuanced and open-minded approach to studying these enigmatic systems.

The Enigma of LLMs: on Creativity, Compositionality, Pluralism, and Paradoxes