Junliang initiated the discussion about large language models (LLMs), which is based on the papers: “Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond” and “A Survey of Large Language Models”.
In this reading group, the development and evolution of LLMs were introduced, as well as the typical architectures of LLMs. The reasons of why LLMs are powerful were also summarized, including faster hardware, large training corpus, new training paradigms and new architectures.
Some open questions were discussed in the reading group:
- How can our research benefit from LLMs?
- What harms can LLMs cause?
- What will be the impact of LLMs on employment and job markets? How should society prepare for potential job displacement?
- How might LLMs change the way society communicates, learns, and accesses information? What are the potential cultural impacts?
Computational Social Science Reading Group Series: Interacting with Opinionated Language Models to Change Users' Views