This blog is dedicated to exploring the behavior of large language models and generative AI through data analysis and experimentation.
Large language models (LLMs) are increasingly integrated into our daily lives (from social media to search engines). However, they are still largely black boxes — researchers and industry experts only scratch the surface in terms of understanding how these models actually work. It is important to probe these models to learn more about their behavior under different circumstances and scenarios.
Telling compelling stories using multiple sources of data is a challenging but rewarding feat. Each dataset on its own may only tell a very limited part of the story, but combined with datasets often lead to interesting results and conclusions.
Large language models such as ChatGPT are increasingly better at generating code. And sometimes, it’s faster (if not better) than I am at creating projects.
I’m Yennie, a machine learning engineer and AI researcher. I currently work at a healthcare startup, where I train large language models on data such as electronic health records. I have previously worked as an AI researcher, data scientist, and software engineer with the University of Oxford, Deeplearning.AI, OpenAI, the United Nations, and Microsoft.