WHEN: 27th March 2024, 6:30 pm

WHERE: Fab Lab 1, str. Rândunica 2, corp TEX6, 2nd floor, sala 2

In this in-person talk, we will explore the intricacies of the latest generation of large language models (GPT-4). He will shed light on how they work behind-the-scenes, delving into strategies such as fine-tuning and retrieval augmented generation, and will address the challenges these models face, including slow response times and restrictions due to context window sizes.

While attending this talk, you will not only gain insight into the theoretical aspects of large language models but will also be able to see a live demo — building the world’s simplest chatbot, using only Python, Jupyter Notebooks, and Azure OpenAI.

By the end of the talk, you’ll have a clearer understanding of:

– The mechanisms that allow language models to provide current information.
– Approaches to update AI models without exhaustive retraining.
– Practical methods for fine-tuning language models to suit specific interaction styles.
– The role of retrieval augmented generation in enhancing the model’s responses with the most relevant information.
– Approaches to mitigate common limitations of large language models.

This session is designed for those looking to practically apply their knowledge, equipping attendees with the fundamental skills necessary to begin developing apps based on large language models using Python and the Azure OpenAI APIs