Hosted by the Interactive AI Centre for Doctoral Training
Abstract: We will be looking at how recent developments in Large Language Models (LLMs) are being adopted by commercial interests. Beginning with what companies/products are out there and what problems they claim to address.
From here, we will specifically investigate Llama Index, a software package empowering users to connect LLMs to their own data and to converse with their data. Assuming the audience already has first hand experience using LLMs like ChatGPT, we explain how Llama Index is able to connect LLMs to a custom knowledge base without any fine-tuning through vector search.
Once we understand how this product is able to connect LLMs to our own data, we will play with a toy example in breakout groups. The goal of this will be to obtain responses from the application that run counter to its intended purpose.
Through interaction, we hope to discover the limitations of this technology, while considering the potential ramifications should these limitations be abused in commercial settings. Together, we will discuss our thoughts and opinions on the impact of this technology being rapidly adopted by commercial interests.