LLMs in Data Analytics: Can They Match Human Precision?

Explore the capabilities and limitations of Large Language Models (LLMs) in data analytics through this insightful 40-minute conference talk by Gerard Kostin, Director of Data Science at DataGPT. Delve into DataGPT's technology stack, examining how LLMs process extensive datasets and integrate with advanced algorithms for modern data interpretation. Learn about the challenges of LLMs, particularly hallucinations, and discover DataGPT's strategies for mitigating these issues. Gain valuable insights into efficient data processing techniques for accurate and consistent analytics, beneficial for both engineers and analysts. The talk covers topics such as LLMs as statistical machines, roadblocks in LLM usage, and selecting appropriate tools for specific tasks, concluding with a Q&A session to address audience queries.

- Introduction
- Welcome to the World of LMs
- LLM as a Statistical Machine
- The Roadblocks: Challenges in Using LLMs Hallucinations
- How do we deal with Hallucinations in DataGPT
- The Right Tool for the Right Job
- Q&A Session