DataHour: Reducing chatGPT Hallucinations by 80%

In this session, we will delve into the techniques and strategies employed to enhance the accuracy of language models, mainly focusing on reducing hallucinations encountered in ChatGPT, an advanced conversational AI developed by OpenAI. Join us as we explore the cutting-edge approaches used to mitigate the issue of inaccurate or misleading responses in language generation systems.

The journey towards enabling useful LLM Apps to be built involved three key areas:

1. Observability: What questions are my users asking?
2. Tuning: What changes can I make to improve results - add more data, finetune embeddings, improve search, change prompt, etc?
3. Testing: Did my changes actually improve results?

Prerequisites:Enthusiastic and eager to learn data science professionals and students.

Krrish Dholakia is a highly accomplished professional currently serving as the Co-Founder and CEO of Berri AI (YC W23), an organization based in the San Francisco Bay Area. With a strong educational background, Krrish holds a Bachelor's degree in Computer Science from the Georgia Institute of Technology. During their time at Georgia Tech, Krrish actively participated in various activities and societies, including being the Director of Product at Bits of Good, where they ensured project completion and maintained strong client relationships. Additionally, Krrish served as the Director of Technology at Startup Exchange, leading web development initiatives and fostering technological innovation.