Decoding ChatGPT – Understanding the Algorithm & Its Application in Finance

ChatGPT is an AI-powered conversational chat interface that understands and responds to natural human language. It is based on the GPT-3 family of large language models. GPT-3, or the third generation Generative Pre-trained Transformer, is a machine learning neural network model that specializes in generating human-like text. The model was trained using Reinforcement Learning from Human Feedback (RLHF). RLHF employs a reward/punishment system to train AI. A desirable action is rewarded, whereas an undesirable action is punished. Thus, the AI processes a large subset of data and learns through trial and error, just like humans. Banks and other financial institutions have taken note and are utilizing it for stock market analysis, client service, and financial calculations. In this session, I will explain the complex algorithm underlying ChatGPT and discuss potential applications in the finance industry.

Who is this presentation for?
Technical professionals, Finance Professionals

Prerequisite knowledge:
Basics of Generative AI

What you'll learn?
Generative AI, ChatGPT

I have over 8 years of work experience in areas of Finance and Data Science – specifically fraud detection, Anti-Money Laundering, HR Analytics, Credit Card and Mortgage Risk. In my current role, I lead a Data Science team at Mastercard that develop AI-based product for Mastercard global. I have authored papers and patents in this finance domain and have delivered several talks at prestigious forums. I was also featured as one of 8 female inventors on closing the gender gap in innovation. Before joining Mastercard, Debasmita was at J. P. Morgan & Chase where she was part of the Innovation Team in their consumer banking domain.