How ChatGPT Works: The Model Behind The Bot | Towards Data Science

The GPT-3.5 architecture-based deep learning language model ChatGPT produces human-like responses by processing input text tokens using transformers and attention mechanisms that have been trained on a sizable amount of data. 

ChatGPT is a language model that utilizes a deep learning algorithm to generate human-like responses to text-based inputs. It was trained on a vast amount of text data from the internet, allowing it to understand and replicate human language patterns and nuances. The model is based on the GPT-3.5 architecture, which employs transformers and attention mechanisms to analyze and generate text. ChatGPT works by breaking down the input text into tokens, which are then processed by the model to predict the most likely response. The bot can also learn from its interactions with users, improving its responses over time. 

Source: https://towardsdatascience.com/how-chatgpt-works-the-models-behind-the-bot-1ce5fca96286