Huge quality discrepancy compared to online

[colingoodman]

I'm curious about the cause (and potential fix) on the large difference in quality between this tool and the online tool at openai.com. Here's an example dialogue:

User: Could you give me a brief explanation of Michel Foucault's Discipline and Punish?
Bot: Some response
User How does it differ from his earlier works?

Bot (openai.com):

Discipline and Punish is one of Michel Foucault's later works, and it differs from his earlier works in several ways. In his earlier works, such as The History of Madness (1961) and The Order of Things (1966), Foucault focuses on the emergence of modern forms of knowledge and the ways in which these forms of knowledge shape and are shaped by social, political, and cultural forces.

Discipline and Punish, on the other hand, shifts the focus from knowledge to power. In this book, Foucault examines the ways in which power is exercised and enforced in society, and the effects of this power on individuals and social groups. He argues that power is not just a matter of laws and institutions, but is also present in the everyday practices and interactions of individuals.
...

Bot (chatgpt-python):

Michel Foucault's Discipline and Punish is a book that explores the ways in which modern societies use punishment to control and regulate individuals. The book is divided into two parts: the first part examines the historical development of the prison system, while the second part explores the broader implications of this development for modern society.

The latter response seems confused by the user's prompt and fails to provide an answer to the question, something the former succeeds at. This discrepancy is consistent across a variety of prompts.

[AlexSua]

Hello Colin, hopefully everything will be solved in the next version. We are working on it. The next version will bring a lot of features.