Articles and Resources for ChatGPT

ChatGPT has been a major topic in learning technology discussion for the past few weeks. It introduces a web-accessible, friendly, interface to the existing GPT (Generative Pre-trained Transformer) system which is created and maintained by the OpenAI corporation.

ChatGPT generates text in response to a prompt. This text is generated based on a set of training data which the system has been exposed to in order to model what a text might look like. ChatGPT was also trained on a set of conversations and ratings by human trainers who looked at several potential ChatGPT answers for a given prompt.

It is important to note that the text GPT generates is based solely on the structure of text and so it has no inherent understanding of any concepts. It cannot know anything about heart disease, only what words and phrases are commonly associated with the words “heart disease” in the training data it has been given. This means that it may generate perfectly correct and helpful text or it may generate plausible but completely incorrect text. Many people have shared that ChatGPT has generated citations for its work, but these referenced works that do not exist, or that do exist but do not relate to the words being included. This is because ChatGPT understands the form of a citation, but has no concept of a citation, it’s simply a thing it has seen in the training data.

Many people have been writing about ChatGPT, particularly in the context of higher education. Several of these are gathered here:

January 10 Updates:

January 11 Updates:

January 12 Updates:

January 13 Updates:

January 16 Updates:

January 17 Updates:

January 18 Updates:

January 24 Updates:

January 25 Updates:

January 31 Updates:

February 3 Updates:

February 6 Updates:

As the discussion evolves we will update the collection here.