GPT-3, Generative Pre-trained Transformer 3, is an autoregressive language model produced by OpenAI. It uses deep learning to produce what appears to be human-written text. It was trained on 175 billion machine learning parameters. GPT-3 started with four different iterations of GPT-3, Ada, Babbage, Curie and Davinci. Ada being the fastest, most afforable, lowest performing model, and Davinci being the slowest, most expensive, yet highest performing, with Babbage and Curie ranging inbetween. OpenAI has since upgraded and refactored these models and these demos utilize the model 'gpt-3.5-turbo' and have newer versions are available via OpenAI's online console. In this example, two stages of commands are passed to the model, the first asks for an essay outline on the user provided topic, this outline is then broken down into individual commands, which asks for a paragraph focused on a stage of the previously generated outline. The more information you provide the tool, the more accurate the response will be.

OpenAI's ChatGPT utilises GPT-3.5 which trained on data gathered prior to 2021, so may not have knowledge of current affairs past that cut off. The results generated may also be used in training future models.