Understanding ChatGPT

ChatGPT, the free chatbot released in November 2022 by artificial intelligence (AI) research company OpenAI, has taken the internet by storm. In its first months of existence, ChatGPT inspired users to imagine a host of use cases for the model, including using ChatGPT to negotiate parking tickets, make workout plans, and even create bedtime stories for children.

Below, we explore this new technology and how it could eventually make money. Put simply, ChatGPT is an AI model that engages in conversational dialogue. It is an example of a chatbot, akin to the automated chat services found on some companies’ customer service websites.

It was developed by OpenAI, a tech research company dedicated to ensuring that artificial intelligence benefits all of humanity. The “GPT” in ChatGPT refers to “Generative Pre-training Transformer,” referring to the way that ChatGPT processes language.

What sets ChatGPT apart from chatbots over the last several decades, however, is that ChatGPT was trained using reinforcement learning from human feedback (RLHF). RLHF involves the use of human AI trainers and reward models to develop ChatGPT into a bot capable of challenging incorrect assumptions, answering follow-up questions, and admitting mistakes.

To put ChatGPT to the test, Investopedia asked it to “write a journalistic-style article explaining what ChatGPT is.” The bot responded that it was “designed to generate human-like text based on a given prompt or conversation.” It added that, because it is trained on a data set of human conversations, it can understand context and intent and is able to have more natural, intuitive conversations.

As mentioned, there are numerous potential uses for ChatGPT. They range from more direct, chatbot-type functions to much more obscure applications, and it is likely that users will explore a host of other possible ways to utilize this technology in the future, including in search engines.

While chatbots have existed for many years, ChatGPT is viewed as a significant improvement on the intelligibility, fluidity, and thoroughness of prior models. One demonstration of the sophistication of ChatGPT provided by OpenAI includes a prompt that was designed to trick the bot: asking about when Christopher Columbus (supposedly) came to the United States in 2015. ChatGPT’s response easily avoided the trap, clarifying that while Columbus did not come to the U.S. in 2015, it can posit some of the ways he may have reacted to his visit if he had.

Investopedia