Content area
Full Text
A Conversation with ChatGPT
ChatGPT is an artificial intelligence chatbot developed by OpenAI, an artificial intelligence research laboratory consisting of the nonprofit OpenAI Foundation and the for-profit OpeiiAI Limited Partnership. ChatGPT launched on November 23, 2022 and is built on top of OpenAI's GPT-3 family of large language models (LLM). It works by predicting the next most likely word in a sentence based on a prompt and previous chat history. While chatbots have existed for some time, the arrival of ChatGPT has been heralded as a significant watershed in the development of AT applications and has become among the fastest-adopted technologies of the Internet era.
There is significant disagreement over both the usefulness and transformative nature of LLMs. It was quickly noted that because this LLM was not connected to the Internet, it could not fact-check itself, leading ChatGPT-supported applications to "hallucinate"-generate plausible-sounding yet factually inaccurate information in response to verifiable questions. This led some observers to suspect that AI is not about to displace the millions of knowledge workers who every day generate text and visual information-at least not until connectivity is enabled. Additionally, there are very real concerns of bias incorporated into its coding and the likelihood that the technology will merely reflect humanity's own flaws without transending its limitations.
Yet in the longer term, legitimate criticisms have been levelled at ChatGPT and other LLMs, including generative AT tools more broadly, for their ability to cheaply massproduce content of dubious quality. In a world increasingly bloated by misinformation, the era of limitless, inaccurate content is very close at hand. Such a reality could significantly alter the information ecosystems in which much of humanity currently finds itself. Less apocalyptically, there are also discussions around the use-value of these tools and their capacity to complement, not replace, existing knowledge work. The ability to instantaneously write a legal document or briefing memo, to automate tasks such as email and coding-to draft an article for submission in the Journal-could radically alter productivity. It is still early, though the promise is great, if uncertain.
In many respects, AI is too "live" a topic to address with much rigor or seriousness in the pages of the Journal; that analysis will come in a later edition. But here we present...