Albert Einstein achieved fame for his development of theories of special and general relativity. He wrestled over the decades, however, with the task of producing a unified field theory that linked electricity, magnetism, gravity, and quantum mechanics. Whenever feeling stumped by obstacles in successfully completing that pursuit, he would state “I will a little tink.” English was not his native language.
November 2022 marked the release of a new open source, natural language processing tool by OpenAI called ChatGPT. This new chatbot is designed to simulate human conversation in response to prompts or questions (GPT stands for “generative pretrained transformer”). In January 2023, Nature reported on two preprints and a pair of articles published in the science and health fields that included ChatGPT as a bylined author. These articles and their nonhuman “authors” already have been indexed in PubMed and Google Scholar.
Two articles published in the March 2023 issue of the journal Lancet Digital Health discuss both the substantial ethical implications of this latest technology innovation as well as how it has potential to improve health care delivery. The publishing industry, for example, is at risk of undesirable consequences because when presented with a query, ChatGPT automatically will generate a response based on thousands of internet sources that may not involve any further interventions by the user. Scholarly manuscripts can be generated that also are accompanied by appropriate references. Ethical considerations abound concerning copyright, attribution, plagiarism, and authorship when AI produces academic text. These concerns especially are pertinent because when copy is AI generated, it currently is imperceptible to human readers and anti-plagiarism software. A significant concern is that the functionality of ChatGPT also has the capacity to cause harm by producing misleading or inaccurate content.
Presently, ChatGPT is available to use without cost, but OpenAI's leadership has affirmed that free use is temporary and the product eventually will be monetized. One commercial option for the platform could involve a form of paywall, which might entrench existing international inequalities in scholarly publishing. Although institutions in socioeconomically advantaged areas could probably afford access, those in low- and middle-income countries might not be able to do so, thereby widening existing disparities in knowledge dissemination and scholarly publishing.
Apart from worrisome aspects of this new innovation, the delivery of health care services may benefit from expanded capabilities associated with further development of ChatGPT. One potential application could be to generate discharge summaries, which may be left under-prioritized that result in delays in patients’ discharges from clinical facilities or insufficient discharge summaries. All things considered, from a much wider perspective it would be advantageous to devote a little more “tink” to contemplating how to optimize the potential contributions this exciting technology has to offer its users.