I recently decided to explore artificial intelligence writing through OpenAI’s ChatGPT. ChatGPT is “Chat Generative Pre-trained Transformer” and is a complex machine learning model that is able to carry out natural language generation. The company describes it as, “…a model…which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.” I am a physician and not a computer scientist or engineer, so I won’t speak to any technical aspects of programming. But, my exploration of the tool shows promise and limitations from a medical perspective.

From what I have seen, the program displays excellent grammar, and content can be tailored by narrowing down and specifying requests. The creative part of using the program will be in how to guide it to give one the exact content that one envisions. It is extremely fast and can generate content within seconds. It will improve the efficiency of the creation of written content exponentially. However, there are limitations.

I thought that it would only be appropriate for ChatGPT to express these limitations on its own:

“As a language model, ChatGPT has many limitations that can affect its performance and accuracy. Some of the main limitations include:

1. Biases and inaccuracies: ChatGPT is trained on a large corpus of text, which can contain biases and inaccuracies that are reflected in its responses. For example, if the training data contains stereotypes or discriminatory language, ChatGPT may reproduce these biases in its output.

2. Limited knowledge: While ChatGPT has access to a vast amount of information, it is still limited by its training data. It may not have knowledge of certain topics or may provide inaccurate or incomplete information in some cases.

3. Lack of contextual understanding: ChatGPT may struggle to understand the context of a conversation, which can lead to nonsensical or irrelevant responses. It may also have difficulty understanding sarcasm, irony, or other forms of figurative language.

4. Inability to understand emotions: ChatGPT does not have the ability to understand emotions, which means it may not be able to respond appropriately to messages that convey a certain tone or mood. This can lead to misunderstandings and miscommunications.

5. Lack of common sense reasoning: ChatGPT can sometimes provide illogical or nonsensical responses to questions that require common sense reasoning. For example, if asked “Can you jump to the moon?”, ChatGPT may provide an answer that is technically correct but not helpful or relevant to the question.

It’s important to keep in mind that while ChatGPT is a powerful tool for generating text, it is not a substitute for human expertise or judgment. Users should always critically evaluate its responses and use their own judgment when making decisions based on its output.”

These limitations correlate with what I have found. From a medical standpoint, the content derived from ChatGPT is relatively general and doesn’t seem to give one the detail of a scientific review or research paper (at this time). However, it will provide nice, solid information for the public. But, as mentioned above, biases in the information that it pulls from its sources need to be taken into account. The information should still be reviewed for accuracy by a medical professional prior to publishing. I have noticed that it may reference incorrect links for requested resources, so it may not be obtaining the most accurate information from its database.

Nevertheless, under the proper guidance of healthcare professionals, ChatGPT (and artificial intelligence programs like it) is a tool that can produce content for teaching and education with a degree of efficiency that is unprecedented. Also, if used and curated correctly, the healthcare industry has a powerful ally that can aid in producing volumes of content to combat the infodemic and dispel medical misinformation.  ChatGPT and other artificial intelligence programs of its kind are important programs to explore and become familiar with because these technologies will only advance and understanding them will be the best way to be able to utilize and guide them.

How do you feel about using artificial intelligence programs with writing?

Do you think there is a role for artificial intelligence programs in generating written medical content for the public or healthcare industry?