Don't be careless with ChatGPT

#chatgpt #ai

Written by Anders Marzi Tornblad

Today, I saw a LinkedIn post that gave some advice on how to create prompts for ChatGPT. The author suggests a lot of different prompts to get the most out of the model, but most of the suggestions are pretty risky. I think it is problematic to talk about ChatGPT and similar technologies too carelessly, for a couple of reasons.

New unstable technology

Since ChatGPT is a very new technology that is constantly changing, any prompt that works in one way today will give different results tomorrow. AI is the type of technology that will most likely never be stable, but will continue to evolve rapidly.

Some of the prompts suggested in the LinkedIn post simply will not work. A Plagiarism Checker, for example, will definitely hallucinate non-existing works, so even very original works will get false flagged. ChatGPT simply has not access to every single written work.

The suggested Advertiser prompt will only generate bland non-creative texts without any novelty. It will simply create "more of the same". The same goes for Stand-up Comedian, Motivational Coach, Debater, Rapper, and so on.

Then there are the more dangerous ones, where AI hallucinations provide safety risks or financial risks, like Doctor, Accountant, Investment Manager and Self-help Book.


While ChatGPT has shown its capability in generating human-like responses and is a powerful tool in a wide variety of applications, it's crucial to understand its limitations.

ChatGPT doesn't have a consciousness or a genuine understanding of the content it generates, nor does it have the ability to verify the accuracy of the information it provides. It's trained on vast amounts of text data and relies on patterns within that data to generate responses. Consequently, it may output inaccurate, inappropriate, or misleading information, making it an unreliable source for critical or sensitive contexts.

Excessive reliance on, or misunderstanding of, ChatGPT's capabilities could lead to spread of misinformation, and flawed or dangerous decision-making. Always use this tool with critical thinking, and corroborate its outputs with other reliable sources when accuracy is paramount.

View this article on LinkedIn