Last week, I shared a few revealing questions that I think showed the true flexibility of the latest ChatGPT model. ChatGPT’s creators at OpenAI admit the model can return biased or outright incorrect information, so it’s best to not depend on it for any planning or final decision-making. And of course, I independently warned this is not the best tool for highly suggestible people.
Still, it is difficult to downplay the impact the model has had just over the past month alone.
And yet, after using the model for several weeks, I noticed the model never does one thing that people ordinarily do…ask follow-up questions.
Notice of Non-Affiliation and Disclaimer: As of the publication date, we are not affiliated with, associated with, authorized with, endorsed by, compensated by, or in any way officially connected OpenAI, or their owners, subsidiaries or affiliates. The names OpenAI and ChatGPT, as well as related names, marks, emblems, and images are trademarks of their respective owners.
External Links: Links to external web pages have been provided as a convenience and for informational purposes only. Unboxing Tomorrow and Voxidyne Media bear no responsibility for the accuracy, legality or content of the external site or for that of subsequent links. Contact the external site for answers to questions regarding its content.
Restoring The Art of Inquiry
Obtaining information is nice, but retaining the most salient points was really the objective here. For example, in a recent conversation with the chatbot, I wanted to learn about the open-source Kubernetes project and how to get started using it.
I started by asking how Kubernetes might relate to leased cloud services. The response from ChatGPT included a few unfamiliar concepts like containerization. As a hardware-oriented person, this is outside my domain. But rather than set the stage for the typical encyclopedic reply from the chatbot, I requested a small addition to ChatGPT’s next response…
For a first attempt, this open-ended question from the model was an impressive response. And this one actually had me wondering what kind of answer it would identify as a correct one.
Conclusions
I think this is the way I plan to interact with the model in the future, as this seems to better utilize its capabilities as a conversational chatbot. With a few other out-of-the-box ways to ask questions, you may get some very memorable responses.
References
| [1] | B. R. Mayes, “Seven Revealing Replies from ChatGPT,” Voxidyne Media, 26 Dec. 2022. [Online]. Available: https://unboxing-tomorrow.com/seven-revealing-replies-from-chatgpt/. [Accessed 12 Jan 2023]. |
| [2] | OpenAI, “chat.openai.com,” OpenAI, 12 Jan 2023. [Online]. Available: https://chat.openai.com/chat. [Accessed 12 Jan 2023]. |
Important Notice: This article and its contents (the “Information”) belong to Unboxing-tomorrow.com and Voxidyne Media LLC. No license is granted for the use of it other than for information purposes. No license of any intellectual property rights is granted. The Information is subject to change without notice. The Information supplied is believed to be accurate, but Voxidyne Media LLC assumes no responsibility for its accuracy or completeness, any error in or omission from it or for any use made of it. Liability for loss or damage resulting from any reliance on the Information or use of it (including liability resulting from negligence or where Voxidyne Media LLC was aware of the possibility of such loss or damage arising) is excluded.
