I am paying for Chat GPT 4 using it applying for jobs. During my experience today, I encountered several frustrating instances where ChatGPT-4 “forgot” content, mishandled tasks, or failed completely. This raises the question: why does this happen, and how can these issues be addressed to improve user trust and satisfaction?
What Went Wrong?
-
Premature Assumptions
When I provided a URL to discuss a job posting, ChatGPT incorrectly stated that the job was no longer available—without verifying the information or giving me a statement why this conclusion had been taken. Of course the job was available. This led to unnecessary confusion and wasted time. -
Content Omission
In a task where I asked ChatGPT to rewrite my CV, an entire section—Education & Certification—was removed without explanation. When I pointed it out, the section was restored partly, leaving away important details, just after pointing these out 3 times they had been restored. -
Misleading Formatting
I requested that changes to the CV be shown in bold for clarity. ChatGPT marked unchanged content in bold, which misrepresented the updates and caused additional confusion. -
Repetition of Errors
Despite identifying earlier mistakes, such as content removal or formatting inconsistencies, ChatGPT repeated similar issues. For example, bold formatting continued to appear incorrectly, even after I flagged it. -
Overpromising and Underdelivering
ChatGPT often assured me that outputs were aligned with my requests or the job requirements, but closer inspection revealed gaps. Assurances were given without evidence, so ChatGPT wrote that changes had been done, but they were in fact not done.
How Can It Be Improved?
-
Enhanced Verification
Introduce automated checks to compare the output against the original request, ensuring no critical details are omitted or misrepresented. -
Persistent Memory
Develop user-approved memory capabilities to retain key details and prevent repeated errors across tasks in the same session. -
Task-Specific Protocols
Implement better logic for handling multi-step instructions, such as clear checks for formatting, alignment with inputs, and inclusion of all required content.
4. Greater Transparency
ChatGPT should flag potential limitations or ask for clarification if there’s ambiguity in a task, rather than assuming and delivering incomplete results. When I ask to be more precise, ChatGPT writes, that it will be done from now on… but instead introduced additional errors, which exacerbated the issue.
I hope these suggestions help developers improve the quality of ChatGPT’s responses, ensuring it meets the expectations of paying users.
As a user, I’m seeking insight into what steps OpenAI can take to ensure ChatGPT delivers precise results.