Source record
POV: You've been arguing with ChatGPT for 30 mins over something simple 😭 Here's the fix that'll save your sanity! #ChatGPT #AITips #TechHacks #ProductivityTips
Do you ever find yourself going to ChatGPT. For what seems like a simple task, only to find yourself 30 minutes later so frustrated that it hasn't been able to do this super simple thing that you find yourself telling it that you hope there's a special place in hell just for large language models that don't do what their users ask them to do?
Yeah, me neither, but I hear this happens to people, so I just want to share some tips I've found to avoid these super frustrating situations with large language models and get the output you want quicker so you're not spending 30 minutes doing something that you could have done yourself in 10. What I found is this almost always happens because the large language model is confused about what you want from it.
There's usually some ambiguity in the initial prompt or one of your follow up prompts, and so it doesn't quite give you what you want. And so of course you kindly explain that to it. Hey, could you please do it this way instead?
And maybe you go back and forth this way a few times. The problem is that all the mistakes it made and all the ambiguity you introduced previously is still in the conversation, and the large language model is looking at that entire conversation every time it responds. So if you want the large language model to stop making mistakes, you need to remove The mistakes.
So if this is a fairly short conversation, just start over, take whatever lesson you can from the bad output, and write a better first prompt. If it's a longer conversation, find the prompt where it went wrong and just edit that prompt. You can do this at least with ChatGPT, and then just start the conversation from there.
The one thing you always want to avoid is asking the model to fix the same mistake more than once. What almost always ends up happening in this case is every output gets a little bit worse than the output before. And that's where you get in this frustrating cycle where you end up saying things that you would never say to your worst enemy.
To a large language model. Not that I would know.
AI visibility work depends on matching the right model, search surface, or answer format to the business outcome being optimized.
Map each AI/search surface to a task: answer extraction, public copy, citation mining, checkout readiness, media generation, or customer-intent capture.
This source is connected to AI visibility and answer readiness.
For what seems like a simple task, only to find yourself 30 minutes later so frustrated that it hasn't been able to do this super simple thing that you find yourself telling it that you hope there's a special place in hell just for large language models that don't do what their users ask them to do?