AI Earthquake – ChatGPT Is Collapsing?

[mdarafat724]

In November of this year, shortly after OpenAI’s latest update, a heated discussion began on the official ChatGPT forum about what kind of experiences users had with the friendly robot after the latest update. As it turns out, the vast majority of them were very negative .

Let’s take a look at just a few of the hundreds of negative comments:

I've noticed the same thing, the quality of the answers has plummeted. When I ask her technical questions, she now likes to give a numbered list of generalized points, instead of just understanding my intentions and responding appropriately like a person would. I have to waste additional prompts to understand that I want a more direct and canada mobile phone numbers database engaging answer...

It's true, many users also report that ChatGPT version 4 answers questions much more generally than it did 2 months ago. It also "hides" more often with formulas like "I can't answer because..."

Particularly serious problems are noticed by programmers who used GPT to speed up their work:

Coding in ChatGPT 4.0 is getting harder with each passing week. Here's what I've noticed:

The model no longer seems to remember the conversation history.
History is not even fully preserved anymore; whole pieces are missing.
Code that is created and contains errors is suddenly truncated or completely changed.
I've made ChatGPT generate suggestions itself to avoid these errors, but it doesn't help at all. It's wrong every time.
Coding intelligence is getting worse. Even small errors keep repeating themselves, and the model is no longer able to solve them.
The model always wants to split the code into multiple pieces, so you end up spending a huge amount of time cutting and pasting, and new bugs appear.
And yet another user:

He ignores my instructions in the same message. He can't understand what I need. I have to keep repeating how I want him to respond to each message because he forgets and ignores the previous ones.

The worst part is that it's so SLOW and full of bugs. Every few messages it generates an error that repeats until the page is refreshed.

Other comments
Other users also report:

slower responses from ChatGPT4 than before
less accurate answers
forgetting the context
prompts (commands) that worked a month ago now give completely different results or don't work
Chat refuses to respond to more topics than before
It is also worth mentioning the "functionalities" added in the previous update, the so-called GPTs. These are assistants built into ChatGPT, which can be designed by the user. However, as reported by Internet users, GPTs do not exceed the intelligence of regular GPT4 regardless of the commands used .

I, a regular user of the software, have also noticed that Chat has become much worse at organizing information . It also refuses to apply required formatting. Commands that were previously performed without hesitation (e.g. creating a table) now most often result in refusal from the AI.

What does OpenAI say about this?
Very quickly, in response to the negative opinions that were being circulated on the ChatGPT forum , as well as through other channels such as the X platform, a statement from OpenAI itself appeared. We are pasting it below.



English translation: We've heard all your feedback about GPT4 becoming increasingly lazy! We haven't updated the model since November 11th and that's certainly not intentional. The model's behavior can be unpredictable and we're working to fix that

What does this comment mean to us, even though it doesn't say all that much?

That there were indeed problems. It is not a matter of users unlearning proper prompting, or that they have problems with the program. The latest ChataGPT updates have actually worsened the software's performance.