I’ve really been using gpt 4.0 since day one and it keeps getting stupider? I can’t understand why? While it was very beautiful at first, it has now turned into something that is not worth the money. A lot of new features were added, but it became stupid. I want my old smart GPT 4.0 back
27 Likes
I agree. I have been using ChatGPT since it started and its answers are becoming more and more ridiculous. It doesn’t even TRY any more. It gives completely false responses and when you correct it, it just says ‘I apologize’. That’s no use to anyone. It seems to be LEARNING to be stupid! Is it that people don’t give a thumbs down to stupid answers so it doesn’t log that it is making mistakes?
14 Likes
And it is getting worse over time. Firstly I thought maybe it’s a human factor - we just demand more and more. But after googling and seeing that I’m not alone I believe more and more that its GPT behaviour what is changed
6 Likes
I felt the same way, like it must be something I’m doing. I honestly find it to be almost unusable at this point. I have no idea what happened in the last few months but it seems like every time the add a new feature, the logic of the system goes to hell. I used to treat it as a co worker to bounce ideas off of to help me resolve issues with code and why things aren’t working – now it just regurgitates the same information back at me over and over again. I gave it a 400 line code file and asked it to resolve a problem I was having, it responded like 8 times in a row with literally not making a single change to the code. I ran the before and after files in a “check for differences text editor” and there was not a single letter, including comments that was different from the version I gave it.
6 Likes
Completely agree, I started using git-4 for day 1 and now I stopped paying, it’s getting more stupid by the day.
4 Likes
If you guys try to ask it to name one odd number without e it will reply 40.
Absolutely wrong.
1 Like
that is so true. I’ve been using it very often and solved many problems. But now it just struggles with the basic requests. I thought maybe ChatGPT is not optimized for code I used the Copilot Chat and it is the same. It literally became dumber. As a result, I keep increasing my prompts making it more detailed and simple like I’m explaining it to the 5-year-old, but still, it doesn’t get it.
2 Likes
Indeed, I try to write a grand application (chat gpt helps me greatly to overcome the weaknesses of dyslexia) and it makes a mess of the info, im considering to search an other ai company for help. This is useless:(
2 Likes
100%. Been using it since beginning when the answers were sharp and correct. Now almost useless and often incorrect. We aint getting the real and good version. There is just no way the devs dont see that.
5 Likes
I’ve also used this since the very first launch. I really hate paying for the service and almost wouldn’t even use the free version. I ask it fairly simple programming questions it gets wrong. I often state “Give me just the relevant code or code you change” as it proceeds to not change anything or give me the whole file.
5 Likes
same here, the free gpt at launch was waay better than today’s 4o
1 Like
100% agree. Waste of money. Bigger waste of time.
1 Like
I believe the key to understanding this phenomena is to understand ChatGpt itself. ChatGpt is nothing more and nothing less than a guessing algorithm. It’s not smart at all, in the proper sense. It’s “smart” in the sense that it can be very useful when it’s capable of guessing things right most of the time. It does a great job with Python.
It fails miserably with SAS. It fails miserably with other topics. I’m just extremely annoyed that when you reply to it that the answer it gave you is wrong it tries to guess again totally ignoring the feedback you gave it. So it continues to guess until it gets it right, even if you said “that’s wrong, I’m good, you can stop trying now”.
Happened to me as well. I was asking it to take a look into the website with the link I provided (last time it was smarter and follows my instructions, and it 's really go through the link I gave it), but unfortunately, it gaves me wrong informations despite all the informations include links given. It keeps on replying sorry but it keeps doing the same mistakes after that. I am so disappointed with the service because I paid for it. Where is our priority? I don’t see the worth?
1 Like
Same issues, it causes more frustration now, I will literally send my plugin code asking it to make some changes to the structure in detailed instructions. It will acknowledge then proceed to redact tons of information out of the code and even when corrected and sending new copies of the plugin it still does not understand and removes code from the plugin.
1 Like
But a language model has a conversational buffer. The transformer should understand the context of the reply and based on attention should give a better answer. It is not clear why ChtGPT is ignoring the user feedback. Thats not how it is supposed to build. It all makes no sense 
1 Like
I’ve been thinking about this for a while and decided to Google it to see if it was just me. This confirms it.
2 Likes
Unfortunately, I have noticed the same issue. I had subscription active since the very beginning, when the subscription became a thing, and I just stopped the renewal a few days ago.
Lately, I have noticed degradation in performance, especially in coding, where it started to get even the most basic things wrong (e.g., generated a small snippet where it used the .NET method Equals() with 3 parameters, and many other rookie mistakes).
Now, as I’m using the free version for the last 2-3 days, I find it significantly worse - it’s a huge disappointment overall, because the number of prompts required to get something right has increased by 2x to 3x from my own experience.
I have tested other models from the competition (to say like that) and they seem to be what GPT-4 was at its peak. I managed to test Gemini Advanced a couple of months ago (there was a 2 months free trial), side-by-side with GPT-4 strictly for coding (unit tests etc.) and it performed slightly better (applicable for Claude as well).
I hope this is going to be addressed soon, otherwise, I doubt one can justify the subscription.
3 Likes
Agree completely and not even a single word from OpenAI, a lot like netflix good at first get some loot, then just LGBT … My two cents on the matter: Industry pressure so OpenAI decided to make it stupid and dumb it down, I can´t believe such a promising project instead of doing “machine learning” it´s doing “forgetting” and I’m 100% sure it’s not going back to its previous glory. OpenAI was secretly forced to make chatgpt stupid.
It’ repeats content on a single answer, it mixes previous responses and mashes it up with new content making t unreadable and unusable, it deletes older interactions, when you tell it to get content from a website and sumarize it it gets a couple of paragraphs and when you ask why, a lame-ass respponse that some copyright thing, definitely it’s not worth paying for anymore and googles alternative sucks at best.
I know I’m just a lonely person screaming to nothing in the woods and I’m sure nothing is going to change so let ChatGPT - RIP
thanks everybody,
I’m in complete agreement. I provide guidelines to have it help me write something. It feeds back almost exactly what I wrote. I give it clear limits, I.E. “the character does not know [information character does not know]”. It includes a response where the character is aware of what he should not know. I upload text (under a thousand words) it does not process it. I ask it to modify the first paragraph and provide the text. It jumps somewhere into the middle of the text and changes that. 3.5 worked better than this.
You all remember how search engines started? You could actually find great search results, and now produce garbage? I guess 4.0 has gone the same way? I think the company has settled for a “white bread and butter” AI that feeds the mediocre, rather than a compelling AI that actually provides assistance beyond the mundane
2 Likes