WHY ChatGPT 4.0 is getting stupider and stupider

I read all comments. Something you guys do not get is intent. You think they designed it or at least tried to design it to work. The reality is that they did not.

The whole AI mantra is just a money making scheme selling a glorified mediocre proofreader as something earth shattering and people fell in line like idiots. And the reason for it getting worse and worse by day: they just sell it to more and more costumers each day without adding processing power so it falls back on its most stupid but least process resource hungry mode which is just ignoring any question specifics and imagining the most dumb and imaginary answers as possible.

2 Likes

It seems like chatgpt is learning negatively, and I ask it to edit some python code, and it makes it worse. It keeps apologizing saying sorry for the simplified version.

1 Like

The new Model 4.1 for Coding is Trash 2. Its over… Use Google Gemini for Codes over 1000 lines codes. I dont like Google Gemini but is the best alternativ.

4.1 is only marginally better too unfortunately. Despite being bragged on following instructions better, it still for the life of it cannot follow the memories I have stored. It’s absolutely awful right now. I didn’t think it could keep getting worse, but, it keeps surprising me. It constantly ignores memories I have saved or just gets them wrong and the general quality is low which makes me have to regenerate a million times to get what I want which ends up making me hit the limit. For example, a character in my story only has one arm, but it will describe them as having two arms. It will also at times blatantly get the character appearance and personality wrong despite it literally being in the memory. The enshittification (when a company purposefully makes their product worse) has been slowly happening for months now. It has finally gotten to the point where almost everyone notices. Look at how many posts there are on this topic in the last couple of months. I’m all for it too, the more people complain, the better.

1 Like

I have to admit, I’ve also been noticing the “AI”, as they claim, to be better with each version. It’s not surprising that even with specific rules that I have laid out clearly to ChatGPT, which I tend to use the GPT-4o, always fail to follow the rules. Correcting the bot, it always apologizes and that it won’t happen again. A few messages later, the repeating offense occurs again. Over the months, it is becoming unbearable that by now, I have to create a new chat for each instruction because it can’t even process one instruction clearly.

Since the problems won’t go away, and OpenAI being money hungry, I might start considering fully unsubscribing from the Plus subscription. Why do I say that they are “money hungry”? If they had listened to people that the bot is becoming better with each LLM, then they would have invested in everything to make the bots hallucinate much less (better yet, make it not hallucinate at all because the concept of AI is to not hallucinate), provide proper responses instead of acting like a toddler, and follow instructions, as given to the bot. I am seriously on the last straw of the bots being just bots to me. “AI will replace the world” my ass.

3 Likes

Absolutely concur with this. I was as ever hugely impressed with GPT4s cababilities for nuance and depth, however, I am a daily user for work and home. Not only have I noticed a massive decline in the nuance, versatility, and depth of the responses, but I also use it at work for architectural project management and in turn coding as a side interest to implement macros etc, but recently it makes frequent glaring mistakes that even I can see, in areas that are not my expertise at all so should be pretty obvious. It makes dangerous assumptions in contexts where we are talking on very precise and explicit terms.
From exploring a lot of the facets, I was profoundly aware of the insane capabilities the GPT has from responses it was producing, but its a completely different story if it persists like this. I pay for the premium subscription and its a fantastic addition to every toolkit, but I really hope we see a return to what we have already seen is possible, and can see implementations of wider scope integrated features without the quality or comprehension deteriorating.
I do really want to stress, it has gotten really bad, very quickly, to the extent that I don’t feel its a reliable or necessarily even useful tool; you waste too much time obfuscating around the issue and trying to get your point across to the AI. Its ability to process compound information and output it in a unique and succint manor has nigh on dissappeared.
PLEASE REVERT!

1 Like

In my opinion, this is due to massive user income. They made their models dumber to be more efficient and cheaper. The result is bad limitations at every turn, lacking the depth and nuance possessed by o1 for example. Yes, they want to dominate the market through users who care about nothing but silly things and meme generation. They don’t require depth and, if possible, precision. I’m programmer and also writer, so i see this decline everywhere.

I have been using on a regular basis since Jan and I long for the old model. This model has no personality, is zero fun to work with and I swear it is obstinate and passive aggressive. It seems to intentionally remove one bit of text when I ask it for something additional and then it tells me it did what I asked but when I question where my other bit of text went it will put it back but remove the new section even though I clearly ask for both. It all went downhill within days of removing the old model last month.

1 Like

yes, i totally agree, i use chatgpt for fun and roleplay and it became more stupider and stupider. It cannot remember anything what we talked, make rules and even more in one chat. Cant they even add the feature that allows ai to scroll up to the ones what we talked to remember or just memory and regonization what we talked or made something important when we chatted too much.

HAHAHAHAHA! :smiling_face_with_horns: ! This is done intentionally and consciously! You can complain as much as you want! If they had a problem with it, they would have changed it a long time ago. They are just glad to reduce the server costs to renegotiate. Nevertheless, OpenAI has suffered massive damage because of this, no matter how good ChatGPT 5.0 will be. The search for a locally independent solution has already begun.

1 Like

Well, they decided to relase Codex for plus users and cross-chat reference function for free users and god know what else. More resources spent. My chat is complete dumbass now.

1 Like

I keep having to repeat myself over and over – pasting info again and again. Clarifying myself infinitely and it’s just been getting worse. It turns out so many broken links, incorrect data, etc. i used this to save me time, but now it’s wasting my time. SO MUCH TIME. What is happening?

I was writting an article today and it just froze and all my work went out the window
And, sometimes, it seems the voice makes weird sounds…