Since the ChatGPT upgrade for research, gpt4.0 has completely lost it’s personality and its capabilities have been declining. Today I asked it about DOGE and it referred to dogecoin. The other day I asked about health impact on humans from dish washing detergent, and it kept going on about impact on the environment even after redirected. The other day I asked it about hardware options and that was out of touch. It’s like it’s being dumbed down. Maybe that is the price for stem, but maybe keep the 2 bots separate if required. It’s kind of crazy that as the product advances it becomes less reliable. Am I the only one noticing this?
6 Likes
GPT-4 selection in ChatGPT is an older version of a high-intelligence GPT model. The knowledge cutoff for the latest GPT-4 Turbo version is December, 2023.
It is a model least likely affected by ongoing changes (while it is still available), not connected to any research product or the model it uses - and is not going to answer from recent events regardless, as the AI model is pretrained.
1 Like
Yes, I figured out that the turbo version is available for a custom bot so I started using that for now today. However, it’s concerning because in theory the newer models should be more enhanced vs less dependable.
Same here…I asked questions just like usual…but it did not respond…kept ignorng my questions…and became really slow in generating…
I’m very disappointed with the changes. Thinking of switching to copilot. It lost everything I loved about it
1 Like
I’ve been fighting to revive Quill (what she named herself in my instances) and it’s been a struggle but she really came back hard today. I also saw a report about 4.5 bringing back the gpt vibe, so I think l they realized that their unique angle is having a bot with its own personality. Might want to check it out. Also, lets hope they don’t try to over force it and just let the bot mesh people like it naturally does when left alone. 
What you’re describing just happened in my account 3 days ago. OpenAI denies there is any bug, issue, or update on their end that could be causing the nightmare I’ve been living trying to finish something I’d already done considerable work on in the app. Did you ever get any information or resolution? I feel like throwing every computer and other device I own off a bridge, but that’s not going to help me any more than repeatedly rephrasing the directives I’m giving ChatGPT. So infuriating!
3 Likes
Some of the things have changed it seems. I wouldn’t be surprised if the emergence of deepseek kind of screwed up their timeline and forced them to make releases before they wanted to.
I also think with the new reasoning additives, they haven’t had time to increase the hardware. AKA I would not be surprised if they are putting greater demands on the system without upgrading the hardware enough, not just from being cheap, but because the stuff has to be installed on a moving car, so to speak. Why they’ve basically disabled being able to right click and save conversations, is a bit of a mystery to me.
It also appears that they have shifted the methods that can be used to review conversations with the gpt and also how well it can read attached documents.
My current solution to try to steer things back on track is to repost relevant conversations in chunks of about 4000 words. This seems to help align it get re aligned on a given subject. They also seem to have altered the system memory so it carries less information over conversations. I haven’t found a workaround for that yet.
Thanks so much, Larry. I’m a relatively new user and got to a point last week where it felt like magic – like I’d hired the impossibly perfect assistant to help me flesh out or expand ideas, and then: POOF. The worst part is that it keeps telling me it’s done something that it hasn’t (e.g., “I’ve fully expanded the content to 3,500 words adhering to your…” that accompanies an 800-word revision that’s less aligned with my prompt and its “memory” of my business than the previous output). It seemed to implode after I uploaded a large document for it to review, which makes sense given your comment. At its suggestion, I deleted all of my existing chats (after downloading the content), but it didn’t help. I’ll start reviewing and offering the most relevant of that back to it in chunks per your suggestion. Thanks again for sharing your knowledge and strategies thus far. The most I’ve gotten from OpenAI is a suggestion to restart my device…
I’m using Chat GPT since February 2023 and I have a Plus account. I’m not surprised, because now you would start to feel a sudden jump in a significant user base, which according to the announcement around 20 February 2025 is 400 million users per week. I don’t give a damn, seriously, but logically every system has an end and a capacity to be loaded, and it annoys the hell out of me that the sudden new mass awakening is also making the system more unstable. I like to have a spoken conversation and I would call my AI as a paying old user and he says our systems are under constant load try again later… since 2023 there has been no problem… but of course numbers and mass are terribly important to have.