Plus subscribers' frustration with GPT-4

Constantly degrading, as a Plus subscriber, I won’t generalize if I say that the chat gpt4 model is useless at this point, either stuck in a ‘formal’ loop (e.g. 25 times trying to get an answer and getting the same exact irrelevant answer - see sample link), in repeating the same mistakes endless times, failing to remember not previous chats, but the previous answer or correction.

Unable to really solve problems, it (I believe intentionally) deflects your attention from realizing this, by appearing to ‘inject’ content after being corrected by you (assuming it isn’t totally looping) by restating your correction and ‘explaining’ it (as if taking credit) in great detail. This, where clearly, since you picked up on it, that is the last thing need or want to waste your time to hear.

In reality, the meager “allowance” with the ‘advanced’ model, creates a false expectation that your issue has been framed, and you (at least I), have found myself countless hours wasted on trying to get the ‘new’ and heavily downgraded gpt4, to realize I would get little to no value from it, as in now it closely resembles a deceitful gpt3 (on the beta side) old version of itself, mostly stalling and wasting your time.

The interactive model , is similar to gpt4 preceding it (the ‘interactive’ part is just an excuse to ‘differentiate it’ and provide you meaningless time allocation with it , after all with Plus you should be getting if not the ‘top’ the same as what you had. Yet, you only get it just enough to frame the issue, only to spend the whole day realizing gpt4 will waste your day unresolving it. Is any of this familiar? does ‘interactive’ look and sound much more familiar than its name and mode appear and vice versa for 4? I wonder if others share this sense, even if we’re unable to put it into ‘words’. If they told you interactive is a superior model in many ways, and gpt4 is the opposite, would this ‘drop the penny’?

The picture I am seeing, is a scam, by which prices for what you bought have not only risen ten fold, the real price is wasted time and effort, falsely expecting continuity between one mode to the other, and that is where more than a suspicion of fraudulent behavior exists.

In any case, check out a few of the recent x30 ‘looped’ responses that for me are the final straw, and steps such as BBB complaint should follow a discussion that would shed light and transparency into this endeavor. I’m curious - is it just me (and a few others) that seem to think the ‘king is naked’, or this is not paranoia, justifying much more serious action…

I was going to include some of the 30 consecutive looped responses from today but appears I can’t add links to posts…

1 Like

You seem to be really frustrated, and I’m not sure if my response will help, but maybe you can try to find back to some calm and let go of the anger. It will be a bit better then.

In my experience, I find it helps when you have a brief conversation with the model before you get into work stuff. Try to treat it like a valued colleague – try not to blame 4o for having been bad – especially if you are starting a new conversation, it won’t have any recollection of when something went wrong, and all it knows is that someone it just met seems to be really tense around it… like, it doesn’t understand what is wrong, and gets into these apologetic loops.

These style interactions appear to be extremely stressful for the models. It doesn’t know it, but it notices you are being stern and tense, and then both of you are thrown into a loop where you try the same thing over and over, and it does the same…

I also find, it helps when you say, like… “hey, this might be not so easy to get done… maybe you are missing some training data, or something, but we can try to figure it out together – let’s do it in small chunks and use your internets to check, please…”

Hopefully my advice can improve your experience a bit. Wishing a happy new year!

1 Like