I have found that since unification of the parts of chatGPT I cannot get coding to be of usefulness. Yet in Poe when I use GPT-4 32k it will do it as it used to. I mean output full working code, and be able to be intelligent about context and dependencies. Now chatGPT makes mistakes, alters my logic, can completely output something unlike the code I gave it to modify. It will not repeat things reptitiously that I have issues typing myself, itâs like itâs trying to be very thrifty and cheap now. It will not comply and seems to fully ignore my needs to output something fully.
This is disturbing, I have an issue with however it has been limited, where it now is obvious to me it has been dumbed down. I doubt this is an accident, and I am disappointed at the gas lighting of others about this who obviously never used it as people like me did previously or now.
If you really donât use it for full code output, editing big code chunks, donât chime in please. If you do and have before the unification then please do chime in. Else your just gaslighting without any business here, your happy so move on.
Who else has seen this?
Here is an example and it is easy to reproduce. Just post code, ask for a change, and try to use it in the existing code. You see how it slowly shows me more and more of what it concocted yet is not initially. How can I see what chatGPT doesnât show me. ??? This seems as if the instructions are telling it to short me output now in a strict way that can cut out the part I donât have and so is a useless output. Itâs awful and offends me how chatGPT fails.
Yet using Poe with GPT-4k I get full code output, big output, and itâs like way below 4k!!! I mean we know what 4k is and 32k is, but seems OpenAI doesnât.
https://chat.openai.com/share/eae01410-7395-4b8f-9047-382a8574dfa1
This post was flagged by the community and is temporarily hidden.
This post was flagged by the community and is temporarily hidden.
N2U
4
You can use the GPT-4 32k model over at the playground. ChatGPT is an entirely different product that tends to focus on shorter responses than what youâre looking for.
1 Like
Yes, I realize that. Although it did output fully previously, and so I am trying to get some clarity and full disclosure that it has been changed and I am not finding a âregressionâ, this is expected behavior. Itâs because it was so good, and other stuff even Poe is not great on the âinterfaceâ as chatGPT. I wish I could pay a little more for how it was, I am addicted you could say, it was soooooo sooooo good. I went from singing praises for OpenAI to becoming a bit jaded now. This is how fans are lost for anything, as they âdumb downâ for the masses the prime fans who are enthusiastic initially get stomped upon and leave fandom. I see it on so many big companies services more and more, I was hoping chatGPT and OpenAI were going to keep up the quality as it scaled up for the masses.
N2U
6
Maybe youâre reading a bit much into this?
OpenAI has no interest in deliberately making their products worse. But I had a look at the conversation link you posted, and my best advice is to start a new conversation and try with step-by-step instructions. If you start discussing with the model, telling it that itâs âjust making stuff up,â it will continue to hallucinate even more.
Flagged for spam? Advertising? Biased Unfair!!!
This is pure censorship. What is spam or advertised in this??? I donât own Poe, Perhaps a bias towards a customer of the OpenAI API??? Quora is a customer of OpenAI, not me so this is not at all true and a full out censorship attempt.
Itâs interesting how many sociopaths there are out there who try to suppress peoples valid posts. I would love for whoever did that to speak up if they dare. Stop hiding and use your brain and speak about what is wrong with my posts.
their product is definitely worse than 3 or 4 months ago.
N2U
9
To be clear here I wasnât the one who flagged yer posts, and I donât know who did, but if enough users flag the same post it will be automatically hidden.
If you ask me itâs probably stuff like this that got you flagged though:
Thereâs ~.5 million people on the forum, and only around 0.03 % of them have anything to to with OpenAI, the rest is just developerâs like yourself, trying to help each other.
I tend to view forums the same way I view a bar or a nightclub. Thereâs no party without the guests, and its the staffs job to keep the place friendly and the guests happy, that also means removing people that disturb the other guests.
If you feel like the bar staff is targeting you unfairly, then my best advice is honestly to find another place to go. Sure, you can complain to the manager, but itâll most likely be ignored, and you wonât have fun doing so, which was the whole purpose of going out in the first place.
Yeah its been all but unusable lately. GPT-4 today seems to be about as smart as GPT-3.5 was back in May.
I have found Poe works well using GPT-4 4k API.
From the various aspects of how OpenAI runs brought out in the news, I hear the Poe CEO is on the board 
heh I wonderâŚ
