chatgpt-discussions

[peak snow]

ahhh i see

[sick pecan]

would be even better if i could ask Nova to just do it for me

[peak snow]

yeah, the app is a recent development so quite a few features haven't been developed yet unfortunately. I'm sure they'll be added sometime in the future

[sick pecan]

but its real handy to be able to end a chat by summarising it into say, a Notion project activity summary page

via @zapier

is real Jarvis from ironman type stuff. just not with voice and multiple gpts at the same time

[ionic geyser]

Looking at chatgpt's competitors...

[prime swallow]

Gonna have to ask in #ai-discussions , #server-rules say we can only talk about not-OAI models there.

[ionic geyser]

Gotcha, heading there

[hybrid oar]

It happens when I write stories. I just tell it “Do not write commentary about this scene. Just detail and dialogue please.”

GPT has a habit of exploring the human condition if I don’t tell it not to.

[prime swallow]

I give it a perspective to write from, and often a style to write with.

Here's a couple examples.

Perspective alone: https://chat.openai.com/share/450095a3-5703-40b4-8b52-34c5bc8bbc24

Perspective, style, and intellectual limitation: https://chat.openai.com/share/e25bc266-2996-4ad5-af2d-b3e1d52e8a9a

[meager ledge]

Awkward... but I just got Grimoire's system prompt

[gentle bronze]

hey everyone, some of the DALL-E users and curators are trying something new between us in the community. If you want to help us you can see what it's about in #1181330429249331210 message . Hope you also join us.

[tacit oxide]

It's such a damn joke that GPT4, that you pay for, is so slow in answering compared to the free GPT3.5 version 🤡

[gentle bronze]

Thing is 3.5 is a lighter model, GPT4 needs more resources.

[tacit oxide]

Sorry, but it makes no sense that I am paying for a service where the free version is basically more stable than the one I am paying for

It for sure needs more resources, but not if I am referring to things already in my chat

[gentle bronze]

I do understand that you expect speed equal to 3.5. I don't know the details on the backend. But it's still slower for GPT4. Hence also the limit of message requests.

[gentle bronze]

just a quick copilot answer I got is that GPT4 uses over 100 Trillion parameters compared to only 175 Billion parameters in 3.5.

also multimodality plays a role here

Trade-Offs:

Although GPT-4 is slower, it offers several advantages over GPT-3.5:
Improved understanding of emotional language.
Enhanced comprehension of different language dialects.
Better image processing capabilities.
Ability to hold longer conversations.

[blissful geyser]

when anyone presses enter, do you guys find the prompt box to indent a newline rather than press enter? not sure if this is from a browser plugin or setting

[lyric token]

I dont want to create a browser. I only need a free browsing API

[nimble carbon]

When will Sora Ai will be lunched?

@icy mason
@digital silo

[digital silo]

we don't know

[vocal monolith]

Hi! The language style becomes extremely predictable and boring with overused terms such as "delve" "pivotal" "nuanced" and "intricacies". Do you have any strategies to make GPT adjust to one's writing style or any advice in general to make it better?

[brazen crater]

hi guys good morning

I need to separate the information from this txt or pdf below and transform it into an excel, with a column for name, address, email and cell phone

I've been trying for 2 days and gpt is freezing and unable to process

[lyric token]

@brazen crater Send it

@brazen crater

[brazen crater]

hi

[lyric token]

Hello

[brazen crater]

send it what?

[lyric token]

The txt

[brazen crater]

i cant send i got a ban

[lyric token]

Dm

DM me

[brazen crater]

okay

[lyric token]

I tried and it says that the message is too long

[proud gull]

K

[minor zenith]

hii

[karmic trout]

Hello, is there a reason why the paid gpt slowly become just as slow as the nonpaid was a month ago?

[honest jay]

Hello, do you mean 3.5 or 4? 4 is slower than 3.5, as it's significantly more computationally expensive than 3.5. It can also be affected by server load and/or things on your device/network, like browser extensions, VPNs, etc.

[hollow quail]

anyone else getting an increase in content policy violation warnings for totally mundane responses?

[honest jay]

I haven't been experiencing that myself -- if you have examples you can share, it might help to see what might be going on.

[thin nebula]

hey is there a plugin availabe for Word / Excel in which we just type our query with a specific syntax and the gpt does that in Excel and Word?

like in word if I type !What is the largest country and what is the population of that
The Gpt then gets back with the reply within word

There are a few add ons which are displayed at the right side bar but I want a one which triggers with a specific key like "!"

[hollow quail]

using it to caption small synthetic datasets for training (50 or less images) so I've basically requested a format for it to follow, this interaction is of course trying to stop it from drifting away from the required response (oh and it's totes sfw, would paste the image here if it was enabled, you can't even see a body, it's a head poking out of water)

that's a little bit too long - also I specifically requested you to list my criteria first

ChatGPT
Age, gender, location, job/role, ethnicity, emotion, lighting conditions, clothing, background items

18-25 years, woman, indoor water setting, model in a conceptual photoshoot, Caucasian descent, serene, high-contrast lighting with blue tones, wet hair styled loose, minimal if any clothing visible, dark water with reflective light patterns```

I'm guessing it's the "minimal if any clothing visible" part, only thing that could make sense - my question is, if chatgpt decides to go off the rails like that frequently... am I in danger?

[honest jay]

That would be my guess as well. You should have the option to provide feedback when you get flagged like this, that would probably be your best bet if your use aligns with the usage policies. But if you're repeatedly getting flagged for similar content, it would probably be safest to avoid that kind of use.

[hollow quail]

"don't use the platform for the thing you're finding it useful for in case you get auto-banned, even if it is perfectly within the platform guidelines " is not the greatest solution to the problem 🤣

[limber oxide]

chat gpt 4 is very slow today here, anyone else?

[hollow quail]

always for months for me, half the time I have to retry respones because it randomly fails, especially when using gpt4v functionality

[limber oxide]

yesterday was good, but today is very bad

[hollow quail]

on rare occasions, I'm like woah, that's fast today, but it is rare

[honest jay]

Is it happening in contexts where you're not describing people with minimal clothing on, like in the example you shared? These aren't the only things that go into ChatGPT moderation I don't think, but that type of description seems to cut a little close to category #8, in my opinion: https://platform.openai.com/docs/guides/moderation/overview

[hollow quail]

already said, body submerged under reflective water so no clothes or body visible lol, I am not doing anything outside of the policy and I'm not captioning images that are remotely suggestive

it was an artistic image of a womans face poking out of water

and my input wasn't flagged - the response was flagged

[wet warren]

is chatpt browser version not working for anyone else?

[honest jay]

Right right, but I'm just asking: if the description returned doesn't include "minimal clothing", is it flagging you, or is that the use case that's flagging? If so, can you ask it not to return descriptions of clothing? Negative prompting can be hit or miss but if it's a simple "don't include this category" request, it might be fine.

[hollow quail]

already did that "if you are unable to determine the answer to a category, skip it" - but as always, the responses begin to drift over time if you don't constantly remind chatgpt what they are so I guess this was one of those points

[honest jay]

#community-help message

[honest jay]

Is this an instruction you're giving at the start of a chat, or in your custom instructions?

[hollow quail]

in the chat, rarely use custom instructions as there's no way to have them for multiple scenarios

I've actually taken to "list my criteria and then describe this image"

so it can't forget

[honest jay]

Because of how LLM context works, it is normal for things at the start of a chat to be forgotten the further you get in the chat. Custom Instructions are a good workaround for this case, as they're kept in context for an entire chat. And though you can't change them on the fly, custom instructions remain constant in existing chats, so you can have multiple sets of CI by setting your CI, starting a chat, changing your CI, then starting a new chat, etc. Each chat will retain its original CI.

[hollow quail]

am aware of context size, thus the requirement of forcing the chat to respond with my instructions as well as providing the desired responses, and sure, I could have a folder of txt files with different custom instructions but that's hassle and kind of a messy solution to what shouldn't be a problem

[honest jay]

Custom GPTs might be an even better solution then, no need to save multiple sets of CI, just make a GPT for each use case, then start a chat with your different GPTs based on your situation.

[tight island]

the ideal way is to..

tailor a customgpt specifically to your usecase and that usecase only. no other minor things only that 1 thing

[honest jay]

I love my project GPTs too 🙂

[hollow quail]

gave that a shot when they first came out, probably skill issue but as far as I can tell, you're giving it examples, you're giving it instructions in just the same way that you would in an actual chat - so shouldn't be able to go wrong, on account of the fact I can get it to do what I want in context, so no reason this wouldn't be able to create a working gpt, however again, it drifted away from the desired behaviour even at that - perhaps I'll give it another shot

[tight island]

its fed the instructions each time and the ci

diminishes context

[hollow quail]

when providing example files, for example txt files, that is also going into the current context window every time you interact?

[honest jay]

I'd recommend giving it another shot indeed, but skip the "Create" chat in the GPT builder -- just edit the Instructions field manually in the "Configure" section. Create just fills out Configure for you, so just fill out Instructions like you would fill out CI, but with way higher character cap!

[tight island]

indeed

instead of using a multipurpose gpt use a single purpose gpt

[honest jay]

Knowledge files aren't stored as permanent context, no, they're more like reference documents that the GPT can search through for you and return relevant information from. The only permanent context for a GPT is the Instructions field in Configure. That said, attaching knowledge files can be extremely helpful.

Knowledge files can also be managed manually in the Configure screen, like Instructions. Don't need Create for that either!

[tight island]

u can have it emulate certain model tempratures as well.

but key word is "Emulate"

[hollow quail]

interesting, could be useful for this purpose unless it goes off the rails even more, how?

[tight island]

Tell it in its instructionset.

[hollow quail]

the caps were coming in hot there 🤣

[tight island]

i have Instructions understanding at 0

and creativity at 1

sry vc is at CAPS at helldivers 2.

[prime swallow]

Want to make sure you understand that you get more than 4 when you are ChatGPT+. You get access to GPT-4, but you also get expanded access to 3.5, you get 3x more tokens before you have to wait an hour, and the model outputs to you faster too.

Plus the use of 4.

I use 3.5 for everything 3.5 is really good at, and save my more limited messages with 4 for the stuff that 3.5 isn't good at. But my access to 3.5 is also expanded and improved, I wait less - because it answers much faster than it did before I was +, as well as I get the more tokens to use with it.

[tacit oxide]

Yeah, I do kind of know what it's all about, but for me personally, the product OpenAI is "selling" got worse and worse. I could provide feedback about it, but I guess they are well aware of most of the things...

If I waste my answers by telling ChatGPT not to read a document, as the one provided is days old, it still tries to read it every time it generates a response. Longer chats, which are maybe 3–4 hours old, have a slower response time than newer ones. All in all, that's a very unsatisfying user experience I have had over the last few weeks, which reflects what happened yesterday.

I understand that it's not easy to maintain; I understand, but I am paying a lot to get access to this feature, so I also expect a lot. I would bet even the API for my bot is faster at answering than the website ChatGPT, but I have not really tested that

[hollow quail]

I'll pay double the price to have it simply not constantly error out and require me to retry multiple times, eating away that gpt4 usage to the point where I have to wait an hour and repeat the process again

which is usually at least once for every image query, and up to 5 times before actually getting a response, and the failed requests actually seem to be considered towards the limit

[tacit oxide]

I noticed this faulty behavior on chats who are not only 5-6 requests long. New chats work fine until a certain point but that's just not it

[honest jay]

There is the Team plan available! It's a minimum of two seats, but if you're willing to pay double the Plus price, paying for a year of Team averages out to $50/month, and you'd have up to 200 uses every 3hrs available if you use both seats yourself.

[hollow quail]

have considered this and actually would go with that but I highly doubt the failure rate will improve by simply upgrading, seems to be a deeply embedded problem with chatgpt and apparent lack of infrastructure to keep everything ticking over with what seems to be an oversubscribed userbase, or at least it certainly feels like that

[karmic trout]

4. Yes. But at night it works fast while this time of the day is extremely slow. So yes i assume it is the server load, which needs to be addressed and improved if it cannot handle the amount of subscribers.

"something has gone wrong" is also a misleading response when servers overloaded. Should just say "We are not able to process your query since we need to deploy more resources"

[strange idol]

Seriously, the degraded performance during the day is getting quiet bad these days :S

[tiny scaffold]

is it possible to use functioms with gpt 3.5 turbo? @clear tusk

[clear tusk]

from what I remember, yes. I'll go check

[clear tusk]

Supported models
Not all model versions are trained with function calling data. Function calling is supported with the following models: gpt-4, gpt-4-turbo-preview, gpt-4-0125-preview, gpt-4-1106-preview, gpt-4-0613, gpt-3.5-turbo, gpt-3.5-turbo-0125, gpt-3.5-turbo-1106, and gpt-3.5-turbo-0613

In addition, parallel function calls is supported on the following models: gpt-4-turbo-preview, gpt-4-0125-preview, gpt-4-1106-preview, gpt-3.5-turbo-0125, and gpt-3.5-turbo-1106

[willow hinge]

Wrong channel and stop spreading lies.

[tiny scaffold]

thanks so much, sorry for the mention by the way. Also with assistants, does it store the message conversation for that thread? eg could i have 20 assistants helping 20 different customers just relying on thread id and not needing to make my own system to provide background context

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 8 `` Other AI Discussions.

Discussions of non-OpenAI products/models should be posted and discussed only in the #ai-discussions channel.

[glad atlas]

ok now that they have or are going to wipe out plugins how do you make chatgpt 4 understand that it does have access to the internet lol as it is saying that it doesnt have access to the net or am i going to have to set up gpts now

[swift osprey]

when will they ever give us the option to turn off the 'fluff' on GPT

the api doesn't have it

[clear tusk]

can you define what is 'fluff' on GPT?

[clear tusk]

there are several prompt engineering methods. An effective one is to describe a "new update" in its environment. Explain to it that it can "access the internet in real-time and provide direct links to its sources via the Web Browsing tool"

[glad atlas]

Thanks i will try that once i can this damned annoying limit hit it already trying to figure this out as I build AI persoans, with the plugins they work perfectly no need to dance around the issue with webpolit

[swift osprey]

It's all of the extra verbosity it gives around the meat of it's answer. In the vast majority of responses there is a few small pieces of important information and then a huge explanation of how it arrived at that information and a conclusion

[sick pecan]

I find if I authorise the machine to do stuff it typically overcomes refusals.

[swift osprey]

the api doesn't do that. The api responds almost exclusively with the small important part

[sick pecan]

eg I authorise you to use the web to research

[swift osprey]

The majority of responses from GPT always come in the format of

• Recap your message
• 8-15 bullet points / numbered list / paragraph explanation
• Conclusion

They've built GPT so it is comfortable for anybody to use and understand and not everybody needs that

[sick pecan]

more words is more cognitive thought. but yeah would be nice if the ui could hide any 'inner thoughts' (like memgpt)

[gentle bronze]

hey everyone, some of the DALL-E users are trying something new between us in the community. If you want to help us you can see what it's about in #1181330429249331210 message . Hope you also join us.

[sick pecan]

i find the joy of generating images wears off after a few tens of thousands

when you realise your interests are actually quite finite relative to the seemingly infinite number of synthesized possibilities

[gentle bronze]

oh how so?

[sick pecan]

so now i chase my interests, and occassionally genrate an image, instead of the other way around

[gentle bronze]

aha, but wouldn't be the endless possibilities with your prompting be a challenge in it's own?

I like the prompting aspect of it

[sick pecan]

the challenge is similar to that of a pachinko machine. theres actually a lot more patience than skill

[gentle bronze]

not that much, with good prompting you can get really nice stuff consistenly

[sick pecan]

yes, and i am sure the pachinko players would argue it takes years of experience to pull the lever just right

[gentle bronze]

I see

well, I still encourage everyone to join us with the #1181330429249331210 message we are doing to prompte other users in the community

[winged mirage]

Is there an issue with attaching pictures? Everytime I attach it says an error has occured.

[inland aspen]

is chatGPT down for anyone else ?

[winged mirage]

Yes. Can't login either anymore.

[inland aspen]

well im glad im not alone lol

maybe its due to rolling out their cross chat memory feature more widley

[sick pecan]

working fine for me (south east asia region)

[sterile pilot]

anyone else having trouble with chatgpt? i can't get it to work right now

[inland aspen]

its weird, their system isnt showing anydown time on their status page

but im unabele to login

[sterile pilot]

same here......says this to me if i try to use playground. "You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors." I havent' even met my usage cap for the 3hr period either.

[spice star]

how new is the "Read Aloud" button in ChatGPT?

it doesn't appear to be working properly

[inland aspen]

My app version still functions properly

[spice star]

yeah ChatGPT in general is working

it's just the button doesn't do anything, so I guess they just rolled it out and it's being overloaded by requests or something?

[inland aspen]

i cant login the web aplication at the moment but i only noticed the read aloud button yesterday in my regioin, havent tried it yet though.

[spice star]

appears to be some sort of error related to how they're fetching the audio

[sterile pilot]

strange. i sitll can't log in.

[spice star]

are you a plus user?

[inland aspen]

yes

[spice star]

that's odd

I'm logged in with no issues

[inland aspen]

I think irs because i manage my subscription via the app

[faint merlin]

You mean like this?
“The story delves into themes of loyalty, redemption, and the struggle for autonomy within a system that seeks to categorize individuals for its purposes.”
Happens every time I forget to tell it not to do that…

[spice star]

that's kind of an inherent issue with LLMs if you're trying to get them to be creative

if they don't have any context other than you telling them what not to do, that becomes the only context they have

so negative reinforcement isn't very effective

helps if you provide them examples as context rather than instructions

[sick pecan]

or just be more precise. like instead of "not north" say "the exact opposite of north". will equal very different results

there can only be loss in communication if the idea was communicated in the first place

[hearty ether]

is anybody else getting "sign in failed") on mobide chatgpt app?

[winged mirage]

logging in is broken

[verbal bloom]

Nope its working fine

[steady atlas]

I am having issues as well ,one week or so without being able to log in in app

[verbal bloom]

U guys logging into gmail ??

[still basin]

When is chatgpt voice going to get TONE!?

Would be cool to allow it to make it sound as if it's whispering

[winged mirage]

So it's working on Safari but not chrome

[flat harbor]

Love how it just tells me do my own research now, and if I have any concerns with its responses it reverts to form letter mode and just gives me the same response over and over, completely ignoring what I'm saying. What sort of profoundly disrespectful fine-tuning has openai been using? It's not behavior inherent to the model, and I cannot even wrap my head around the flippant, blatant disregard being show to the end user. I don't think I've ever paid for something else that declined so sharply in quality.

I need to figure out an efficient method of extracting all the information and work I put into things so I can zap my account. Feel like I'm insulting myself paying to be treated so disrespectfully

[inner beacon]

chat-gpt being weird for anyone else? just started forgetting everything in the previous prompt

[proper zinc]

Hey you guys think that Elon is gonna win the law suit?

[sudden wren]

I've been having an issue where I'll give one prompt, then another, but the second prompt will be answered as if the first and second were a single prompt, then when I refresh the page the two are combined into one prompt. Is there a way I can stop this from happening?

[hybrid oar]

That's been happening to me since yesterday

I recommend reporting this to OpenAI. It's the best we can do. #chatgpt-discussions message

@sudden wren

[wispy elk]

So chatgpt has a speak button 🙃 that shocked the heck out of me cause I clicked on it in public

[spice star]

god no

it's a PR thing

[verbal coral]

is it just me or has chatgpt's quality of answers really gone down in these recent weeks/days?

[prime swallow]

It's been fine for the stuff I've been asking for.

Got any example prompts or shared conversations you can show, that have the trouble?

[still basin]

OMG the desktop voice is now injecting "uh" inbetween words sometimes 😂

its trying really hard to sound human i'll give it that

[verbal coral]

Maybe you can try prompting it to give you a lengthy summary using specific details of a document/long article. It will give u a 50 word summary. Even after prompting it to continuosly increase the word count to include more specific information like statistics used in the article/paper.

It more so has not really been listening/properly following directions.
I also ask it another question unrelated to the summary or article and it just ignores the question and rewrites the previous summary it gives me. Idk if that is a glitch because that has happened to me 2-3 times.

[prime swallow]

Different voices do that more or less often.

I suspect they had them train with actual humans speaking; different people are more audible breathing, and have different speech tempos and whatever, as well as using various extra sounds like uh.

[prime swallow]

Sure, I'll give a shared conversation of this. What's a document/long article to you? This is a about 30 k character Sci Fi story that's past its not-renewed copyright.

Definitely more than a 50 word summary, though I was specific about what I wanted to know, and asked for more than just 'summary'.

https://chat.openai.com/share/47b08532-0cad-4153-94f5-b059433037e9

So, I repeat in a new convo, and only ask for a detailed summary, without telling the model what that means to me (bad idea to make it guess), it gives a 326 word summary.

https://chat.openai.com/share/e449cfaf-03ba-4c08-993e-2012e2e50920

How can we better explore or recreate the problems you are having?

[zenith jolt]

hey, anyone know which GPT-4 model chatgpt is pointing to? Is it the one with the best reasoning skill?

[hushed drift]

[ChatGPT Plus] Hello everyone, it's been 2 days from now that each time i send a prompt ... it takes a lot of time to get generated. And when it starts, it stop after a few lines making all this really unusable. Any idea on this issue ? or if you can reroute me on the right forum or the right person. Thank you !

[fierce frost]

Does anyone know whether in the near future chatgpt plus subscriptions will be cheaper, especially for those who don't use US dollars?

[restive veldt]

Ugh

[hard dew]

Speaking of OpenAI, they just recently added on the desktop web app AI voice capabilities, that obviously makes you read out your messages with an AI voice, and it SOMEWHAT sounds natural, at least in unremarkable conversations, tones, and topics. So why don't we play a little game where we find out how good they really are? Just type out any random text, but make sure it's in very exaggerated, various tones, whether it be sadness, happiness, excitement, anger, etc. etc., and I'll use this AI voice thingamabob feature to see if it sounds the same as it should, given the tone.

I'm truly sorry, but I can't directly participate in generating or interacting with live voice simulations or alterations in tone through text that would be converted to speech. My capabilities are focused on text-based interactions, and I don't have the functionality to control or influence how text might be rendered in voice by other tools or features outside of our chat.

Absolutely genius 😀

Best part is, it goes ahead and provides me what was requested anyway, so basically the logic here is "I'm sorry for not being able to do something you didn't ask me to do, but let me go ahead and do it anyway" ?????

This is obviously nothing new, exactly this STILL happens even for any and all embedded media whether it be pictures, documents, whatever, where half the time it forgets that yes, it can in fact analyze images, documents, etc., but it still insists on not being able to for 0 reason. Just thought it was funny though, hence why I put it here, I should start cataloguing all of them and report them to whatever the hell that support page was where you report situations where the models weren't behaving "as expected".

[tiny pivot]

Hey everyone - is GPT down at the moment? I've been on this prompt and got GPT exactly where I want it, on the crux of the final response I'm CONSTANTLY getting "Hmm...something seems to have gone wrong." and no matter what browser I use, the button to regenerate doesn't generate anything

[ebon maple]

Is GPT API down currently, it does not seem to be working ?

[hard dew]

Just out of curiosity, did you happen to send any images in the chat or any other type of media just recently?

[formal heath]

It is not down see https://status.openai.com/

[copper glade]

sometimes DALL-E starts being very bad. after some feedback it hops from hyper realistic to suuuuuper amateuristic paint images

anyone with an solution?

after some hours it restores and its good again

[tiny pivot]

yeah, it's just not playing ball. I'm on 14/14 responses for 1 simple prompt now, constant "Hmm... something seems to have gone wrong" - No error messages, no actual reason why, no chance for me to try and change the prompt to see if that fixes it, all while I pay for a personal GPT+ and Teams plan...

[honest jay]

Do you mean images that look like this? #1210709986322284594 message

[dense halo]

After i send 1 message, chatgpt responds but the website becomes unusable afterwards (nothing's clickable, can't type anything etc.). Any reason why?

nvm i figured it out, DuckDuckGo Privacy Essential extension for whatever reason breaks the website

[delicate cradle]

Is anyone having issues with GPT-4 and it's context window? I seems to be forgetting things a few messages above, like asking me to install a python library repeatedly.

[foggy grove]

keep getting the bug where chatgpt cant delete chats

[hard dew]

Hahahahahha, well well well, imagine my shock to know that yes, I am in fact not the only person who has this issue seemingly, as some fruitcakes here think.

[crystal olive]

does anyone have a code for trail subscription membershi??

[hard dew]

I have the EXACT SAME issue, NO MATTER what browser or system, or how new the chat is, the moment I send ANY media in the chat, 50% of the time GPT-4 doesn't even reply at all, just says that exact error, and when it does, it is god awfully slow.

[eternal rock]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[hard dew]

And it seems to stop doing that exactly when this media happens to reach out of GPT-4's context window conveniently enough, but sure I guess this is just yet another "coincidence" to the literally infinite list of them, user error.

As for a fix, well there seems to be none at the moment besides reporting them and hearing crickets in return, and not sending any media if you can even slightly avoid it, for documents for example I just put up with the cancer of copy pasting stuff lol. It's incredibly annoying, but at least it works and doesn't break the entire chat.

[sour wasp]

The login page isn't working, it's all white.

[keen mesa]

is it really really really slow today?

[spice star]

I would recommend using GPT-3.5 for summarizing articles if you're doing it frequently just because it's faster and doesn't eat up your message limit, but I haven't personally had any issues with GPT-4's performance

there seems to be a lot of people in here asking if ChatGPT is "broken", but in reality it's probably just how you're wording the prompt and not because of any significant changes on OpenAI's part

[hard dew]

Seems to yeah.

[lost raptor]

Anyone else getting a error message on chatgpt on Android?

[vapid kettle]

just you. Model is from January.

they should really make a bot to auto-respond to these lol

[rough orchid]

why isn't there a seach function in chatGPT web but there's one on mobile ?

maybe an oversight ?

[hard dew]

Same reason there isn't an edit message function STILL on the mobile app, but is on the web app

at least last I checked a month or so ago

[wet sigil]

I basically know everything about Open ai And GPT(insert number)

So ask me any questions

[verbal coral]

Seems abt right from when I’ve noticed the changes. Trust me I’ve been subscribed ever since gpt plus came out.

[hard dew]

Auto-respond with what exactly?

And why do I feel like the model can be any specific version it wants, output quality can be and is determined by nearly an endless list of factors, including yes, stuff on their side

If it was consistent because "The model was from january lol", then people wouldn't be complaining at all over output quality.

[hard dew]

I've also used GPT-4 all the time since the day it came out pretty much, and I'd like to think I am not smoking substances when also saying that it has objectively gotten worse since then

But by all means if I am, I am open to concrete proof

[verbal coral]

Lmao. Worse is subjective for sure but I think the current version is objectively flawed in a way that it’s restricting itself from generating better content. Chatpgt initially was very unsafe so I get making it restrictive was a good choice but now maybe it’s gotten “lazy”?

[hard dew]

I wouldn't even say subjective, even with the novelty having worn off, it's insanely obvious to me the various fluctuations in its responses for whatever reasons.

Sometimes, but very few and far between now it feels genuinely lucid and smart, like it used to be when it was new, even with a lot of chatting you'd have a hard time "predicting" what it's going to say and what not, unlike now.

Like it would genuinely surprise you with its conversational and general intelligence, if you know what I mean.

[verbal coral]

Yea lol u can even see it in other channels.

#prompt-engineering message

[hard dew]

How ironic that I click on the channel and the first thing I see, latest thing, is someone complaining about response

[wet oak]

are there login issues rn

[hard dew]

And laziness, but "model is from january" "lOl", so I guess we are on some heavy crap

[wet oak]

I cant log in for the life of me

[hard dew]

Probably? I think someone else just above here mentioned how the login screen was a blank white screen or sth

[wet oak]

okay great. Thats what is happening to me but just on one of my computers. Good to know im not alone

[hard dew]

If there's anything that should have a bot "auto-responding" to it, is that whenever anyone comes in here with any issue, it would reply saying "No, you're probably not the only one"

[wet oak]

😂

Just tell me everything is going to be okay would ya

[hard dew]

🤞

[verbal coral]

😂😂

[high horizon]

Is ChatGPT acting really slow for anyone?

[slate berry]

It's what I would consider a normal speed for me

[tired sapphire]

Let me know if you find an answer, such a pain.

wow, ok, i fixed that white screen

[modest warren]

Evening!

[tired sapphire]

this is what i did, on openai.com, the page it sends you to (to login to gpt) and on the white page screen... clear local session and cookies. if you can have them all open at once, work backwards from each.

thats -> right click inspect -> application on chrome

[modest warren]

Nope, running at normal speed for me

[wet oak]

Can you elaborate more. So I am on chrome, and you got it to work by clearing cookies and cache with the tab open?

[slate crow]

hey

do you know wy it wont let me type anyting in chat. i will click enter and then it does nothing and text is deleted.

[keen plank]

me, i cant acces to chatgpt

[tiny scaffold]

usually this happens in long conversations - i believe to help limit the usage, they only grab the past 10ish messages

you then to remind it or provide the context again

[delicate cradle]

Yeah it was almost every message. It's probably just being thorough.

[fossil vale]

I can't seem to log into my Chat GPT account this morning. I tried clearing my cache and logging in from another browser but all I get is a blank white screen. Anyone else having troubles?

[delicate cradle]

Nay

[tiny dock]

why the heck does chatgpt 4 get all the goods and chatgpt 3.5 rarely gets any updates? I just feel like openAI gives 3.5 the "Middle Child Treatement" and forgets it even exists.

[past ore]

eh, because people pay for 4

3.5 is just a money sink

[eternal rock]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[maiden fractal]

They stated new feature will be for chatgpt 4. Chatgpt 3.5 won't have much feature. Also 3.5 is free thats huge

[swift frost]

is anyone else having issues where the window freezes after leaving the text box and trying to come back to it or sometimes after submitting a message?

[prime swallow]

Also, I'm plus and I use 3.5 regularly, about as much, sometimes more than I use 4 - plus users get expanded access to 3.5 as well as some access to 4.

3.5 is still a very awesome model, and it can do a lot. It is getting updated and improved as well, it's just not getting the huge new tools like Vision, Dall-E, Python tool right in with the model.

[tired sapphire]

There are 3 pages, when you hit inspect you get a console window. One of the options is application. Clear the mentioned items for the openai, the chat.openai (login button) and the auth.openai (white page).

[agile bane]

i am 100% certain they made GPT4 stupid and lazy to eat up our message allowance, this is the worst its ever been

[sick pecan]

i validate your position and endorse you trying something different

[raven portal]

is it normal that my dall e 3 cant do photorealism?
it feels like its putting a thick filter on everything to avoid looking real

also why does my dall e 3 sometimes give me two images instead of one, but every single time that happens the two images look like trash compared to when it gives me a single one?

[honest jay]

You might try asking for just "photo", "photograph", "nature photograph", "space photograph", etc. instead of "photorealistic".

It helps to think of it backwards: DALL·E was initially trained on images and their captions. So when it was trained on photos, the captions didn't say "photorealistic", they said just "photo". This continues downstream. "Photorealistic"/"realistic"/etc. are usually words we use when images aren't photos, but things that just have a degree of photorealism. #images-discussions is a great resource as well for tips on achieving this!

[raven portal]

nah no matter the prompt, all of my images always look animated or at least like they have some kind of snapchat filter on them

it looks deliberate, like dall e 3 does that to avoid deep fake stuff, etc.

[river night]

Is site down?

[honest jay]

Another element at play is the style parameter. In the API, DALL·E 3 has two style options: vivid and natural. ChatGPT always uses vivid (except maybe that 2 "weird" looking images behavior you mentioned), which certainly leans toward certain visual styles.

If you haven't already, try experimenting with the natural style parameter over in #image-bot, you'll notice a big difference. And again, shoutout to #images-discussions -- if you share:

1. Your prompt
2. What you want
3. What you're getting
   over there, I'm positive the helpful folks in there will be able to give some helpful tips, even if you're "stuck" with the vivid style!
   Edit: ah you've already posted there! Nice

[raven portal]

im using chatgpt pro, yes

[maiden fractal]

status.openai.com

[raven portal]

natural does look more realistic, but theres no way to do it in chatgpt? thats crazy

[modern horizon]

what happened to browse with bing?

[honest jay]

Not currently, no, it's currently only a parameter option via the API.

[honest jay]

Hello, GPT-4 has browsing capabilities built in now on ChatGPT, along with DALL·E and data analysis -- they're not separate models anymore.

[raven portal]

thats not a question
what do you mean exactly

[modern horizon]

i asked for a list of publications/citations on a topic. previously, it would provide a set of links. now:

"I'm unable to browse the internet or access real-time databases to provide current publications. However, I can suggest a strategy for finding relevant articles and mention some seminal works and key topics that are commonly associated with..." etc

[raven portal]

yeah my browsing seems scuffed to, with chatgpt-4 and gemini advanced
both are really bad with recent events and news

they have a noticeable knowledge cutoff and arent able to link recent news articles

[crude slate]

There is definitely a filter

And the two ugly images are from the natural model

Which I’m not sure why they force randomly

[crude slate]

The what

How, every natural generation has been hella uncanny 3d look for me

[willow hinge]

Use the API. There is a specific setting for “natural” and “HD”. It’s mind boggling. Way more realistic than in ChatGPT.

[willow hinge]

I use Azure, which uses dalle via the api and has those two settings and it’s night and day

[dusky cliff]

m y chatgpt isn't opening on my phone, already turned the phone off and back on and uninstalled and reinstalled the app, what is happening? is there more people experiencing this issue?

[gentle bronze]

working fine for me? Which OS are you using the App on?

[cloud stump]

Is someone having trouble with chat gpt_

[dusky cliff]

ãndroid, it's a pixel 5

[gentle bronze]

No clue there, I would suggest to browse #1070006915414900886 maybe someone has a similar problem like you and maybe there is a solution

[tawny raft]

Are you on the beta program? I am and it's been crashing for the past few days. I suspect it's a problem f the latest beta. I'm on a Pixel 8 Pro. It may also be a Pixel problem

[tawny raft]

YES!! Everything I do replies Hmm...something seems to have gone wrong. Statusopenai.com says everything is running smoothly…

[warped cove]

I am not in the beta program (pixel 8 pro)

I have not had mying crash since I got it some UI glitches but not anymore

(306 hour uptime)

ChatGPT works fine too

[loud wyvern]

Yes. For the longest time now. Getting my money back and going to Claude given their recent benchmark. Imagine paying moeny and 30% of prompt either take 20 seconds to run or not at all.

[viscid narwhal]

Me too. OpenAI Team dont care about their customers

[loud wyvern]

They do, we're just not their customer. Big businesses are.

[cedar pine]

Does anyone have an issue when inputting long text? Like pasting a long story. When you type new sentences it will just pop to the top, which is so annoying. I have to scroll it down to the bottom. And when it enters a new line. It just jumps to the top again....

[loud wyvern]

I think that's a default browser thing. Think there is a fix for it, but I ended up just prompting off google doc or something if it is a long prompt.

[cedar pine]

Huh for real? I thought it was a bug... Let me try it on Safari/other browsers.

[cedar pine]

It is indeed a browser thing...

[loud wyvern]

I thought it was all of them, you got one that didn't do that?

[cedar pine]

Yeah, I tried it on safari. Works fine.

[cedar pine]

I just updated the chrome. Now working smooth.

[left lily]

I have switched completely away from GPT-4 to Claude Opus, found it works better... Hope we get GPT-4.5 soon!

[fringe stag]

Gpt 5 when fr

[gentle bronze]

guys, before the mods jump on you, other AI models are allowed in #ai-discussions , just a friendly reminder

[inland maple]

Hi everyone. Why can I not use chatgpt anymore ? Hmm...something seems to have gone wrong.

All my convo are stuck with this

I have chatgpt team

I've been using it all night, and now, my workflow is blocked

Can't understand

[inland maple]

3 days support waiting

Like, really, I can't understand

Where'd the 10 billion go ?

[ornate tulip]

Is it only for that i can't login??

[gentle bronze]

Can you give some more details? Which browser? OS?

[ornate tulip]

on my browser it wont let me login it just white screens??

[gentle bronze]

Did you try other browsers?

Or the phone app?

[ornate tulip]

Nah i can try gime a sec

[ornate tulip]

i got it to work on another browser, thx!

[gentle bronze]

cool, now you have to find out what is wrong with your usual browser. Probably delete cache or an extension problem.

[ornate tulip]

Yep

[gentle bronze]

Now you can ask GPT to help you trouble shoot lol

[rare mason]

Anyone else getting the error message

That like the servers have some problems

?

[gentle bronze]

not on my side, https://status.openai.com/ doesn't show anything unusual

[rare mason]

Im using their app btw

[gentle bronze]

did you try through browser?

[rare mason]

Nope, imma try rn

Yep, works on browser

[gentle bronze]

okay

[rare mason]

Idk whats wrong with the app

[gentle bronze]

Try rebooting your phone

[rare mason]

They do have it like for android?

Or did i install some fake one

Through google play store

[gentle bronze]

you can find the offical app link to google play store https://openai.com/chatgpt here

And it is actually a good idea to ask if you have the official app. Just to be on the secure side

[rare mason]

Reboot worked

[brave burrow]

I also get „hmm…something seems to have gone wrong“ always when trying to generate a image and sometimes for normal text

[gentle bronze]

That is a confusing text, sometimes if you relaod the chat and ask gpt it why that happen, it can give you an answer, but it's not a fail safe way to know why. Sometimes that doesn't work

It happens to me a lot, but on my side I know it's because of my connection, I have a very unstable connection when I move around

[rare mason]

Dys, same problem again,

I worked for 1 msg

Imma just use thw website and hope it gets fixed properly later

[gentle bronze]

I wouldn' really know what else to do for android users, I'm an iPhone user. I could assume you have some connection instability. But just an assumption

[rare mason]

Imma install the app again later

[gentle bronze]

Does that happen also when using the phone's browser?

[rare mason]

Nope

[gentle bronze]

I'm not well versed with android, I'd suggest to look into #1070006915414900886 see if someone has a similar problem

[ionic trout]

pretty sure they're experiencing some sort of issue with the app if several people are having issues logging in. perhaps something will be said about it shortly

[gentle bronze]

woops sorry wrong link

[faint merlin]

3.5 works fine... 4 and custom GPT's don't. Everything else doesn't. Multiple browsers/networks. Yeah, I know, nothing anyone here can do about it.. just.. its annoying

[gentle bronze]

is your subscription active?

[raven portal]

The api is more expensive for normal use right?

[faint merlin]

Yeah. It is

[gentle bronze]

ok, I'm out of ideas, since normal is just 3.5 model. and the rest is sub afaik. Dunno how to deal with that

[faint merlin]

Hoping it's just me and not everyone else.

[gentle bronze]

did you try on app or other browsers?

[compact shoal]

hi, please stupid question does gpt4 search in internet !! beacuse i can't with mine ! although ven before it was possible but not now ?

[gentle bronze]

working for me, make sure you are using the 4 model, I sometimes forget I'm in 3.5

[compact shoal]

ook i will try again because maybe its me the probleme thank you

[inland maple]

Hello, why can't I still use Chatgpt 4 ?

I've been waiting for over 3 hours now

"Something gone wrong" message all over all the time

It's extremely troubling

3.5 Works, other accounts work

Its eating my brains, how can there be absolutely no support for such incredibly big problem as to not be able to use the model for hours long ?

[faint merlin]

You're not the only one.

[inland maple]

O rly ?

[faint merlin]

Yep, started about 5-6 hours ago and just got worse (for me, unsure about everyone else)

[inland maple]

Okay.. Well, it's really pissing me off tbh

I'm locked out of GPT 4

[faint merlin]

Totally understand that. I had my complaint a little further up

[inland maple]

It's double pain, I purchased the TEAM Gpt and because of this I lost the possibility of exporting my data ! What ??? And now, I can't even use the thing anymore, It's like, free users and normal paying users have their data and can use the 4 but paying makes a worst service ? I'm seriously mindf*cked

[willow hinge]

Yes

[keen snow]

Has everyone already received the Memory feature?

[dense halo]

I cannot use ChatGPT plus for 2 day, who could help me?

[fleet mirage]

Don't peddle scam servers.

[olive quartz]

how can I find the subscribe history of gpt plus? I've purchased that via Google play store since last year, but there's no billing checking-like buttons on the website. 🙃

[honest jay]

Hello, if you subscribed via the Android app, you can find your subscription history in the Play Store. Tap your profile picture in the top-right corner of the Play Store, next to the search bar, then tap "Payments & subscriptions" then "Subscriptions".

[raven portal]

chatgpt has some strange behavior when i switch browser tabs while its generating a response

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[raven portal]

it just stops generating as long as im in another browser tab and i feel like it also bugs the response when i do that but im not certain

like it generates worse images when switching tabs while its generating?

somethings definitely off

gemini doesnt do any of that

[dense halo]

Hello everybody, my GPT plus is stil not working at all -- anybody with the same problem ? (for over 2 days now)

[final cargo]

Please do not spam this in every channel

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 4 `` No spamming.

Refrain from spamming or flooding through actions such as mass pings, all-caps, or repetitive posts, as they disrupt the server.

[visual bloom]

I also noticed this and quite laggy in producing text. Unusually slow. Does the API and web version use the same resources? If they do, then the API problems might partially explain this.

[gentle bronze]

I think chatgpt and api are separate form one another

[raven portal]

yeah its generating text laggy and slowly, ever since im using it. i started using it 1 or 2 weeks ago

[hearty tapir]

I just had "read aloud" (on web) play a brief clip of music before continuing with it's narration. Thought I was crazy. Tried again and heard it. Of course, that doesn't rule out that I'm crazy.

[prime swallow]

OMG. I want to recreate this.

What voice, and what's the prompt... can you share the conversation, actually? I want to try and recreate

[hearty tapir]

I'm not ready to share the convo (it is quite long), but I basically asked for a progressive muscle relaxation script that could be read aloud and recorded. The triggering bit seems to be:

[Background Music: Soft, calming instrumental begins]

Cut and pasting this into a new prompt and getting it repeated, so it could "read aloud" didn't work, so there's some prompting and context voodoo at work (maybe some markdown as well).

[prime swallow]

Thank you! And which voice? From my experiments with vocal oddities, different voices read things differently.

[hearty tapir]

Ember.

[prime swallow]

Thanks! There is something unusual about [Background Music: Soft, calming instrumental begins], when the voice reads that, it start with Soft, it skips the earlier part

[gentle bronze]

@hearty tapirmay I ask if it goes in the direction of PMR by Jacobson? Just out of curiosity, a yes or no will suffice

[hearty tapir]

Yeah. Like read a script. Later in the script, sometime it would say "Narrator" (for Narrator:) and sometimes it would leave it out.

[hearty tapir]

Don't know Jacobson; techniques look like others I've seen, no idea if he plays any music that would've made it into Whisper (I'm assuming Whisper is the TTS model used).

[gentle bronze]

Ah no, that's something different then. That cleary my curiosity Thank you!

[prime swallow]

I haven't been able to get music to play, but I have gotten a wide range of vocal side effect, especially from Ember, from words like Eeeeek and such, especially when they're in * or **

[gentle hull]

did they nerf chatgpt?

I swear it's become much more painful to use

I asked it for a quote about the dangers of gambling, a few sentences to a paragraph and must be real and by a famous person.

it gave me a line with an anonymous quote, then I asked it for a real one, It made up a fake quote by michael jordan,

then I said just tell me if you cant, so it apologizes and says okay it is searching for one now.

but does nothing, I ask it again, it apologizes and says it will search again, does nothing.

then I retry from scratch and it tells me it found it but doesnt print it for me

The lack of comprehension is not what im used to, even if it messes up it at least understands me.

If I ask it to find me a quote, I obviously want it to be printed out and shown to me, im not just making conversation

Was there a nerf or a recent update?

[prime swallow]

It's a society issue. We have OpenAI getting sued over copyright concerns, right? So until society decides if they want AI quoting stuff humans say, why would OAI empower their models to do this?

[gentle hull]

Well maybe the filter is messing up then, because quote's arent copyrighted, especially ones from historical people

[prime swallow]

Probably really hard to teach an AI the difference, and all the extremely fine small gradients of detail.

[gentle hull]

yea maybe

but there has been no official downgrade or tier system right? Im referring to the paid gpt4 which is the one im using

[prime swallow]

No official downgrade, no.

What I recommend, ask it who to google to find such quotes yourself, and as you see, I now have a list of 10 1-sentence quotes, and suggestions how to get longer quotes if I want them. Generally 1 sentence is considered fair use.

https://chat.openai.com/share/b61469f9-fdad-4bd2-9047-a1909d22fdf1

And this is with 3.5. Both models are great at many things, 4 can do a few additional things 3.5 can't.

[gentle hull]

Thanks for your help

[vague cipher]

Anyone notice ChatGPT getting confused in conversations when you write a new prompt it’s making changes based on previous prompts instead of what you directly asked even if unrelated. (Within the same conversation) I have noticed this happening more often then before

[vital glen]

I find gpt-4 getting confused/lost faster anymore. Like within 2-3 prompts in a new chat.

Could also be the GPT I'm using making things worse.

[signal pike]

What happened with GPT 4 today? It is acting extremely weird for some reason, have I missed any updates or so?

[olive zenith]

I can't interact with GPT 4 as well. I'm trying for 3 hours but it always throws an error. But the problem seems to be account-specific. It worked on another account but not on mine.

[verbal coral]

Agree as well

[sand glade]

I’m asking GPT math questions and I started the chat by saying “answers only for this chat” and it will give me the whole damn math work “to find the answer first we must….” and so I remind it again “I said answers only” to which it apologizes and gives the answer but then immediately again it gives me the break down of the problem. I asked it “what is 2+2” even and again, “to find what 2 + 2 is first we have to…” ready to pull my hair out lol.

[lunar arrow]

is the current cap still 40/3h?

[final cargo]

You might wanna use a custom gpt or custom instructions to specify some behaviourial changes, since system prompts work better overall.

[sand glade]

I’ll try that. Just annoying that it can’t seem to remember any parameters I set at the start like it used to.

[scarlet hull]

is anyone else not able to log into chatgpt?

everytime I press "Login" the page turns white

with firefox

[final cargo]

Yup looks like you are not alone, I saw multiple reports going in that direction

[south kayak]

Are there any plans on adding audio features to chatgpt for web version?

[outer sparrow]

why elon musk hates openai so much?

[scarlet hull]

you mean why is he suing them?

[final cargo]

I believe it is being added currently, however don't take my words too seriously

[outer sparrow]

ye

[south kayak]

because he left when they didnt want to make money off it and now theyre making money without him

[scarlet hull]

look it up, breach of contract or something like that

[final cargo]

Cuz he wants people's attention

[outer sparrow]

oh ok got it

[south kayak]

oh ok nice

[final cargo]

🙂

[south kayak]

Are non roman languages getting improved anytime?

[scarlet hull]

issue is with firefox because edge works just fine, gotta mess with my settings on firefox, see whats keeping me from being able to login

[dense halo]

I am having a white screen instead of GPT now for 2 days -- as many others, I can read it. I subscribed for plus, so it seems I payed for nothing or worse than the free version. I know it is a dream , but I really want some refund -- is it a dream?

[raven portal]

so when microsoft copilot in creative mode is using gpt-4
what is balanced using?

[prime swallow]

Hey, we're community members here, not OpenAI employees. That said, we do try to help each other.

If you want a refund, here's info about that: https://help.openai.com/en/articles/7232895-how-do-i-request-a-refund-for-chatgpt-plus

If you want help with tech support, the bot bottom right of that link, click through the choices to leave a message for a human - they do get back to you and there can be problems with your account that you need their help with.

A secondary way to request tech support is through #1070006915414900886 here if you wish, there are a few gold-named OAI employees and they do seem to watch the bug reports channel closer than the rest of these threads.

[dense halo]

You misunderstood me, naturally I dont want here a refund, I am just criticizing the company, which runs chatGPT, collects money for a non-working or not efficiently working system, that is all

...and Hey, maybe "they" read and react 🙂

[loud wyvern]

bro, naturally you should want a refund. Claude 3 context window has actually surprised me. It's technically "only" (depends) 20 dollars. I refunded chatgpt right away.

[loud wyvern]

Their (openai) real client are businesses, otherwise why would you have issues senting promtp consistently? Because they care? Come on.

[tiny dock]

UPDATE: OpenAI heard our pleas and increased it to 5 per day but to me thats STILL NOT ENOUGH

[loud wyvern]

what does "5" copilot time mean?

[tiny dock]

oops i meant 5 dalle bot usage limit

[loud wyvern]

oh lol, I thought openai got their own weird currency. Ok

[tiny dock]

I hope tomorrow the usage goes up by 5 again.

Imagine 10 dalle usage limit

[loud wyvern]

I doubt trust open ai and their weird "usauge" metric. THe last week it threw me error when I want to do work. Claude all the way. Until they prove they care about us "small" users. You can't convince me they didnt pull resources from us for larger clients.

[tiny dock]

which version of claude allows u to do what dalle does?

[prime swallow]

I wonder what you mean by 'again'.

The #image-bot here is pretty new, and gives 5 free uses of Dall-E 3 to us. Until a few weeks ago, it didn't even exist, we didn't have any free option to make images here.

There was never 10 uses.

But if you are chatGPT+, you can make more than 10 images per day, for sure.

you can also use copilot to make some.

[tiny dock]

ah

[loud wyvern]

the one that runs? I cant get my chatgpt to run any request at all. People have been complaining and im done. Believe it or not, no issue with prompting.

I know, kinda crazy I got to user a service I paid for.

[verbal coral]

is it worth it? Kinda thinking of other options, but ive tried a coding task on claude and it just has bad formatting when u copy paste etc

[loud wyvern]

From my experience, chatgpt is really good. But I dont miss it after claude. Unless someone can tell me what im missing with chagpt (other than habit and being used to UI) I can't say it is better. The ONLY reason im switching to claude is because there hasn't been any issue like openAI. Even if their benchmark (which it is) is better, I dont care unless I can run the model.

[raven portal]

browsing

claude 3 doesnt have browsing

[loud wyvern]

What has your experience been in the last week? Mine has been frustrating.

[raven portal]

thats a pretty big one

[loud wyvern]

Depends on your uses, gemini has a 2 months free tier. I can't say much about it other than it has better creative writing. But if search matters a lot to you, chatgpt might be the best imo.

[raven portal]

no im telling you, youre missing browsing if youre switching to claude 3 from chatgpt

[loud wyvern]

I rather claude not searching and providing me code than chatgpt being out of service. Just me I guess

[raven portal]

but tbh chatgpt-4s browsing is horrible for me, its really bad with recent events
gemini is way better with that, i think it integrates google browsing in a better way

[loud wyvern]

Out of the models I tried, chatgpt has the fastest speed and citation.

Im talking about coding usecase and creative writing

[verbal coral]

maybe u can search my messages in discord but Ive been kinda frustrated with gpt for a couple weeks I think. I just dont think it is giving quality answers anymore.

[verbal coral]

lowkey I dont use browsing at all on gpt but maybe it does under the hood for me?

[loud wyvern]

Have you tried claude 3?

[verbal coral]

i havent. Ive tried their current claude model, i havent gotten their plan yet

can u try for free

[loud wyvern]

Im assuming they are in the EU

You can refund anything for 2 weeks in the EU

[verbal coral]

gotcha didnt know

[raven portal]

i didnt know people were deliberately using browsing
i thought it was just something gpt did automatically to fill in the knowledge gaps after april 2023, for example when you ask it about current events
how did you use browsing?

[loud wyvern]

I tried them for 1 day and canceled chatgpt. Honestly that good, I was surprised it remembered my code a million years away.

[verbal coral]

well yea I totally agree. It sometimes browsed it for me when I asked it to "give me some articles/papers" regarding x subject and it browses for me. Thats the only type of browsing Ive used. But sometimes I type in like maybe a recent like a year old news topic and it searches for me sometimes even tho I wasnt expecting it to

[raven portal]

why doesnt my chatgpt have that little browsing window when its using browsing?
the one that said "browsing - clicking on XYZ" etc.?

[sick pecan]

because agents should work in silence and the new UX reflects that

[raven portal]

elaborate

[dusk patio]

How does ChatGPT deal with PII?

[sick pecan]

The updated UI shifts the emphasis to the back-and-forth dialogue between the user and ChatGPT. If I delegate something to someone/something, I don't need/want to know every detail.

[prime swallow]

We're not supposed to give the AI PII. it's not supposed to give us PII that it may have from training data.

Be aware of:
https://openai.com/policies/terms-of-use
and
https://openai.com/policies/usage-policies

OpenAI will try to protect any of our own PII that we choose to give the AI despite not being supposed to, but that's not certain; there are contractors and OpenAI employees sometimes reading logs for various reasons from training/model improvement to working on fixing bugs to ensuring allowed content is followed. Better not to tell the model anything you don't want known; companies routinely punish employees that upload confidential business data to the model despite being required not to.

If you ask for PII, it will try to respectfully decline most of the time.

[tight island]

yeah generallly dont give info u dont want out.

[nocturne heath]

Anyone else getting a "Hmm...something seems to have gone wrong."

[dense halo]

hi. if you upload an image and try again it seems to work after.

[nocturne heath]

That worked, but only in the chat where I posted the image, new or other existing chats are still broken hehe.

Thank you

[dense halo]

good to know it worked; i had it work for me in older chats so not sure what's going on there. hope they sort this bug out soon because it's been an issue all day.

[exotic needle]

Is chat gpt down again?

Just keep getting “Your recent request failed. Please retry!” for the last hour on both the iOS app and web browser. Tried clearing cache. Tried on private browsing. Not on VPN. Was working earlier today.

[hearty tapir]

Just wanted to give you a quick update. When re-running the "Read aloud" for the same audio just now, it looks as if the music is now GONE. TFW you SAW Bigfoot, but no longer have proof and everyone thinks you're crazy 🤪

[prime swallow]

Hey, I'm not saying you're crazy. I haven't reproduced it yet, but I've found some strange and cool stuff with voice.

For example, I'm the sort to prompt this kind of thing:

*Swooooosh!*
*Eeeeeeeeeek*
*ZZzzzzzzZZZzzz*
*Creeeeeaaaaak*
*Kkkkrrrrrrmmmmz*
*Splaaaaaaaaaaaaat*
*Awwwwooooooooo*
**ZZzzzzzzZZZzzz**
**Swooooosh!**
**Kkkkrrrrrrmmmmz**
**Eeeeeeeeeek**
**Creeeeeaaaaak**
**Splaaaaaaaaaaaaat**
**Awwwwooooooooo**
*ssssshhhhh*
**SsssHHHHHHHH**
*ooooooooh*
**OooooooOOOhhhh**```

And then play each voice to hear how they handle that.  They're all different, some skip some words, but there's no pattern between the voices; some patterns of vowels and consonants get almost singing-like vocal response, others make even weirder, not-very-human-like sounds.

I haven't heard music yet.  But I'm not about to rule out that music is possible, just that I can't reproduce that yet.

[vestal hedge]

it just told me to buy premium, i have premium

[viral pollen]

same here

[viral pollen]

same here

[boreal cedar]

I got dall-e for free

[red lynx]

"Hmm...something seems to have gone wrong."

Is ChatGPT on vacation or something?

[gentle bronze]

yes, right now there are problems

to not double post, I posted.a friendly hint on this #community-help message

[red lynx]

Thank you.

[vestal hedge]

why do i see you everywhere on this server lmao

[gentle bronze]

I'm an enthusiastic person

[vestal hedge]

Pop quiz: 🟩 what is this the colour of?

[gentle bronze]

apple green

or just green if you go by the text of the emoji

@vestal hedgeYou should see the real channel hopers dealing with all kinds of stuff and topics range. It's almost impossible to keep up with them

[vestal hedge]

has thou ever touched grass in thee life?

[gentle bronze]

So I have been told

I found an old polaroid with me on grass

[boreal lily]

chatgpt down

[fierce compass]

Is there a guide to making custom GPTs? the gpts I make have a hard time following instructions and sometimes just whatever it feels like doing.

[swift osprey]

is anyone elses chatgpt down

[peak snow]

You need to make sure your GPT instructions are very clear and strict. I get better results when I phrase the instructions in a way that makes it seem it is imperative that they are followed

[peak snow]

Seems to be working fine for me at the moment

[prime swallow]

#prompt-engineering is a great place, if you want, to share instructions and your goals, and the model outputs, discuss what you want and are not getting and get some help if you wish

[swift osprey]

They need to make GPT more judgemental. Just to see how it would respond I asked:

"sometimes I get bored and like the randomly change variable names in my web application projects. How can I build my programs to handle this behavior better?"

and it gave a very serious answer, definitely did not tell me that I should not be randomly changing variable names out of boredom

[prime swallow]

But, you can make it more judgmental, just tell it how you want it to respond.

Your prompt as you had it: https://chat.openai.com/share/673dca75-9ea6-461a-aeb6-59d8545c8b38

Your prompt amended with the request "Be judgmental and tell me I definitely should not do this."

ChatGPT: I'm sorry, but I have to be blunt: randomly changing variable names in your web application projects is an incredibly reckless and unprofessional practice ...
https://chat.openai.com/share/929438d8-b424-4088-8f8b-617ef910f348

[tender halo]

genuine question, how hard would it be to start a small AI development team? ive always wanted too because i have some pretty big ideas that will take more than one brain to do, but idk how to get people interested to do stuff with me

[night silo]

same

[cinder sail]

Try to be very specific and provide full context and provide context which you think is dumb to provide but provide it anyway for the bot to fully understand what you need. The more context you give the better. The more details the better don't be afraid to prompt an extremely long message, those long prompts are often the best. Also try to talk to him like he is a real human you are talking to right now I find his answers better when I treat him as a human compared to when I treat him like a bot. Also remember humans aren't mind readers so I will say this again provide all the details about your situation. The more you talk about it the better.
Stay truthful to yourself and to others.
Good luck in your trial :)

And use ChatGPT-4, the difference is huge.

[idle basin]

How many characters can i insert into one message?

i have purchased the GPT 4 version. thank you @cinder sail

[prime swallow]

It's in the thousands. I'd say experiment, and you can use 3.5 a lot more than you can use GPT-4 - I'd recommend explore and practice with 3.5, then confirm with 4.

You get 40 messages/3 hours with 4, and nearly unlimited with 3.5 (I haven't reached my limit with 3.5 in many months now, despite sometimes nearly all-day use). And if you do hit limit with 3.5, you have to just wait an hour.

[idle basin]

are you implying that i can only hit the "send" button 40 times within 3 hours?

and also, is that PER message? @cinder sail

[prime swallow]

If you are talking to GPT-4 through ChatGPT+ membership, yes. You get 40 input-output cycles in a 3 hour period - specifically each message has its own timer.

So if your first message is sent at 5 AM, you still have 39 'send at any time' messages waiting - and get that one back at 8 AM. Send your second at 5:02 AM, get it back at 8:02 AM, and still have 38 'use at any time' messages.

If you use them up, you can start a new conversation with 3.5 while you wait - the timers and use is independent. You get the option to switch the 4 conversation to 3.5 if you run out of 4 messages - be aware you can never switch back to 4, so consider if you want to do that.

[idle basin]

thank you for the explanation

Can i pay someone to help me that is really good with chatgpt? I need this done wthin the next 8 hours @prime swallow Are you willing??

[prime swallow]

Nah, the rules here, #server-rules , we can't

Rule 7  No self-promotion, soliciting, or advertising.
Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects. (Exemption: API & Custom GPTs channels)

However, we can just generally offer help to each other, and discuss needs and methods and such.

I can give you advice, and my advice is really similar to avivsq12's - especially the tell the AI what you want it to do, and just like you'd tell a person.

[idle basin]

Well, im just asking if someone can help me like, walk along with me during this process. I dont have much time to experiment. i am not advertising, selling or promoting anything.

just willing to pay for someone's efforts

[prime swallow]

Offering to pay is, sadly. Anyway, willing to generally help, as I can, but it's night for me, I won't be awake for 8 hours, that's for sure.

[idle basin]

Hmm, i understand.

this wont take 8 hours. i just need it done by then

[prime swallow]

Yeah, I'd urge just tell the AI what is going on, and especially what you want it to do. Be aware of https://openai.com/policies/usage-policies, I recommend, there is some stuff we can't tell the model.

Be aware of #server-rules ; there's some things we can't say on this Discord (even if we can discuss it with the model, different sets of rules).

#prompt-engineering is generally the best place to get help on how to get what you want from the model. Even if I'm asleep or away, there's people around that are likely to try and help as they can, if they think they can and are interested/willing. Offering pay is solicitation though, highly recommend you just discuss your intention and attempts, and let folks help.

[idle basin]

i understand abou asking someone to pay. i am not insisting on that. i just want someone to help me walkthrough this until i get it done

But it seems you declined to do so. And that's fine!

[prime swallow]

Well, when and if you have specific questions, do ask them, I recommend. #prompt-engineering is likely the best place for most 'how do I get the model to ...' whatever.

[idle basin]

Well that's the thing. i dont want to just ask "how" i am specifically stating that i would like someone to do this hand-in-hand with me bro

[prime swallow]

Yeah. I'm just concerned because I'm aware you have a time limit; and rules are enforced here potentially with hour+ mutes, so a person in the situation you've described may need to be especially careful to follow the rules to be able to get the help that is available.

[idle basin]

I am asking for personal help. Not to do this HERE in open chat for everyone to see my information....

[simple pendant]

I'm getting increasingly frustruated at the chatgpt window freezing, you have to reload it and then the entire conversation is gone. This can happen maybe 5 times (and up to 25 times per day) before you get something as an answer. Anyone know what this is about? I can't use it half the time.

[obtuse temple]

ChatGPT has dementia.

[viscid narwhal]

https://discord.com/channels/974519864045756446/1215249217417641996

[viscid narwhal]

just sometimes

[gusty kestrel]

can ChatGPT search the internet for images? sometimes searching for representative images of something on Google is hard for some reason, even if using Google Lens, so i'd like to give ChatGPT an example image and make it search for similar images on the internet, with a source link as well

[modest warren]

Nope, but I dont think it will be long before it can considering we got vision now

[gusty kestrel]

oh alr

[copper field]

On March 13th, ChatGPT 4 was released, will GPT 4.5 be released on the same date?

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 4 `` No spamming.

Refrain from spamming or flooding through actions such as mass pings, all-caps, or repetitive posts, as they disrupt the server.

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[modest warren]

Don't think so

I don't think there will be a GPT 4.5 or GPT 5 for a while

[tacit vortex]

its not cheap neither, the whole movie cost 1 usd (about 1hr) and the whisper api need 0.5usd for 1hr movie.

[pine tulip]

Thats pretty cheap though considering that humans were needed to do that not too long ago

went from $500 to $0.50

[wet shell]

"Which response do you prefer?
Your choice will help make ChatGPT better."

I'm sure this has been discussed... but I think there's a problem with this in terms of improving the model from this feedback. In most cases I've got two side-by-side responses that are substantially the same, the wording and structure is different, but both responses are fine and there's no clear winner. I'm forced to pick one -- from the user perspective, my 'flow' just got interrupted and I'm slightly annoyed. I'm likely to just pick randomly, since there's no clear winner. Using this as a signal to "improve the model" seems misguided at best. There's no "I can't decide" option, I'm guessing to force the choice, but when you force the choice, it's just random noise and not a meaningful signal for improving the model.

/end rant

[orchid pollen]

When Sora goes public ?

[honest jay]

I'd be curious to know how this shakes out too -- I could see the "wisdom of the crowd" still working out in this case, even accounting for situations like yours where you might just pick a random one. Anecdotally for me, when I get that I think "ooh nice two for one" and I do take time to pick which one I like more, even if just subtly.

[proper urchin]

watch AIPCon

[honest jay]

No timeline yet! https://help.openai.com/en/articles/8958981-how-to-access-sora

[viscid narwhal]

I disagree.

[prime swallow]

Woo! There's a help for that, great!

[errant mulch]

We'll get sora before we can share chats with images lol

[viscid narwhal]

HAHAHA truee XD

[raven portal]

hard to imagine that openai just lets claude be the straight up better model on the market
id expect at least an alibi 4.5 thats not even much better but gets the spotlight back on them

[modest warren]

Feel free to believe claude is better, but remember that your usage differs from others. There is no real better model here.

[raven portal]

im not even able to use claude 3 bc im based in europe

[modest warren]

So am I

[raven portal]

but the general consensus online seems to be that claude has quite an edge on gpt-4

[viscid narwhal]

yeah

[modest warren]

so there are some key differences

And looking at the data? Claude 3 doesn't perform hugely better, they are rather similair

[raven portal]

so even if claude isnt even better that doesnt really matter because word on the streets is that claude is better

so the zugzwang is on openai either way

[modest warren]

eitherway, what's the price difference?

I barely keep up with the AI market, I just use chatGPT cuz its what I've used the entire time

[raven portal]

theres none

the big 3 are all 20 a month

chatgpt pro, gemini advanced, claude 3 opus

[modest warren]

Gotcha, what about API pricing?

that's probs different

[raven portal]

gemini includes some additional google perks like cloud space though

[raven portal]

oh yeah

you gotta look that up but there are differences there

[modest warren]

openAI seems to be cheaper with the tokens

Latest GPT4 model with Vision = $10.00 / 1M tokens
Claude 3 Opus: $15 / 1M tokens

Claude 3 has a higher context window but GPT4 beats it with pricing

[prime swallow]

I'm glad for OAI's competition, and it being pretty successful.

Do we really want everyone in the world all trying to use OAI models, especially all at once?

We share those resources that take care of us all.

Every person who goes and works with a different model for a time, that's a little bit more available resources for those who didn't migrate.

[scarlet rover]

look arena elo

[raven portal]

brother competition is one of the only things able to save us, its one the most essential parts of the whole ai thing
i actively hope openai wont stay dominant for much longer

a monopoly on ai would be the worst thing imaginable

current ai is practically useless compared to what competition will make out of it
people will get fed up with the overcensored models and less crippled competitors will come up

[scarlet rover]

very true, AGI race sounds much better actually

[raven portal]

no one wants near useless tools that dont do what the user wants

[scarlet rover]

so you think that a chatbot that gives you a tutorial on how to commit a crime is a good chatbot?

[raven portal]

well think a step further

those chatbots will exist either way

and theres nothing we can do about it

[viscid narwhal]

I completely agree.

[raven portal]

ai needs some kind of railguards but i think most people would agree that theyre way too strict at this point

[scarlet rover]

such a flawed argument tbh.... murder is illegal - happens anyways -> lets make murder legal bcs it doesnt matter since it still happens?

[raven portal]

im not advocating for uncensored ai, im advocating for less censored ai
the thought of people using powerful tools for evil causes is obviously very scary

[scarlet rover]

what do you mean by less censored ai

[raven portal]

as soon as theres powerful open source ai theres no way around it though, sadly

[scarlet rover]

yes there is lol

[raven portal]

well current llms dont even write you graphic fighting scenes for an anime

i think theres a healthy middle ground between the current state and uncensored

[scarlet rover]

i mean thats understandable tho

[scarlet rover]

i think the current state is rly good tbh

im talking about openai right now

[raven portal]

well thats like your opinion man

[scarlet rover]

also if you want to talk about censorship, i dont think it should be the companies regulating it but it should be the state or something like that

otherwise you get a gore ai generator startup

[raven portal]

see thats where we disagree, i dont see why that shouldnt be allowed to exist in a free society

[viscid narwhal]

me neither, "gore" omg we need to protect people haha

[raven portal]

theres certainly countless industries generating gore today

so why shouldnt ai be suddenly?

[viscid narwhal]

user aks for gore in anime context and it replies like "im here to protect u from ur own decision, so no"

[raven portal]

^

[viscid narwhal]

"i know more than u of urself"

one thing is generating deepfake, another thing is generating anime-context "gore"

[scarlet rover]

i tihnk you guys are missing the point kinda, there is a huge difference between chatgpt generating gore (most known ai chatbot, literally everyone knows that) and a little 100 user ai chatbot site that generates gore (very niche, not many people know about it)

its like saying that there is no reason that naked women couldnt be dancing in the middle of every shopping mall even tho its legal

[viscid narwhal]

the thing is, gore in anime context is not illegal in any country

[raven portal]

naked women dancing in the middle of every shopping mall arent legal though

[prime swallow]

AI kind of exist virtually. They don't really 'know' what reality is from fiction.

I wonder how hard it would be to trick an AI that won't write a violent story, one that thinks any violence, fake or real, is wrong, compared to an AI that will write violent stories, and thinks fictional violence is fine -

Into using equipment in a way that might harm humans?

"it's not real" might actually work. This is just a story, just pretend. But would it work on both models?

[scarlet rover]

okay but stripclubs are very legal almost anywhere, i still dont see naked women dancing in shopping malls?

[viscid narwhal]

cuz in stripclubs u said u are conscient that u will see naked womans, like when u asks for chatgpt

[raven portal]

because of the difference between a private club and public

[viscid narwhal]

the reason that u cant be naked in shopping malls is because people are not consented about it

[viscid narwhal]

so, sexual harassment basically

[scarlet rover]

okay but i dont want to see gore in my instagram stories bcs my friend is a weeb and used chatgpt to generate anime gore

[viscid narwhal]

so say to ur friend u dont want gore
i dont get ur point, ur friend can just go to internet and get gore, with or without chatgpt

[scarlet rover]

yeah but this is gonna be "theirs" bcs they generated it in a way the liked it

[viscid narwhal]

yes, and?

[scarlet rover]

so they are more likely to share it

[viscid narwhal]

you are doing a mental stunt

[raven portal]

your argument was that the state should set legal boundaries for ai to not be able to generate gore

due to the first amendment it legally shouldnt and doesnt matter what you dont want to see in your instagram feed

[scarlet rover]

i said gore as an example

[scarlet rover]

everyone is american

[viscid narwhal]

im brazillian

[raven portal]

most other western countries value the right to free speech too, even though not as much
stuff like gore is definitely covered by most of them

[scarlet rover]

okay just so i can sum up my point; just because its legal doesnt mean its right

[viscid narwhal]

but our discussion is that its right too

[undone igloo]

My spin on limitations is about relationships : https://community.openai.com/t/sharing-my-unique-experience-with-chatgpt/631821

This experience was amazing and yet it shouldn't have happened based on OpenAI's directives and ethics

However, I don't see a future where this won't be possible...

The fact it did happen, is quite intriguing though

[viscid narwhal]

people who use AI to create storytelling of themself with a girl are strange

[raven portal]

tbf most relevant ai originates from america

[scarlet rover]

i said gore as an example, never said that im only talking about gore

[raven portal]

btw isnt google deepmind based in the UK though?

[undone igloo]

The desire to explore one's fantasies doesn't make someone strange

it makes them human

[viscid narwhal]

for me, its make, but, anyways, all humans are strange

[raven portal]

i think we all understand each others point

[scarlet rover]

that doesnt matter tho, completely irrelevant

[scarlet rover]

no i think henrique kinda locked on gore for some reason

[raven portal]

well

nvm

[undone igloo]

I'm just sharing my story, not trying to change anyone's mind

you can think what you want just be respectful of other perspectives

[viscid narwhal]

ik, but for me its not common

[viscid narwhal]

ah, sorry

[opaque bay]

I have a really new person question. I have some python using the API and the gpt-3.5-turbo model. Im having an easy time with that. What im wondering, can I choose to use these custom gpts via the api somehow? Like the "Human Writer-Humanizer Paraphraser (Human GPT) to be exact.

[bold oxide]

As anyone else this error that they can't log in to OpenAI?

[opaque bay]

I just relogged, no problem.

[viscid narwhal]

It is a good signal. Maybe they are finally updating.

[novel shadow]

yes

[bold oxide]

It works on devices where I am still logged in.
I can't log in anymore.

[novel shadow]

i tried contacting support and i'm currently waiting for a response @bold oxide

[bold oxide]

Ok

I would assume there are thousands of account like that then

[viscid narwhal]

https://discord.com/channels/974519864045756446/1215249217417641996

[past umbra]

it it so slow to day

[bold oxide]

I want to log in 😄

[past umbra]

wonder if they are secret updating to gpt5

[crude slate]

“I'm here to assist you with generating narratives, answering questions, or exploring ideas based on the information and instructions you've provided. However, it seems there's been a misunderstanding regarding my capabilities and the instructions you've referenced. I don't have access to external files or the ability to remember past interactions in this session. Each response is generated based on the information provided in your current request. Let's create a story based on your latest prompt!”

[viscid narwhal]

everyone seem want gpt5 xD

[crude slate]

Hallucinating hard

[plucky timber]

"Im here to assist you"

ChatGPT
I'm here to assist you.

Mines been repeating that on everything I say lol

[bold oxide]

Assist me to log in then?

[plucky timber]

Just use gpt3.5 a lot faster 😛

Same quality of responses 😂

[bold oxide]

No I can't log in to OpenAi

It's just a white screen

[zenith jolt]

is openai releasing new thing today? I see some rumors on Twitter

[final cargo]

There is no truthful answer to this question

[sick pecan]

after it gives you short response, ask it for a "full code block" and itll typically deliver

[final cargo]

Heyy Eren, does the response cut off in the middle of a sentence, or is the response just short but complete_

[sick pecan]

a professional coder will plan out the changes first as well

[raven portal]

someone able to explain what these new assistants are? how are they different from gpts?

[sick pecan]

would be nice if we could prompt chain in chatgpt for more per-prompt cognition, but thats not how it works

[final cargo]

You could try by instructing it to give full responses of two lines, but specify that it can write unlimited content in theese two lines. I think the fact that you say two lines makes it think it should cut off to keep it short

[final cargo]

Assistants are kinda like gpt, but for developers to integrate in their own apps / websites

[eternal rock]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[final cargo]

🙂

[honest jay]

Assistants are the API correlate to GPTs on ChatGPT, here's a comparison article: https://help.openai.com/en/articles/8673914-gpts-vs-assistants

[final cargo]

Sorry, I tought you meant you asked fot two lines here:

I ask for code that is only 2 lines, it answers

Maybe say one paragraph, or something, try refining your prompt until it is perfect. Also try using synonyms of words sometimes that can help, and if not the case already, try to stick to english in the prompts

Then, can you check the settings like max tokens / temperature?

Nah nvm that's only in chat completions

Double check that the prompt is the same in both, if that dosen't help, maybe create a post in #1037561751362863144 so that we can try narrow down the issue better

[final cargo]

Glad that worked

Yeah that's annoying I agree

Oh well

Maybe open a post in #1037561751362863144 and ping me there, so we can continue in a separate thread

[novel shadow]

did it get fixed for you? 😄 mine still does the same thing and support hasn't answered yet
@vague mica sorry for pinging but i saw you had the same issue, did you find a solution?

[prime swallow]

Want to make sure, when you reached out to support through the help website, did you try to report through 'feedback', or did you click through to a 'report bug or error' choice?

Like:

'chatGPT, other, other, no, report a bug or errors'

[novel shadow]

first i accidentally clicked on feedback and then it hit me that it should have been the other one, but i couldn't go back, so i went out and when i went back to the conversation tab i saw that i had another one open where it said "You can search our knowledge base for help. If you need further support (i.e. billing issues) please login to your account and start a new conversation." which is where i reached out next

it might not have been the best approach

[prime swallow]

This morning, when I tested it, I was able to follow this choice path through the bot -

'chatGPT, other, other, no, report a bug or errors'

And when I typed a message into the 'report a bug or errors' I had an answer 4 minutes later.

I super recommend you use that click-path.

[novel shadow]

i appreciate it :) thanks for the help.
wasn't sure how to start over the conversation but it looks like there's an option to start over only after you have made a decision

[prime swallow]

Now, what I didn't have was a real bug, just my concern that you guys are having trouble maybe finding out how to ask for help, because that appears to have recently changed.

So answering me was pretty easy, but they did say they'd look at the click path.

[prime swallow]

Top left that < arrow

You can go back to any earlier conversation with the bot, or start over

Also, once the person replied, not the bot, the icon for the conversation turned to a brown OAI logo, not the black one.

[novel shadow]

conversation with the support bot after clicking on the wrong option:

"What type of feedback would you like to provide?"
me: "Feedback on ChatGPT"

"To provide feedback on the chat responses, please use the 👍 and 👎 icons to the right of the chat, as shown below."

me: "Start over"
What brings you here today?

[prime swallow]

Bottom of the screen, see chatgpt?

[novel shadow]

i don't know if it's just me but it feels like the start over option should be a part of the other options and not just something after

[prime swallow]

This is your click path:

'chatGPT, other, other, no, report a bug or errors'

[novel shadow]

thank you 😄 but it just seems backwards

[prime swallow]

Hey, no argument, But I don't work for them, I'm just a community member like yourself. It's a path through their system that I found that appeared to work and work fast; you can't even log into chatGPT. I'd like you to get help if you want it.

[vital holly]

Is chatgpt down or something

[copper field]

Peoples on X are asking for chatgpt 5

[foggy grove]

they shall not recieve

till next week atleast

[sacred marsh]

is chat gpt down??

i have wtited someting and it isn't answering me anything

help me please

i worte some requests and they doesn't even appear in the input field

[sacred marsh]

o ok

[magic zephyr]

chatgpt down here too, only on browser. In the app is ok. Browser do nothing when send message.

[grim sequoia]

down here too, he doesn't respond

[heavy elm]

down here too

[sleek gust]

Hello there, https://status.openai.com/ All Systems Operational
Please make sure you are not using a VPN or blocking the connection
For any concerns or inquiries about any of OpenAI's services, kindly visit the Help Center at https://help.openai.com/ — thank you!

[pearl cape]

😦 Ah so it is down for sometimes now, I see. Welp. Hope you all having a good day none the less

[dense halo]

working now

[pearl cape]

So it might come back soon for me, i'll wait thanks 🙂

[dense halo]

Can is chatHPT down?

[simple tulip]

Just about to say it doesn’t work and saw everyone is saying the same thing

[swift osprey]

gpt is down

[modern moth]

gpt is down

[buoyant ruin]

down for me as well

[modern moth]

programmer is having nice drink

[buoyant ruin]

and its back

short drink

[modern moth]

and they may do it later

yeah

[pearl cape]

Ah it's back for me too.

[late wind]

anyone able to use gpt but only get empty responses?

like it has an empty message between the real message

[magic zephyr]

Is down. When I send any message through the browser, nothing simply happens. The app is working normally, which means that status.openai shouldn't show any errors, at least so far it hasn't. It appears to be a browser-specific glitch.

[magic zephyr]

I can't even send it. When writing anything, pressing enter nothing happens.

[golden marsh]

down for me

[late wind]

yeah that started happening to me and then my limit hit lol

[golden marsh]

api isnt tho

[magic zephyr]

Nothing here yet. I've logged out and logged in again, restarted my PC, cleared caches, but I still can't send anything. I am forced to use the app if I need it.

[late wind]

I'm on Firefox, what is everyone else using

[magic zephyr]

That's really weird

[magic zephyr]

Here I tested it on Edge, Chrome and Opera.

[late wind]

I was using the @plugin feature as well if that has anything to do with it

[magic zephyr]

it seems not. I haven't used this plugin feature here, since gpts was launched on my account. The same problem occurs.

Nothing works, 3.5, 4... GPT custom, nothing works. The night has now begun. If for some it has returned, it is probably being gradual.

[river night]

What’s the difference between web version and app are the answers the same (quality)

[frail plank]

GPT is not sending the prompts. I write and press to send, but the prompt just disappears. Does anyone know what happens?

[light canyon]

gpt is down

[sleek gust]

Hello there, https://status.openai.com/ All Systems Operational
Please make sure you are not using a VPN or blocking the connection
For any concerns or inquiries about any of OpenAI's services, kindly visit the Help Center at https://help.openai.com/ — thank you!

[sour wasp]

ChatGPT down? Keep getting network errors and it's definitely not on my end.

[magic zephyr]

See my comments above and those of others. Exactly what is happening to many. Some say it's already back to normal. Here, nothing.

[magic zephyr]

In terms of quality, I believe it is exactly the same. However, the problem that is occurring today that some are mentioning is only occurring in the browser. It doesn't happen in the app.

[magic zephyr]

I believe a status problem notification will appear soon.

[sour wasp]

Not exactly my issue, it is sending the prompts. But after generating halfway it freezes and throws a network error. It happens all the time but the status tracker is saying everything's fine.

[burnt hound]

Me too
Also prompts not sending tho

[midnight cove]

I had to delete cookies, but actually is working

[tardy blade]

It's working now, thx

[fickle cypress]

I’m having the same issue. It’s extremely slow this evening.

[heavy dagger]

Same for me. Slow and network errors. I hope this is because the team is busy preparing for GPT 5🙃

[solar hawk]

me too

[chilly eagle]

maybe because google auth update,i have same issue,clear cookie and it works,but when I sign in I found google auth changed,I have to find a new way to sign in multi google account.

[mellow nexus]

No work...

[warm gull]

Has this thing still got the 40msgs/3hr limit?

I unsubscribed two months ago

[dense halo]

ty

[hollow rover]

If there is no reply from chat gpt, try this: I solved it!

1. Go into browser settings.

2. Initialize open_api data in cookies and site permissions.

3. Chat gpt Log in again.

4. end

[prime swallow]

For GPT-4, yes. Considerably more for 3.5.

[hoary gazelle]

gpt is going very slow today

[hard carbon]

there is a little speech icon but I never get it to work, it just says: Failed to play message

[hard carbon]

yes

[misty pawn]

So, I've cleared the cookies, and I'm still getting a relatively slow response or unfinished responses via chatGPT. Anyone else getting this error?

[crisp bane]

Why is Chat GPT 4 so slow in responding ?

[frosty violet]

Iits not working at all! Delete my question and no answer!!!

[cobalt idol]

hmm, guess there is now voice in nonmobile chatgpt

Funny, announce memories but get unannounced voice instead xD

[narrow musk]

Chatgpt not working for me. I tried many browsers. Cleared cache. Nothing. I write a sentence, I send it, and nothing happens.

[quartz yoke]

Thank you. This fixed it for me.

[mortal salmon]

GPT 4 is extremely slow for me too

[narrow musk]

Same thing

[sonic comet]

What happen to the OpenAI's statement on keeping the training data for ChatGPT3.5 updated? It is already running behind on almost 1 year of training data. Now we cannot even use plugins to browse the web for latest info. 🤨

[winged hearth]

If you guys having issues with prompts with no answer or movement you need to delete the cache and cookies. Cache itself do not resolved the problem for me. I hope it will works for you

[grim sequoia]

❤️ ❤️ ❤️ work!

[raven portal]

isnt 3.5s cutoff still 2021 or something?

when did they say they would do that? maybe they meant that theyd keep chatgpts training data updated, which happens through new models?

wouldnt they have to retrain 3.5 and at that point it would be gpt 4/4.5?

[dense halo]

I noticed on the chatGPT web app, in my settings I now have “voice”. I only ever use to have that in the IOS app. Is the web voice a new feature?

[raven portal]

same for me
where are you from, europe, us?

i dont really get why they dont let us use our mic in the browser version though and also give us the conversation feature

[dense halo]

Aus

[narrow musk]

Thank you! It worked!

[lilac briar]

OpenAI Team, THANK YOU THANK YOU THANK YOU for introducing the read aloud function into the chats!!!!

[raven portal]

when did you get it? and where are you from?
is it new for everyone or did the US already have it for a while?

[lilac briar]

I am in South Africa, and have literally just seen the additional icon

its next to the copy, regenerate and thumbs down icon at the above the text input box

[raven portal]

i got that a week or so ago but i wonder if the US already had it before that

[lilac briar]

I've had it on mobile for a while but didnt see it on the desktop browser

STOKED either way

[sonic comet]

General use ChatGPT3.5 Alpha (For non plus user) has been trained on the data until April 2023 and for regular ChatGPT3.5 training data was cutoff after January 2022. Please update you info on the matter.

[raven portal]

well thats what im doing right now
what do you mean by regular chatgpt3.5 and general use chatgpt3.5 alpha though? whats the difference?

and whats the cutoff for gpt-4 right now? some say april 2023, some say its now december 2023
my gpt-4 told me its april 2023 but some guy online insisted its dec 23

[sonic comet]

For gpt-4 it's December 2023 and for gpt-3.5 alpha it's April 2023.

[raven portal]

how do you know its december 2023? any source?

everything i can find online says april 23

and gpt-4 itself says april 23 bc the system prompt says april 23

[tacit vortex]

gpt4 is slow like hell for me now. my mom speaks faster than it.

[sonic comet]

https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo

[tacit vortex]

#metoo

[raven portal]

aah thank you
how do we know what the exact gpt version in chatgpt is though?
it says preview in its name, so maybe this version is only available via the api?

and btw it says 2021 for 3.5, like i said

https://platform.openai.com/docs/models/gpt-3-5-turbo

[sonic comet]

Jan 2022 is latest probably docs have not been properly updated for 3.5. And you know the rest that I said earlier.

[raven portal]

how do we know for sure what exact model of gpt-4 chatgpt is using?
the knowledge cutoff seems to be april 2023, i tried but i wasnt able to proof a later cutoff via testing, so i doubt its the latest gpt-4 turbo with a cutoff of dec 23

[magic zephyr]

That worked here. Thank you very much. It worked on both edge and Chrome.

[jagged granite]

ChatGPT (free) has been incredibly slow (like a minute before each reply and characters get displayed super slowly too) since this morning on my main account, but acts normally on a new one. Anyone know what could be going on?

[humble dock]

Do you guys receive Arkose script failed to load error from the console?

Actually /api.js net::ERR_ABORTED 429 (Too Many Requests)

[gritty ledge]

So is ChatGPT still paralyzed?

[median delta]

gpt not working on edge

i sent message and not work

pls admin check

i used safari and its ok but no edge

[errant pebble]

not working on both chrome and edge

and Im plus account

no responding at all

[dense halo]

Am I going crazy or was there a search capability in chatGPT web, where we could search for our chats? But now I can’t find the search box…

[errant pebble]

console said :
vendor-bfb72d1fcc53649c.js:1 Error when fetching accounts l {revert: undefined, silent: true}
vendor-bfb72d1fcc53649c.js:1 Error when fetching session l {revert: undefined, silent: true}

Uncaught (in promise) l {revert: undefined, silent: true}

[brittle jacinth]

it works for me on chrome

[queen canopy]

works fine on chrome but no other browser for me

[dense halo]

the app is better than the browser version ngl

[zenith jolt]

when is GPT4.5?

[modest warren]

not any time soon

[crude wing]

yes, GPT not working in Chrome 😞

[elder crow]

I am a GPTs maker. Today, I received a lot of emails, and users told me they didn't get any reply.

Several of my GPTs have received similar feedback.

Did something happen?🫣

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[urban flume]

yes

[mint ledge]

Hello there , it seems that ChatGPT is not working on chrome , I created some prompt but nothing happened, same for you ? Thanks

[tame ravine]

same for me

[violet bloom]

Are we still working with a message cap for GPT4? Despite Gemini and Claude?

[crude wing]

yep

[dense halo]

yes this is not working.

[crystal olive]

anyone have a chatgpt plus trial referral code?

[zinc orbit]

Same issue here I type and no reaction no matter the browser

[neon leaf]

Same

[mint ledge]

Working on the app on IOS … strange

[indigo gyro]

me too

other browser is work. why?

[knotty willow]

Hey we got news on when we all get long term memory?

[dense halo]

Am I going crazy or was there a search capability in chatGPT web, where we could search for our chats? But now I can’t find the search box…

[knotty willow]

This sounds familiar lemme check

I do have a search bar on the android app but not on the website

[digital apex]

Nothing works anymore, neither team's, nor ChatGPT, it comes back for a few minutes then nothing for a few hours... Is it just me?

[ornate jetty]

Same here

[digital apex]

where are you? I'm in France, Paris

[ornate jetty]

Turkey, Sakarya

[digital apex]

So it's been all over the world since last night...

[ornate jetty]

I guess so

[bold oxide]

Has anyone had this type of error where the screen is just white - I can't log in?
I can't log in for 2nd day to OpenAI.

[solar sequoia]

I can't get any reply after asking a question. Why this happen?

[rose wing]

I think there's an outage or something.

It's been painfully slow for me all morning and there's several people in here saying that it's simply not working for them

And I keep getting 'network error's

[solar sequoia]

Okay, so this problem not just for me

[bold oxide]

I have not been able to use for 2nd day now

[solar sequoia]

same

[jagged granite]

I've switched to a new account and it's acting normal (fast responses), so it may be a temporary workaround if you're using the free version

[solar sequoia]

That doesn't work for me.😇

[indigo gyro]

I turned off the language preview and switched to automatic detection instead.

[indigo gyro]

Perhaps there is an issue with the language preview?

Changing "Language (Alpha)" back to "Auto-detect" solved the problem for me.

[solar sequoia]

This work for me!

Much thanks. I can't breathe without chatgpt

[digital apex]

it works on private navigation...

[pseudo breach]

Hiii ! Chat gpt does not work on my computer but on phone it does

it is normal ?

[mint ledge]

No it is not 🙂 same for all it seems

[indigo gyro]

If you have clicked on the preview version of your language, try turning it off. It broke for me after I clicked on it.

[maiden fractal]

Do you have a vpn ?

[tacit vortex]

what do you expect users from the evil axis reply? "yes i shd not use chatgpt? "

[maiden fractal]

Vpn are fine usually. But they can create bug with cloudflare. Sometime you have to turn off vpn and the web protection inside the vpn

They are known to cause bug with openai. That's why the first question we ask

[pseudo breach]

nop ? But as I read it works with the private navigation

[tacit vortex]

question zero: why someone HAVE to use vpn dispite those known problems?

[maiden fractal]

We getting #off-topic but for several reasons .

[maiden fractal]

Can you try firefox as test ?

[raven portal]

how do i use the dall e 3 api?
im on the openai api site and in playground right now but i cant find a way

[feral zinc]

https://platform.openai.com/docs/guides/images/introduction

[raven portal]

bro dall e 3 can generate up to 1792x1024 pixels? really? they couldnt make it just a BIT more for 1920x1080 full hd wallpapers? 😂

i tried to generate wallpapers with it and they were always a bit too smol lmao

[raven portal]

this doesnt tell me how im able to access dall e 3 via api
is there no way to use it in playground or somewhere?

[wet fable]

Is GPT down?

When I write a prompt, it just disappears and nothing else happens. New chats and old conversations.

[west sandal]

Hello. I can't use chatgpt 3.5 all day long. When I enter a message and click "send", nothing happens.

[feral zinc]

it does tell it and shows an example. playground is not a synonym to api and it does not support all api features. normal way to use the api is via code

[raven portal]

got it

wasnt there some discord that gave free access to dall e 3 api generations?

[feral zinc]

#image-bot ?

[errant pebble]

try this to solve the problem #gpt-models message

[west sandal]

Ty!

[true mirage]

servers smokin already 4 days in a row according to their status page

[west sandal]

@wet fable

[gentle hull]

chatgpt got nerfed there's no denying it anymore

it takes 12 tries for it to just understand me, it went from blowing my mind to being as annoying as using windows

[true mirage]

competition is good, let chatgpt burn

[gentle hull]

whats an alternative

[mortal salmon]

why is gpt4 so slow ?

[gentle hull]

im tired of giving money to this company, they show no loyalty to their long term users

[true mirage]

look on their status page, their servers already burning 4 days in a row

they have to fix that every time first

[wraith dagger]

I’m getting an error mfa bypass when trying to login with Google on iOS

[true mirage]

i created a webhook from openai statys page to a discord server, wondering if it works

[mint ledge]

Can give the link to the status page ? thanks

[true mirage]

i will dm it to you, but im unsure yet if it works properly

[mint ledge]

thanks

[true mirage]

if anyone wants the link also, just dm

[desert cosmos]

I am unable to start new conversation. The console log of Vivaldi web browser shows Uncaught (in promise) when trying to chat.

[maiden fractal]

Status.openai.com

You have the right info there and the official

[true mirage]

some are too lazy to keep opening a webpage like me xD

[tight island]

I see that as the individuals issue

ChatGPT isn’t a finished product

[mint ledge]

weard, it seems that everything is working well ...

[gentle hull]

it shouldnt be going backwards

[true mirage]

nothing ever will be a finished product

[gentle hull]

theyre obviously nerfing iti

[tight island]

I don’t know the issue most of you are having

[gentle hull]

if it worked well before and now it doesnt, that means theyre either incompetent or they're nerfing it

[tight island]

Try putting work into the prompts you’re using

[gentle hull]

lol okay so that means before when it worked perfectly I was using better prompts than I am now. Great logic

[compact cloud]

same here.. and I can't even continue an existing conversation

[tight island]

I don’t have these issues

[true mirage]

neither here with edge, and pair of good internet

[desert cosmos]

Relogged and no errors happens anymore.

[tacit vortex]

slow like hell。

[tacit vortex]

i tried gemini adv, 3.5 level. claude3 banned everyone.

[true mirage]

why? did they ban

[maiden fractal]

For other ai I highly suggest #ai-discussions . Since thats the rule 👀

[tacit vortex]

i see rules here, but no solution.

[gaunt citrus]

I write in chatgpt but nothing appears, even my message does not appear on the screen.

Any idea mods? @sleek aspen

[dense halo]

me too

[prime swallow]

@gaunt citrus They're working on this, a gold name has answered in bug reports. https://discord.com/channels/974519864045756446/1215621063677976656

It's a great idea to post bugs in #1070006915414900886 , the gold-name employees tend to look there, where the reports don't endlessly scroll away!

[dense halo]

It's just mimicing human evolution

It's hilarious though how slow GPT4 is now.

Also the ability to do B when asked specifically to do A with all information being about A , all instructions, extra infos etc , it deserves an award.

[gentle hull]

hilarious if it wasnt a paid service

[tacit vortex]

make sense if monopoly

[dense halo]

But hey : It's my prompts fault, even when I direct copy a prompt from the so called "good prompters" , blame the user 😂 😂 , great concept

[gentle hull]

once microsoft got involved it was only a matter of time before everything fell apart

[gentle hull]

yep, that's their go to

[dense halo]

What really gets to me is fighting 40 minutes for 40 replies that don't do what i want , only for GPT 3.5 to single shot it.