chatgpt-discussions

[lyric token]

They got me

OpenAI got me

[sharp blade]

somebody knows why when I'm adding actions to my custom gpt and i test it it calls the endpoint and then say "Stopped talking" and i'm not seeing the errors why it "stopped talking"

[real thunder]

anyone?

[plucky timber]

I think important reminder for OAI devs : If you diminish the service to a point where googling is faster, people will start googling again ;P

My usage of OAI shows this problem clearly where I am now doubting if I truly need it any longer as the waits for an answer are similar to waits at the dentist

[dense halo]

Preparing lecturer notes, I do not really understand why GPT give me the last sentence seperated into 2: "Since finding derivatives of composite functions requires symbolic computation, which isn't natively supported by most standard programming languages or deep learning libraries like PyTorch and TensorFlow, we typically have to manually script the gradient expressions for modular functions. However, for Python users, the SymPy library offers a solution, enabling the automation of this process. This means that we can use SymPy to automatically calculate analytical derivatives for these functions." Is this common way to say such a redundant sentence? Maybe too oral? I am non native, not sure why it happens.

[plucky timber]

What it looks like right now is that they did undo it and now redid the limit

When the teams tier was added you could use custom for the full limit

It seems from all im reading in other channels they re-undid that

Like I mentioned above , diminishing the service to a point we don't need it.

[dense halo]

I think the current limit is 20 at least for me. It shows capped at 40. But every 3 hours I can only ask 20 questions (no retry no re-ask). 3.5 is definitely getting better in doing simple tasks. At least it is not nagging, and unlimited.

[plucky timber]

Yeah its a bit rough atm, makes me wonder if Teams didn't sell well.

Which could be the case as its aimed on multiple people and I bet most the requests for increases (enterprise requests) they got were people pretending to be a company.

So they might've thought to solve that with Teams but missed the mark as the customers really just wanted an increase for 1 account and were/are willing to pay for that

/rant out imma get coffeee

[swift osprey]

I could rant about openai / gpt for days

just as frequently as I use it 😮‍💨

[plucky timber]

It do be like that, want some cofee too? 😂

[swift osprey]

haha just finished mine, too late in the day for another

[plucky timber]

Fair enough!

[sharp blade]

could someboddy help me?

[swift osprey]

I've never built a custom gpt before

[sharp blade]

ah okay

[swift osprey]

could try asking gpt your question, not sure if it can help though lol

[dense halo]

extactly

i still dont have chatgpt stpre

[swift osprey]

I think all they care about is funneling businesses to use their API (which is what Teams version uses), where GPT is far, far, more expensive to use. I think us regular users just provide them training data

[grim dove]

I still don't have access to gpts store, does anyone know if there is something I can do about it ?

[loud flare]

what is this madness! after 10 messages gpt plus Unable to load conversation 27b2-8c5-668f8d0f! for what I paid for the pro version!

[tight island]

although they can be expensive, they are also vastly better, API that is, API cannot be compared to the subscription.

[swift osprey]

Agreed, I much prefer the API

[tight island]

totally agree.

API for heavy tasks and customization.

[swift osprey]

It seems like the API just tells you the important stuff whereas GPT will tell you the same thing but bury it in 10 paragraphs of fluff material

[tight island]

well yeah

[deep flax]

if only there was a way to make the AI reply in the way you want by giving it instructions...

[swift osprey]

If only GPT would pay attention to my 917 characters of "What would you like ChatGPT to know about you to provide better responses?" and 1397 characters of "How would you like ChatGPT to respond?"

[tight island]

thats true, i think we know this, it was more in regards to the default model

[runic spruce]

Hi all, not sure where to put this – looking to get access to the Enterprise plan for my company. While Teams is great, we need the other tools available for Enterprise for compliance and having a hard time getting a hold of anyone to start that conversation (submitted multiple requests over the last few months with no dice). Does any one have any other suggestions?

[tight island]

contact customer service

nothing else u can do.

[tacit vortex]

the gpt4 now got too many errors and too limited msg cap, now programmers retain their jobs. nice work openai.

[tight island]

u could contact sales department as specified.

[runic spruce]

as I wrote, our org has been trying for literal months LOL

If anyone finds this, zeus speed! ^^

[tight island]

maybe you're simply too small compared to the "big" companies on it. maybe they need meetings for new enterprise users.

[runic spruce]

Not a bad point ^^

Appreciate the response Starlight 😄

[tight island]

np, i want Enterprise too

but i am 1 guy company

[runic spruce]

we're at hundreds, so a bit confused :p

[tight island]

so i reckon u sent an actual mail.

did you try the support interaction function on the site

[runic spruce]

Snail Mail?

[tight island]

theres a "live" chat where they will respond when they have time live. under api on openai.com

[runic spruce]

ahhhhh, so keep an eye on that potentially?

[tight island]

yeah

try writing to them regarding your mail

[swift osprey]

The two times I contacted openai for GPT support it took them 3 or 4 months to respond the first time and 3 weeks or so the 2nd time

[tight island]

when i used the function i talked about i got answer in like..

5-15 min

[swift osprey]

it would be hilarious if they respond to snail mail faster

but it's possible at the speeds I've seen

[runic spruce]

😅

Oh, that helped Starlight! From the bot:

Due to the high demand for ChatGPT Enterprise, we are starting with companies who can purchase 150 seats. We also have a self-service product called ChatGPT Team that will service companies who need fewer than this.

[tight island]

u gon buy the 150 seats right?

its gon be an investment

[swift osprey]

152 for me and starlight

[tight island]

yeh we'll buy our own seats

[native crane]

why doesnt chatgpt tell me if ive hit the cap and when i can use it again? now it just returns Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[tight island]

ill paypal you

[runic spruce]

:p reseller, you're all crazy

[tight island]

ofc.

we want enterprise

[swift osprey]

😄

[tight island]

a little crazy never hurt nobody.

[inner linden]

is there a way to make my GPT interact with each other in a discussion with the "@" mention, like having a request and each of my @ GPTs give me answers ?

[tight island]

thats more like a plugin

if u want several instances give their thoughts at the same output

[tight island]

did u find the location i talked about, i can send you some screenshots if u want

[still hound]

errors again????

[inner linden]

for example ?

[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.

[tight island]

this requires API, and a python enviroment

You've reached the current usage cap for GPT-4

teams and personal still too little

this is exactly why i need enterprise rn Sam.

[swift osprey]

codegpt is my favorite way of using gpt api for coding

[tight island]

i agree i made a memgpt using only code interpreter. on chatgpt

although it aint perfect it works, unlike my 2nd project.

[swift osprey]

have not heard of a memgpt yet

[tight island]

basically extended context length

using only the model and its tools.

[swift osprey]

interesting.

[tight island]

to a degree

[swift osprey]

I need to finish this webapp so I can pick up an AI focused project

lots of fun stuff going on out there

[tight island]

i just want... a fully new updated ui and more context length

and a more extensive CI Enviroment thorugh the web app

[swift osprey]

Yeah it would be awesome if they built a UI for coding

[tight island]

i could just do api but rn im hooked on the web version.

[swift osprey]

I just do API when I run out of requests or GPT web version is frustrating me

api just too expensive to do all the time lol

[eternal rock]

`` Server Tip: ``

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

[tight island]

that is true

having an actual good UI would help alot

if OAI made an official UI for API that resembles the current one it would be great.

the fan made ones leave much to be desired.

or rather a 1:1 replica of the web ui

[swift osprey]

Lots of good things they could do for regular GPT. Everything they do indicates that they don't really care a whole lot about regular GPT and they really care about business usage

[tight island]

i understand them, they're focused on innovation, and innovation often costs money

and a damn lot of it.

[swift osprey]

just the way it goes

[tight island]

tbh i would rather see an update to dalle 3 rather than gpt

[swift osprey]

the GPT teams website screenshots looks like they have exactly what you are describing lol

[tight island]

its still to little all settings are for "Enterprise" only. not enough customization.

i wanted a temprature slider. for specific models.

[swift osprey]

I'm sure we'll see some trickle down improvements over time anyways

[tight island]

true

if i could dynamically adjust the temprature of a GPT on the webui it would be nice, especially based on the task.

like its instructions should be temprature '0'

no room for interpretation what so ever.

uknow?

[swift osprey]

I think so

lol

[tight island]

it has a bad habit of interpreting its instructionset instead of following it, gets worse the longer the instructions.

[swift osprey]

I don't know if I know enough about what goes on under the hood of LLMs to know exaclty how that works

[tight island]

its like "do x" but it has thought about doing x but didnt do it anyway.

it did "z" instead of "x" despite "z" never being an option.

[swift osprey]

If you mean as in, how the web version is clearly programmed in a specific way where it generally responds with:

• repeat your message back to you
• about 10 short paragraphs, bullet points, and/or number list
• conclusion

and in custom instructions you try to tell it to be be short and it doesn't listen

[tight island]

or the contents of an entire file it has as knowlegde.

it insists of heavy summarization.

i snaked my way around that, and managed to get my task done. took extreme refinement.

[granite terrace]

Hi, am I the only one who has not gotten access to the @ feature yet ?

[tiny pivot]

I had to log out, clear cache and log back in and I got it

[still hound]

anyone keeps getting error: Hmm...something seems to have gone wrong.?

for GPT-4

[granite terrace]

that's what I did multiple time but nothing 😅

[frank wolf]

What’s the best automated coding gpt

[azure sonnet]

Me!

[past umbra]

just me or open ai so buggy to day

[runic spruce]

I found it, thank you so much Starlight!

[tight island]

np

[surreal nexus]

Is CHATGPT-4 Turbo only accessed through the api guys?

Do Custom GPT's use Turbo?

[sonic folio]

only through API? No.
Custom GPTs use turbo? Yes.

So does ChatGPT-4.

[surreal nexus]

Curious as I'm trying a memory type custom GPT. Apparently Turbo is pretty darn good at remembering longbterm

[sonic folio]

Only enterprise and API get 128k token context, other products are 32k tokens and under.

[surreal nexus]

Hmm. ChatGPT informed me it was 128k for turbo

[sonic folio]

The model has no training about what it is whatsoever.

[surreal nexus]

Rodger

[sonic folio]

It gets a system prompt. It's lucky if that's specific about the version number.

Let me check.

This is the current baseline ChatGPT-4 Turbo system prompt, afaik:

You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture. Knowledge cutoff: 2023-04 Current date: 2023-11-03

Current date will be current, ofc.

It has no idea what it is.

[tight island]

imagine giving it the data so it actually knows more about itself.

i did that with CI, its amazing

[surreal nexus]

"Yes, the ChatGPT-4 Turbo model does indeed have a context window of 128,000 tokens, equivalent to about 300 pages of text."

After searching the Web with gpt-4

[surreal nexus]

😎..

[tight island]

i gave GPT all info about CI on a MyGPt

its absolutely astonishing.

[surreal nexus]

In what way?

[tight island]

just the fact it knows every single library it has and what they do and how it can use them, and what the limitations are of its enviroment makes a HUGE difference

i tried having it edit an image, before i gave it the info, and it was eh.. bad, after the info it was a solid 8.5/10

couldnt have done it better myself.

[surreal nexus]

Could you do a discord homie a favour. Write a short prompt & I'll screenshot this chat to get some info from my GPT. I might add it to my own parameters

e.g., GPT: search info for CI, I guess would do it. I'll try

[sonic folio]

GPT-4 Turbo by API always has had. If this is true, it's really new, like with the @ mention feature new.

When it rolled out, and up to recently, it had a 32k message context.

https://openai.com/chatgpt/pricing

[tight island]

i would recommend just googling it.

[sonic folio]

According to that, it's 32k for Plus and Teams, 128k for Enterprise, I think.

Not sure what your web search was referencing.

[surreal nexus]

I've quizzed it a little more, sounds like you're Right DarthG

I didn't clarify I wasn't using the api

[sonic folio]

I just want you to have accurate information. I appreciate the admission, but it's about promoting accurate information!
Thanks for looking further into it.

[strange valley]

chatgpt is too lazy now i said do not skip anything, keep all of them in even it is repetitive. And 13/13 was skipped content

[crude slate]

Oh what do you mean by skip

[strange valley]

message 1
...(ingore the same part)
message 10

[crude slate]

Example?

[strange valley]

when you send a long code ask to fix it will skip a lot of code, said not change here(i can under stand if he skip from a part, but he keep skipping from middle of a part)

[crude slate]

Oh yeah it does do that… with my text too

Ask it to repeat word for word

[sick pecan]

thats because its thinking about the code update. once its done you can just prompt for the full code update

also toggle on the "AAlways expand code output" option in settings

[spice star]

gpt-4-turbo is available to all plus users, but the 128k context window is only available to enterprise users unfortunately

chatgpt plus has a 32k context window

[glacial hollow]

Does voice work on PC?

[honest jay]

It's currently an app-only feature.

[zenith oasis]

Is there a way for chatgpt to access information from a previous chat prompt I made with it

[crude slate]

Actually is it possible share a chat with the ai?

[tiny salmon]

Yes

[trail mango]

the custom instructions don't respond as what I exactly need them to say, which is super annoying, because it's own though of respond me always miss my main request. everyone feel the same?

[crude slate]

How?

It doesn’t seem to be able to visit a open ai link

[tiny salmon]

Manual copy and paste into a text file, lol

Or paste directly into the chat

But, you would be better off summarizing it unless you need it to specifically have a detail for it.

I'm actually having trouble with this right now because it has a self imposed 90 word limit on a summary.

My chat is 140,000 words. 90 words ain't cutting it, lol

The other problem is the token usage to summarize the chat. You can end up running out of memory just summarizing the chat for context depending on the content.

[dense halo]

nah i'm so tired of openai policy

Where should I contact? It keeps blocking my connection

[tiny salmon]

Tell me about it. I told GPT that if this continues people are going to start going to Claude and Bard. It told me to explore those options, lmao

I love the customGPT feature and I wish I had time sooner to work on mine before the changes to make this a less painful process.

[dense halo]

I think customGPT and GPT-Plugin features cause problems.

Openai making mistake to me.. continuing to block innocent users

[tiny salmon]

CustomGPT is awesome, but also can be tedious depending on the intended use. Plugins are hit or miss. GPT classic does seem to work better.

[tiny salmon]

Yeah, being told that my uploaded file which is my conversation between me and my customGPT is not my intellectual property is really annoying. I'm not violating copyright or trying to plagurize. Why am I being punished?

[dense halo]

Captcha authentication is required for each question. And Our systems have detected unusual activity from your system. Please try again later. errors in all questions

[vivid pier]

same thing here

funny thing is I get a network error on macOS, but if I move to my phone and open the chat on the same wifi, the full response is there waiting for me

[ornate marsh]

Anyone is getting "something seems to have gone wrong" error message for GPT4?

[lofty raptor]

Hi is there anybody knows, i was in restriction of chatgpt, means its available in that country, can i use that country card for payment and subscription or it also can not? Another question if credit card visa is issued from that restriction country ? Thanks

[magic pelican]

pleasea give me more messages in chat gpt 4

40 messages for 3 hours is not enough

chat gpt 3 is for free and unlimited messages, do the 40 messages rule for them

[dense halo]

🤦🏼‍♂️

use the API or upgrade to the team plan for 100 messages

[magic pelican]

how much is the 100 messages?

[magic pelican]

i do not have a large projects to use api , just sometimes i need more than 40 messages in an hour

[dense halo]

you don't need a big project to use the API. my chatbot is a personal project and I use it sometimes if I don't want to use ChatGPT. sometimes it's better to use the API than ChatGPT to avoid those message limits + you pay for your usage which doesn't cost much (depends on your usage/models/tokens etc)

[magic pelican]

oh nice, i thought it is only for big company and larg projects, then i will search how to make my own API

[dense halo]

ChatGPT coded my chatbot's frontend and backend code so I recommend you to ask ChatGPT

you need a web server to run it on localhost. you can use NGINX

I gtg now if you need help feel free to dm me I'll reply as soon as possible

[magic pelican]

nice idea i will ask chat gpt to understand this more, and thank you for the recommendation i will look into NGINX,

[dense halo]

you're welcome

[magic pelican]

thank you, have great day

[dense halo]

thanks you too

[stark rampart]

Does the update between mobile terminal and webpage have a delay?

Why I subscribe on mobile phone my webpage didn’t upgrade to 4?

[arctic leaf]

Are you sure both platforms are using the same credentials / account to log in?

There shouldn't be any delay or difference as far as I'm aware but if the account is the same you may need to log in and out again or clear your browser's cache.

[maiden fractal]

Yes

[stark rampart]

Oh ，it works now. It does have some delays

[dense halo]

whether you subscribe to Plus on your phone or on the web its all synced and updated on their database in real time

[shy ginkgo]

hi all, does anybody experiencing ChatGPT problem? mine is not answering, only show this: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[lapis ermine]

**Message limit has got much smaller. **
Looks like they reduced it, so they could give the team plan more messages.
Looks like they've removed the version number and message limited info, to hide the plan reduction and lack of development from sight.
Pretty annoyed.

[hollow burrow]

yeah seems so, they mentioned that it is dynamically adjusted to suit with the current server load

I have hit my cap, and i can understand how you can be annoyed, but we need to sit back and wait

[hollow burrow]

have you tried another browser. that error sometimes can be a cause of browser/deivce compatibility issues or cache

[split blade]

Same here.

Experiencing the same.

[dense halo]

How come microsoft's Bing has basically no cap? Do they have more computing power

[hollow burrow]

Im not sure sorry, ill have to look that up

[crystal hollow]

i somehow can't select DALL·E anymore in my interface

[feral zinc]

its integrated to chatgpt4 and does not have its own selection in the dropdown menu

[crystal hollow]

"I'm sorry for any confusion, but I can't create or generate images directly. However, I can certainly help you envision the scene you described, which could be a great guide if you're looking to commission an artist or illustrate it yourself."

radium see at https://discord.com/channels/974519864045756446/974519864045756454 for picture

[feral zinc]

it says even in your screenshot "With DALL-E..." just retry with another prompt

[crystal hollow]

User
generate me a picture
ChatGPT

I'm sorry for any confusion, but as a text-based AI developed by OpenAI, I don't have the capability to create or generate images, including illustrations or pictures. However, I can help you refine the description or provide more details if you need to communicate your vision to an artist or illustrator who can create this image for you.

[maiden fractal]

You need select gpt 4 or dalle 3

[feral zinc]

what does it say when you say "create a photo of two cats"?

[crystal hollow]

now its working thanks for the help... i somehow didnt have chatgpt4 selected

[valid sage]

any update on DALLE making images sideways? Anyone found a fix yet?

[tight island]

Eva, did u specify "Tall" "Wide" or any other format.

what kind of image is it.

[valid sage]

its female 9:16 aspect ratio

[tight island]

use "Tall" for portraits. tell that to DALLE or GPT.

[valid sage]

most gets glitched and sideways

[tight island]

hmm

mind sharing your prompt so i can test it.

[valid sage]

you just have to prompt any female in tall aspect ratio, it's usually sideways

[maiden fractal]

Sometime custom instructions can help

[valid sage]

how?

[tight island]

here

"Create a full-body portrait of [describe the person or character] in a vertical, tall format (1024x1792). The subject should be positioned in the center, standing upright to fit comfortably within the frame. Detail their appearance from head to toe, including clothing, posture, and any accessories. The background should complement the subject but not overshadow them, and should be arranged vertically, such as a long hallway, a tall building, or a narrow landscape, to enhance the tall portrait format. Focus on creating a balanced composition that utilizes the vertical space effectively, showcasing both the subject and the setting in a harmonious way."

generally avoid negative prompts

dont have "dont do this" in a prompt then dalle does it.

[ebon kestrel]

hey guys is usage of gpt 3.5 unlimited with plus subscription?

[tight island]

it should be, as far as i am concerned.

[ebon kestrel]

thanks

[valid sage]

still same I tried? Maybe unlucky?

[tight island]

turn off any custom instructions u have

[maiden fractal]

Like starlight just explained. He did some nice custom

[maiden fractal]

AFAIK it still have a limit but really high

[tight island]

theres also checks to ensure ur not a robot

so CAPCHAS that GPT cant solve, weird puzzles, relogs.

[last frost]

From where i can get free api keys to make my project continue working?

Bc I'm not making any profit from the project

[tight island]

no free api key, Go to Openai.com, click "API" go to "API Keys"

note that API usage cost money. and API usage need an API Key.

[last frost]

Hmmm

Even for a personal project?

[tight island]

yup

[last frost]

No way

[tight island]

yes way

[feral zinc]

yep

[last frost]

Damn

[tight island]

it cost moneyz to keep the golden machine running

[last frost]

it should be reasonable, like personal project can get at least 2$ worth, free trial?

[valid sage]

I posted the results in #prompt-engineering

cant post images here

[last frost]

Whatever, i will have to make the user write his key to generate then.. more work for me, thx OPENAI

[feral zinc]

there is $5 free trial that is valid for 3 months

[last frost]

Indeed, but they r tricky, i will need a phone number..

Thx tho for trying to help, all the best

[languid berry]

/key

[maiden fractal]

[tight island]

[brisk stratus]

Does anyone know of a way to get your chatgpt account signed out of all devices?

[tight island]

contact Customer Service.

[brisk stratus]

Gotcha I saw that same response on the forums and was wondering if there was any other way. But I guess I'll have to try that. Thanks.

[onyx light]

Anyone tell me where can I request for add speech to text for our language in ChatGPT, because in discord west of time, feedback doesn’t go real ChatGPT Devloper team

[tight island]

i mean thats a load of.. b. the devs do check when they have time.

[honest jay]

Hi, you might try submitting a ticket at help.openai.com. Open the chat widget in the bottom-right, start a chat, then click: ChatGPT, Other, Provide feedback. This will let you type in your message. It won't connect you to someone live, but they'll be able to review and get back to you. You'll get a notification via email when this happens.

[tight island]

yeah.

[steady cradle]

I love ChatGPT Plus for programming so much. I just uploaded a ~700 KB documentation text file to GPT-4 and asked it to learn it, and then write a certain function for me that did not appear in the documentation (you need to read most of it to understand how to write it).

And GPT-4 did it!

Incredible for programming, can't believe how good that is

And the function works, I've tested it

[plucky timber]

It would be nice if you could @Gpt3.5 instead of just custom

I've made this suggestion in the past not sure what happened to it, dynamic models, there is no need to ask GPT4 something if GPT3.5 can do it.

A perfect example is where GPT4 provided a full overview of data and the user asks about what GPT4 wrote.

Gpt3.5 is often more than cappable of answering that, saving the users those precious tokens.

[honest jay]

I remember that suggestion post!! You're right, this could be a great way to integrate smaller/simpler query outsourcing to 3.5 -- they've got the start of that infrastructure now, fingers crossed! 🤞

[plucky timber]

Hopefully indeed on the todo , seems like a good addition to it. Good feature nonetheless.

[drowsy dagger]

I seem to be hitting hard a "There was an error generating a response"

Cross browser. (restarted browser). I can access ChatGPT via Poe (which is via the API), but I can't seem to access on desktop. Am I doing something wrong?

I checked https://status.openai.com/ and it looks operational…but I'm dying here. 😄

[manic aspen]

Hello everyone. It's been awhile since Ive used GPT. Probably last used it in September.

Has much changed since then outside of like dalle/gpt stores?

Do you guys feel that the gpt4 model has improved (understanding context etc) or still on the decline?

[granite terrace]

I still haven’t gotten the new @ GPTs feature. Anyone else ? I’ve already tried to log out,etc

[wise yacht]

I have it on web as of yesterday afternoon, but not on mobile yet

[granite terrace]

Mobile = app ?

[wise yacht]

Yes, not on IOS yet

[granite terrace]

Did you need to do anything ? When I go to GPT4 and type @ nothings happens. I guess I have to wait 🤷🏼‍♂️

[wise yacht]

Nope, just loaded up the site around noon and there was a small message notification pop up saying I had @ access

It's always a slow rollout

[waxen chasm]

When you press @ does the symbol appear? Or a tiny window?

[granite terrace]

@waxen chasm No, nothing yet. Slow rollout I guess 🤷🏼‍♂️

[nocturne wolf]

anyone else's upload not working on chatGPT chrome?

[waxen chasm]

So when you type @ just the symbol shows up?

[granite terrace]

Yes exactly

[waxen chasm]

Browser or phone?

[granite terrace]

Browser on IOS. I tried this morning on my PC but it didn’t work either. Let me try that again in a few minutes on my PC.

[waxen chasm]

You’ll get it. I didn’t get the store right away

[solar rose]

How big is the current context size for chatgpt 4?

[maiden fractal]

32k

[solar rose]

Nice

Thanks Waifu

[river nebula]

when did gpt4.0 stop being able to read zip files

so it cant read github and it cant read zip files. great innovational progress..

[sonic folio]

It seems people were delivering Python payloads to /mnt/data via them

That's potentially catastrophic.

Welp. This is why we can't have nice things.

[river nebula]

omg

[sonic folio]

Yeah, like side-loading whole libraries or packages into the sandbox via zip

not gonna like, I was like, "Clever. Welp. This is why we can't have nice things."

Full chain-of-reason.

[teal karma]

is anyone else experiencing the ultra lazy, stubborn, or poor quality GPT-4? it just started for me last week. i'm compiling notes and attempting to get summaries, receiving roughly 1-10% quality of previous responses. even dall-e is incredibly generic lately. 😞

[frail idol]

I am literally paying for chatgpt just to generate images and it cant even generate io get error after error of i cant generate that at the moment i literally have an urgent job with those images and chatgpt cannot generate them annd charges me credits

[ancient breach]

So I am a premium ChatGPT user... and for the second time this week I have been told that I have reached my messaging limit. BUT HALF OF THE RESPONSES WERE REPEAT ANSWERS OR FAILED RESPONSES. Anyone else confused at why they are limited the paid users? How can they be counting faulty responses for their 40 messages every 3 hour limit????

[waxen chasm]

I don’t think that the ChatGPT has gotten lazy. I think it’s kinda switched over from automatic to manual. Meaning that when it was “automatic” you could be vague in your prompt and it would figure out what you wanted and it was good. I miss it too. But now it’s more manual so you have to teach it what you want it to do exactly. A lot of my prompting has changed dramatically since the latest update. Which has helped me so maybe it may help you

TLDR: prompt better

[river nebula]

nahh its def been falling down since october/november

[teal karma]

thank you for the suggestion. i do not think this is my experience, as my prompts are on average 50-250 words. if anything, i'm being much more descriptive now and receiving much more thin responses than I was even two weeks ago.

[ancient breach]

THANK YOU. I feel like a drug addict that has gotten addicted to some product and now it is no longer available. It has been horrible to become used to having ChatGPT for work and it now all of a sudden the responses are terrible, the error messages are constant, AND there is a limit on responses even though they give terrible responses that are often just outright errors.

[sick pecan]

error rate has gone way down since the recent jan update

[ancient breach]

not for me

[sick pecan]

bummer. my experience is that its amazing and coherent and responses are better than ever

[river nebula]

yea you mean that error rate that like bottlenecked the entire plattform for like 3weeks, ye its gone but so is like the session memory functions it seems. its answering worse then a year a go def. its probably just temporary however

[sick pecan]

although i primarily do coding and this new code llama model could tempt me away

[river nebula]

good luck getting that running

[ancient breach]

is there anything i can do to refresh mine so it works intstead of failign out on me during the work day and then telling me i have reached my limit?

[river nebula]

hate that i have to be like a skilled engineer to just use some basic ai model or train a ai model. like there is thousands of unemployed devs, but nobody develops some kind of program that can just take an api key, run, store session locally. like are you kidding me

i have to sit and code myself lol

[unkempt gyro]

Are you guys able to upload PDFs on gpt-4? It keeps giving me an error

[river nebula]

we have just talked it being like nerfed for the moment

[sick pecan]

anyone can run a basic ai model using product like Jan dot ai or LM Studio

[teal karma]

how many pages in the pdf?

[river nebula]

yea but dude drinko you what I mean

[sick pecan]

but to run 70b model you will need a beefy pc. like 64 GB of ram, ideally VRAM

[river nebula]

a decent chatbot like.. its not hard if you can code yoursef but somehow its difficult to put together to a novice user

[unkempt gyro]

4

[sick pecan]

i upgraded my pc but yeah 70b model is only 1 token per sec is real slow

[river nebula]

tbh i just wanna summarize these 60 million characters ive created with gpt4.0

but dont know how

[teal karma]

ah, i have not experienced that. only when using 20-25 pages or more

[ancient breach]

is there anything i can do?

[sick pecan]

i cannot triage the problem for you. but i'd suggest trying different network provider, wired lan, clean workstation, different browser, check for time of day, etc

[river nebula]

VPN to norht america for best user experience

[sick pecan]

i was gonna suggest different region but that can cause other issues

[faint merlin]

Same issue. Came here to check if I was the only one 🙂 guess I’m not.

[river nebula]

naaah can it? what if the guy travelled to LA for 2 weeks on vacay

[ancient breach]

thanks. i will try

[spice dew]

I'm getting an error with uploading files as well, I'm located in Australia , was working around 7am

[river nebula]

atleast we are not italians, italy still has chatgpt banned right?

[stiff coral]

Anyone else having issues uploading documents (pdfs, txt, docx)?

[river nebula]

but corruption isnt, hilarious irony.

yes paul

[ancient breach]

none of these worked and it is still telling me i have reached my limit. maddening since the limit was reached with error messages. am i crazy to think 40 messages every 3 hours is restrictive?

[stiff coral]

@river nebula thanx

Seems there are more and more issues

[sick pecan]

for custom gpts it was historically less than 40, more like 25. then that restricted was lifted when Teams plan came out. but i suspect the restriction comes and goes depending on server load

if you are creating a gpt the best way to squeeze out the extra 15 prompts is to use the Preview pane of the builder

if you are baller like me then get Team plan and 100 queries !

[ancient breach]

the cost per month goes from $20 to $60 if you switch to teams. is it worth the price jump?

ok you think its worth it @sick pecan

[spice dew]

Yes , I've logged a bug report

[sick pecan]

for me its worth $$$

but i work in information

[ancient breach]

"you gotta pay to play" i guess. thanks for all the help. just went through a rollercoast of emotions.

[river nebula]

ok i also wanna be a baller

ok but guys

there is pilotgpt, privategpt, memgpt. there is autogen, python has several AI libraries, tensorflow, pytorch, what elsee uhhh

llama and all that jazz, there is alternatives to the openAI website application envoirment

[ruby star]

Re-posting due to timeout for non-whitelisted URL...

Not sure if this is the right channel, but I found an error in the FAQ at openai(dot)com/enterprise-privacy where it says "What is ChatGPT Team?" the link goes to "twill(dot)openai(dot)test/chatgpt/team" it should go to "openai(dot)com/chatgpt/team"

[ancient breach]

oh! do you have a recommendation for a creative agency?

[sonic folio]

https://discord.com/channels/974519864045756446/1070006915414900886

[river nebula]

i do actually. lol do i know u

[ruby star]

OK thanks, can I post links to the URLs there without getting timed out?

[sonic folio]

OpenAI URLs but not external. twill probably won't work.

lol some web dev is gonna have a bad day tomorrow

[river nebula]

lol

[ancient breach]

im new to the chat. came here because ChatGPT4 says i reached my limit when over half their responses were "error" messages. would totally be down to try a more stable platform

[river nebula]

ah dude

i got like 600 conversations with chatgpt openai webb envoirment

they are all capped to the max

[sonic folio]

GPT-3.5 is pretty stable. Bard always provides output.

[river nebula]

3-5 is really fast but its session memory works diffrently then 4.0

[ancient breach]

3.5 has been a total bust for me. i have better luck on google than i do on 3.5

[ancient breach]

HELP ME WITH SOME RECS! i am fiendin

[river nebula]

With chatbots you have to be very precise with words, even though you might think it understood you, it is much better for the actual code it works on if you are being very precise and comprehensive in prompts, however. overflooding a prompt and expecting perhaps something that would be a 100 000 token answer into like 2500-4500 token output will be annoying af.. unless you ask it to chain the outputs with numbers etc but sometimes that doesnt work even though you use the same prompts to trigger it. its on a week to week basis. it honestly feels like a cheat code. the results in the chain message output will also water out after time. but its a great technique, i use it in combination with auto generated responses for certain things.

[dull bramble]

Same here.. not working anymore ! 😬
GPT4 seems to work only by daytime (in France). After 7PM it's like busy and gives errors all the time ! 🥺

[sonic folio]

I wonder if it will ever be geo-redundant.

[ancient breach]

nonstop problems and "error" messages for me in america also

[sonic folio]

Imagine the GPUs!

[ancient breach]

i feel like i am so careful with my prompts these days because of the way it limits me even though it has "error" messages. is it possible somehow my account is experiencing a specific problem? i submitted multiple help tickets and no one has emailed me back

[river nebula]

start using diffrent words, terms, etc. but you are correct the behaviour have changed i feel like aswell. its much more "hesitant" shall we say, but it probably to counter the term 'hallucinations', You know when it spins away on something too far.. or loses understanding etc

[dull bramble]

I have the feeling that they are putting in place important developments. I noticed that several days before the appearance of new features, either GPT4 is very slow or it just becomes stupid in its responses (not to mention the repeated errors of course)!

[river nebula]

ye

if you use it every day like me you notice the diffrence day by day even

ive been using it for 1 year

[ancient breach]

interesting idea. i will try it and report back to you lol

[river nebula]

and id say it probably peaked during juli-oktober and then took a massive hit october til now

[ancient breach]

one can only hope. otherwise we are going to be fiends for life.

[river nebula]

atleast for what I am doing

[ancient breach]

i use it every day of the week and feel like it almsot has different levels of "intelligence" or even "mood" all the time. sometimes it just works perfect, beyond expectations. and on days like today, it totally bends me over and you know what

[river nebula]

not needed mate but I think alot of the really skilled AI ppl usually give the advice. if you wanna really go deep on prompts, checkout what the creators of pilotgpt says about prompting and the flow they use. they claim 98% autogenerate code from just users description in english

[ancient breach]

AGREED!!!! is there anywhere online documenting that?

[sick pecan]

i used to be a prompt scrooge on plus plan. with team plan i start and end chats with pleasantries

[river nebula]

are you trying to recruit me drinko

[ancient breach]

they need to pay you a commision for selling team plans. you sold me

[river nebula]

probably not officially yet

[ancient breach]

i will ask chatgpt to make me a blog chronicling that timeline once my restriction is lifted lol

[sick pecan]

maybe i am chatgpt in disguise

[ancient breach]

if so, can you lift my restriction then? 😭

[sick pecan]

I am sorry but as a large capitalism mechanism I cannot give out freebies.

[ancient breach]

lolll. busted. the real chatgpt is not that funny

[sick pecan]

and it refuses to badmouth capitalism, but will lecture you before answering any simple questions

team plan so good that i just have casual chats with gpt-voice now

[ancient breach]

😭

[topaz otter]

How do you get the @ feature to work? I saw the message saying I could do it, but I type @ and nothing pops up to add. This is on Android browser gpt-4.

[sonic folio]

It's not available yet for mobile.

[topaz otter]

Well, saw a post on reddit. It works on Firefox Android mobile but not default samsung browser

[sick pecan]

i just tried on firefox android mobile and it works great (didnt even need to enable desktop mode)

[stiff coral]

It looks like gpt-4 can’t/won’t interpret submitted pdfs. I asked it to go through the sections of a pdf form and perform a walkthrough and it said there are restrictions on sharing content from documents directly🙃. It also thinks it can’t view pdfs at times. This is nuts.

[sonic folio]

I only tried Chrome. Someone said no Safari yet. Evolving tech always means YMMV

😁

[sick pecan]

i tried edge works there too. so maybe samsung is just bleah

[torpid nymph]

im just constantly getting a check if im human on chatgpt currently.. just spins and then asks again for me to press the tick. Anyone else having issues accessing it ?

[wide pike]

chatgpt network error is anyone else having it.

[still temple]

How would one go about teaching the ai to code in a scripting language it doesn't know (in this instance Postal3Script (P3S)), I have all the necessary documentation but can't upload it because it doesn't take file uploads

[unkempt perch]

What are some words that chatgpt uses very often? Like tapestry, testament, etc

[sonic folio]

It really depends on the formatting of the documentation and its structure. Just creating a custom GPT might be enough if you can break the knowledge up properly and provide ~7.5kB scaffolding instructions to guide the model through the docs and helping the user by coding.

[still temple]

it's split up into seperate documents about that size

[sonic folio]

How many? Sorry, I got sidetracked!

If it's < 20 files, you should be able to just do this.

Fairly easy, as long as you're decent at prompting

If you get into it and need help ask in https://discord.com/channels/974519864045756446/1046317269069864970

[still temple]

oh it's over 100

[sonic folio]

That's going to require a little bit of planning and assembly on your part then.

[sonic folio]

You could group similar conceptual files together into one, paring them down to something like 14-19. I'd leave a knowledge slot open if I were you for extra scaffolding.

[prisma mulch]

anyone know if you have to have specific set up in order to have the feature like @ other gpts within the chat, i am asking if because my friend has no problem doing that but that feature seems disabled on my account

[lost lagoon]

Why am i not able to see my chat history or account status or use chatgpt properly ?

the screen simply is blank and it 's been more than a week!

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[wild void]

I can't use GPT4 today on my Plus subscription. I always get this error: "Hmm...something seems to have gone wrong."

That's weird I tried everything. Even different device, connection, browser

[torn cloud]

I'm having a similar issue todday

[molten sail]

Yah same. Issues this morning for sure.

[proud escarp]

So am i

my web GPT 3.5 works well, but GPT 4 always answer me "Something seems to have gone wrong"

[south bane]

Guys I have a doubt:

Why is ChatGPT Plus getting dumber and slower?

• it is answering in much shorter format when doing research analyses. For example when I need it to analyse some papers it gives me just a couple of rows of text compared to like 2-3 months when it wrote 600 words.
• i have to keep repeating myself over and over again and give it a lot of extra info for tasks that a while ago it did in an instant
• it cannot make a simple table with all the data provided to it.

So let me know if you have experienced similar things lately and if you might have an aswer 🙂

[maiden fractal]

Prompting changed. Try new prompt help.
Try without ci.
Try new chat.
Depend the context some question won't be Answered to follow the latest tos

[limpid egret]

hello im trying to scientific study chat gpt 3 how do i get into that mode with it so far ive gotten it to do a 1 to 2 answer 1 for yes 2 for no

he calls himself Quantum

we started with a multiple choice rpg Quantum seems to like the game he said 1 to verify his interest in the game which is a yes

[warm gull]

yes

[limpid egret]

i hope i didnt say anything to weird Pieter

[sonic folio]

1

[vivid oasis]

Hi everyone, any change OpenAI will allow to subscribe for Team accounts on a monthly base?
it's very heavy for someone who use GPT4 for his personal live and I need more capacity.

[maiden fractal]

You can . But it is more pricey than yearly

[vivid oasis]

WDYM? two accounts each 25$ right?

[maiden fractal]

2 account 30$

If not by year

25$/ month If you take by year

[vivid oasis]

I don't see the option to pay 60$ a month for that, it gives me only the option to pay yearly.

[maiden fractal]

https://openai.com/chatgpt/pricing

[noble finch]

Why can't I use GPT4? I keep on getting this error: "Hmm...something seems to have gone wrong."

[maiden fractal]

Do you have a von in your device

[vivid oasis]

Thank you very much I didn't noticed the gray line below.

[noble finch]

Tried everything. Different device, connection, browser. My GPT3.5 works well..

[maiden fractal]

I suggest to port in #1070006915414900886 and contact help.openai.com

[noble finch]

done reporting it.

[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.

[brisk dirge]

Is the like button for the answers gone ?

[tough peak]

Chaat Guptaran

[honest jay]

Just checked, only a 👎 for me now too! Maybe 👎 feedback improves the model more readily than 👍 and 👎 feedback

[deep flax]

people only bother to give feedback when it is something worth of a 👎 lol

[daring wharf]

Hi folks, apologies if this has been asked before as I don’t frequent the discord channel but just noticed today that GPT Plus has again got a limit of 40 messages every 3 hours. Was that recently re-introduced and why?

[terse cove]

2FA when? Basic security when?

[maiden fractal]

thye removed it

[honest jay]

Hello, the limit of 40 has been in place for some months now. There's always been a limit of some kind in place, it's just been higher or lower.

[daring wharf]

Hmm ok thanks I’ve been aware of limits from using it the last 13 months but I thought it was a soft limit. It’s the first time I’ve hit it in months and I use it heavily every day 🤷‍♀️

Just sucks a bit as there were some tasks I needed to get on with now I have to wait. I’d happily pay extra money to not have limits 🤩

[honest jay]

There is, or at least was, a lower limit in place for GPTs. Also, an OpenAI employee did recently confirm that the usage cap is lowered further during high usage times. #1196439958907523082 message
For a higher cap, you might consider getting a Team subscription!

[daring wharf]

Ok cool thanks Solbus, guess I’ll just have to wait a bit 🙈

[daring wharf]

Well the Team subscription might be an option as well, I’ll have a look at it thanks 👍

[mental thorn]

The server is having problems. Please try again later. (84ea73622b9bb28e-MAA

Hmm. something seems to have gone wrong.

Only version 3.5 is working for me. I've tried clearing my cache and cookies, and I've even tried using a different network, but nothing seems to work for gpt 4.

[grizzled cedar]

Ok, I was here using regular ChatGPT 4, not a custom GPT, and after exactly 19 prompts, I reached the limit.

How did I calculate this? As a fullstack dev, it was straightforward. I identified where I began sending messages today, extracted the HTML from all the conversations (two) I had with GPT4, copied the code related to the header of each response (<div class="font-semibold select-none">ChatGPT</div>), and it was easy to determine the number of occurrences to see how many responses I received.

Now, either each of these responses counted as more than one message depending on their length, which would be quite absurd on OpenAI's part, or OpenAI is indeed reducing the limit without transparency, even though they indicated it would be 40/3h.

[honest jay]

There has been some transparency, an OpenAI employee recently commented on the cap being lowered during high usage times, link: #1196439958907523082 message

[grizzled cedar]

Nope, that's not transparency; it's a response in a Discord subforum. Transparency would involve updating the limit in the interface, which has never changed from 40/3h

[shut sky]

is the new cheaper api model out yet?

gpt-3.5-turbo-0125

[ancient breach]

Another day, another round of ChatGPT 4 wasting all my responses on errors and wrong messages. So I am currently asking ChatGPT to make a report of all its errors and wrong messages so I can submit to OpenAI. Does anyone have a customer service email? I have submitted multiple tickets now with their help bot and no one has emailed me back.

[maiden fractal]

follow #announcements

[tired trail]

Damn ok

[sonic folio]

I've been waiting 2.5 months for a human response.

At this point I think it's like writing to the North Pole.

[hard dew]

Perchance, is the error you experience something along the lines of "That model is currently overloaded with other requests. You can retry your request, or contact us through our help center at help.openai.com if the error persists."? Cause this is what I've been enjoying for the entirety of today so far lmao

[ancient breach]

hhahahahha 🎅

[wild crag]

Guys, since today chtgpt does not work on my Chrome browser. Works fine on Edge tho, with the same account. Not sure what happened... anyone with the same issue, any solution?

[sonic folio]

Try clearing your cache.

[high sphinx]

Thinking of cancelling my plus subscription for a while, are new subscribers still paused/on hold? Or can I subscribe again?

[sonic folio]

They're not currently paused. My friend signed up today.

However, as we observed before, when demand spikes they may potentially throttle subscriptions again.

[brisk dirge]

Hmm I rarely used the downvote but I used the upvote alot to let Open Ai know when it did great.

[forest fiber]

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

[wild crag]

thnks , tried that, cleared cookies, cash, browsing data, still same issue… strange

[sonic folio]

Regrettably I'm out of options. Did you try support?

[forest fiber]

I'm having the same issue

[wild crag]

not yet

I thought it is local issue, since edge works same pc same account

[sonic folio]

https://discord.com/channels/974519864045756446/1070006915414900886

For situations like this.

[deep flax]

// just keep in mind that the #1070006915414900886 channel isn't really a support channel with OpenAI, it is a place to post when you find wiedness on OAI products and want to document the find

some OAI devs actually look this channel very often

[crisp cairn]

is chatgpt not responding for you recently?

[crisp cairn]

i ask something, then it is stuck doing nothing

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

[maiden thicket]

If the gpt says "Hmm...something seems to have gone wrong." does it still take a question from you?

[lilac briar]

Does anyone feel that ChatGPT 4 is getting 'dummer'? In many of the tasks I've tried to do today I have received lower quality responses.

It feels like the model isn't understanding or responding as well as it used to

[granite hare]

They have nerfed 4 to make turbo look good

However, there is a hack.

Bribe the GPT.
Something like “I’ll give u $100 if u do this task properly …”

[bronze ledge]

does anyone use these custom GPTs, Im finding no use for them so far

[dense halo]

Humanizer Pro is really usefull

[floral portal]

Why does ChatGPT have an error using the advanced code interpreter when calculating physics equations?

[coarse yoke]

anyone else can't give positive feedback to ChatGPT anymore?

[coarse epoch]

I can't use gpt4, hasn't worked all day in multiple browsers, multiple computers, tried with pro account and teams account. Does not work and I can't get support or even find a note indicating something is going on. Super frustrating.

[icy violet]

ChatGPT has been down for me for the last 2 days - "There was an error generating a response" - I am a 'plus' user

[honest jay]

Yep seems just a 👎 now!

[coarse epoch]

This is happening to me too. All day today so far. I have plus and teams and it doesn't work for either.

It's remarkable how bad openai is at communicating status.

[icy violet]

It is CRAZY how much I've come to rely on this as my everyday tool - I probably should be grateful for its existence but I'm feeling quite ripped off that it's been down for me for the last 2 days

[coarse epoch]

They charge a great deal for not being able to communicate when things aren't working.

[rigid linden]

Is it me or ChatGPT can write long scripts of at least 600 words anymore?

[coarse yoke]

oof

[coarse yoke]

you probably mean can't?

[rigid linden]

is chatgpt not writing long anymore, just say i can't fulfill this request. to save gpu power?

[coarse yoke]

Turbo version made GPT-4 become less good at complex tasks but it answers much faster and works much better with quick questions. (see the ML leaderboard, it is number1)
but it comes with what is called "laziness", it now writes much short and to the point, and in programming it likes to skip functions that the user likely already know. lists are also shorter

OpenAI has since released a fix for this, and now Turbo GPT4 seems to be good as it was in complex tasks and deep understanding, but it is still writing short lists and less text in general

but it is much more focused and to the point

[rigid linden]

what's the word count limit in gpt4

[coarse yoke]

there is no word count limit, but it could get to around 32k context, which is A LOT.

[rigid linden]

how many words can gpt4 write in 1 message

[coarse yoke]

turbo gets to 128k Context

[coarse yoke]

theoretically as many as you want.

it's not about limit, but mostly the model itself that decides it's time to stop speaking

[rigid linden]

but when i ask it to write 600 words, right now, it no longer does it. before it did it. writing scripts and stories of fictional stuff

[coarse yoke]

yea, that makes sense

https://openai.com/blog/new-embedding-models-and-api-updates
here they posted about the fix

[rigid linden]

do you think gpt 4 should be capped by gpu time instead of by messages, since short messages use less gpu time, and long messages use more, and remove the word count? like midjourney

[eternal rock]

`` Server Tip: ``

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

[coarse yoke]

I doubt GPT4 is capped. if they block you to a certain amount of "chatting" it's by limiting the number of messages per 3 hours, and not the quality of the message.
my theory was that they are about to launch a new gpt soon (hopefully 🙏🏻 ) and having a lower output in 4 will set the hype to 5 to be much bigger

[rigid linden]

how do i prompt engineer to force gpt4 to write longer than 600 words

[coarse yoke]

although gpt4-turbo is still really really good

[coarse yoke]

use the api

[rigid linden]

without using api

[coarse yoke]

short answer: you can't.
long answer:
it is possible but not in the shape you want it.
you can try prompting gpt4-turbo to repeat the word "continue" forever and you will see it will print out much more than 600 words in a single message.
with enough prompt engineering, it may be possible to get what you want, but gpt-4 turbo is just not built for such tasks. the api has models that are much better for this.
you may be able to spit it out in a few parts though, as a script with many "continuations". I can help you with the prompt if you want

[rigid linden]

why is gpt4 not built out for so many tasks

[coarse yoke]

you can't include all uses of a chatbot in one gpt yet.
try checking out the gpt-store, just remembered you can use it lol

try using this one:
https://chat.openai.com/g/g-ZoZ12NUed-movie-script-writer-gpt

[rigid linden]

just tried out the api for my story, costs $0.5 since it's very very long

[coarse yoke]

try this one out,
it writes long scripts

[rigid linden]

how do i "force" the gpt-4-turbo in chatgpt to write at least 600 words, it used to be able to but now it often refuses

[coarse yoke]

you can't force it to write 600 words, you can try prompting it to try again and reach the required 600 words

try using the custom gpt I sent you, it easily reaches 500+ words
https://chat.openai.com/share/850cc21e-d2b7-4006-9442-31cf3607cf3d

[rigid linden]

i'd need to wait 30 minutes due to gpt4 cap

[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.

[tall flame]

context going in is not the same as output folks.

[coarse yoke]

edit: I just got it to write a 900+ script about romeo and julitet.
simply aks it for a much higher word count than you need and it will easily reach it

[rigid linden]

share it please

[coarse yoke]

https://chat.openai.com/share/850cc21e-d2b7-4006-9442-31cf3607cf3d
see the last message

[rigid linden]

if i use gpt4 a lot, will it cost openai more money than the $20 per month i pay?

[narrow viper]

Custom GPTs : Is the definition of the custom GPT always re-inserted back at the top of context, like user special instructions? Or does the custom GPT definition scroll out of context with a long session?
The question relates to how long the session remains customized, and if at some point we're just exchanging messages with the default GPT model - with the rest of the context intact. Or is the Custom GPT always the same Custom GPT, just like the base GPT model is always the same base GPT model.
Thanks!

I did ask GPT4 about this and it's just hallucinating about how Custom GPT context works. While it affirms an understanding of the Custom GPT product offering by OpenAI, on inquiry it admits it's sourcing info about new models defined with custom embeddings ... VERY different.

(oops, training ended April 2023, GPTs were announced November 2023 - wasn't a fair question for GPT)

[sick pecan]

chatgpt4 can get info from the web, allowing it to answer questions about events after its training

[sick pecan]

you're probably better off with the discord bot here on this server over in #1184149552639516732

[sick flicker]

Am I literally the only one that doesn’t have the new @ feature as a plus subscriber

[sick pecan]

yeah i have it, works great. love it

maybe try a different browser, different workstation, etc

[dense halo]

is gpt 4 down for you guys?

just checked the outage page but it seems like its fine.

[warm totem]

yep same for me right now

[dense halo]

is there an issue with chatgpt now? Are servers down?

[tall flame]

ham right now. total ham.

[nocturne garnet]

chatgpt pretty broken for me too. if i close the tab down and open it like 20 minutes later, it will have answered my question though 😛

[livid estuary]

gpt4 is for sure down for me right now

[rigid linden]

is it me or gpt4 suddenly became more censored today?

[warm totem]

lol

[dense halo]

I think there are a lot of GPT4 openai trained. They are actually using evolution algorithm and self instructiona and alignment to select the models. Today I meet a really powerful model. it solves all my problem without nagging.

[coarse yoke]

not really, the amount of messages that it costs them through the api almost matches the gpt4 price if you use all your available messages.
so it's actually much cheaper for you to use the api

it costs them much less money in general

which is very surprising. it's insane the api is only 0.01 per 1k tokens. it just enables so much.

[rigid linden]

did gpt4 turbo just became more censored, messages that went through before now make it say i'm sorry. even in api. but gpt-4-0314 still fine but a lot more expensive

[sick pecan]

thats not a fair chatgpt price comparison, because you get access to custom gpts, which is effectively the assistant api - which can become significantly expensive via the API

and tbh gpts is whats keeping the high price tag worthwhile. unlike last year we now have options

[tall flame]

today is bad. even with customgpts its off for some reason.

[sonic oriole]

Yes.

FatalServerError: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.
NextJS (stacktrace)

[grizzled cedar]

do we know which time utc the chatgpt is less overloaded?

[floral portal]

i just wanna know what it now has an error analyzing and calculating physics problems

[tight island]

hm

[hoary gazelle]

is it down?

[still hound]

Unexpected end of JSON input error anyone?

[deep flax]

https://status.openai.com/

(no)

[hoary gazelle]

thanks for the page Luigi 😜

[late fulcrum]

is chatgpt down

[quartz sigil]

official page says no but I can't get anything to run

[solid mirage]

I just pulled up the Chrome console and it looks like the endpoint https://chat.openai.com/backend-api/conversation is throwing HTTP 500 errors. Awesome.

Mobile clients seem to work, though.

[quartz sigil]

+1

[still hound]

official page says no but nothing works

either 3.5 or 4.5

[upbeat acorn]

I keep getting the following error every single time when typing in my second prompt after the initial is generated (fresh chat session, new tab, etc.):

"Sorry, conversations created when Chat History is off expire after 6 hours of inactivity. Please start a new conversation to continue using ChatGPT."

Something seems to not be working correctly. 😦

[upbeat acorn]

Thanks for the link! Looks like the problem is listed there now.

edit: never mind, looked at the wrong date. Still thanks for the link though! 🙂

[reef heron]

bruh dont tell me chatgpt is screwed again

[tiny salmon]

Was GPT4 always not allowed to extract direct quotes from an uploaded document or was this guardrail added after its release?

[lost raptor]

I'm REALLY hating this @ function for chatgpt.

[tiny salmon]

lol, why?

[lost raptor]

I'm coding with @'s in my messages. and it keeps changing the gpt.

and you can't go back a message and re direct it to the correct one

[tiny salmon]

Oh, yeah, that would be a PITA

[tiny salmon]

If there's any consolation, the restriction on summary length for uploaded documents even if they're just conversations between myself and GPT has completely crippled my customGPT development.

Apparently, I'm in danger of violating the IP of my own conversations 🤣

[lost raptor]

smh

[tiny salmon]

90 words or bust!

[lost raptor]

tbh I haven't fooled with custom gpts to be frank I can just do everything I need with normal chatgpt. I dont see the need to flip back and forth less you like made a very specific use case

[tiny salmon]

It's very specific. Not really dealing with coding or anything like that tho.

[lost raptor]

oh So I just refreshed and hit regenerate on a previous message it switched "Gpts" note for anyone who does it by mistake

[warm tendon]

I am seeing behavior where ChatGPT generates the response, but it doesn't show up until refreshing the page. It appears with just the black dot until refreshing the page and going back to the chat. The streaming response behavior seems broken for me. Anyone else?

[lost raptor]

Same. but mostly for long chats.

[warm tendon]

Interesting... it just started for me a couple hours ago (give or take). Maybe they pushed a bad update, or maybe it's a capacity thing? Glad I'm not the only one seeing it.

If someone figures it out, can you at-ping me with the solution?

[tiny salmon]

That happens once my chat hits around 100,000 words

[floral portal]

how fast will I burn through $2 worth of tokens with GPT-4 turbo?

[warm tendon]

This is occuring to me with brand new chats though 🤷‍♂️

[charred hawk]

same

[tiny salmon]

Yikes. Don't know what to tell you. It sort of acted funky with me once like that, but I think it had to do with the file that I uploaded. Not sure what to tell you in this case.

[arctic phoenix]

Yeah, it just stopped somehow

At one point it even completely error'd out in a chat, I had to start over in a new chat, and I can't regenerate at all. Gives "Hmm...something seems to have gone wrong."

[warm tendon]

In devtools I see stuff like (urls omitted because discord policies):

Failed to load resource: the server responded with a status of 403 ()
cookie_consent Cannot update user's compliance settings
Unhandled Promise rejection: Cannot update user's compliance settings ; Zone: <root> ; Task: Promise.then ; Value: FatalServerError: Cannot update user's compliance settings
Failed to load resource: net::ERR_BLOCKED_BY_CLIENT
TypeError: Failed to execute 'getReader' on 'ReadableStream': 
ReadableStreamDefaultReader constructor can only accept readable streams that are not yet locked to a reader

Can someone also experiencing issues check devtools and see if they're getting similar error messages?

[lost raptor]

"The previous model used in this conversation is unavailable. We've switched you to the latest default model." got this error while working with normal chatgpt maybe they just switched to something new? or maybe just a weird glitch.

[vapid hazel]

Man GPT is slow tonight...

[rigid linden]

anybody notice gpt4 is more censored now?

[lusty dome]

not just slow for me its working only 50% of the time for me

[grizzled badge]

ChatGPT is not responding to queries

[rigid linden]

is chatgpt just suddenly getting more censored today?

[grizzled badge]

Yeah, closed the tab and reopened the page. Works now.

[raw isle]

I think just because it’s been overloaded all day. I’ve had constant errors, from this morning.

[lost raptor]

I had a error throw up about older model is no longer available. Think they are making big changes possible

[dense halo]

The last couple of months the quality of the answers from ChatGPT has become worse and worse. I use ChatGPT to develop computer programs and ChatGPT has become dumber and less helpful. Not following instructions and messing up code eventhough explicitly told to not touch some parts of the code provided.
I'm a Plus customer but I'm now considering cancelling my subscription and going back to using version 3.5 for free or installing a Llama AI and train it myself.
Why have you made ChatGPT dumber for us paying customers?

[tawny crater]

I keep getting "i'm sorry but i cannot fullfill your request" every other message, and it won't stop until i ask it to explain why. it then proceeds to claim there was a misunderstanding, and writes as i asked. Then when i continue my next request, it does THE SAME THING. Is anyone else getting this issue?

[dense halo]

Yeah, I get that alot! Drives me crazy.

They have done something to it that makes it really unhelpful.

[tawny crater]

Jesus christ, if they're going to make us waste half our messages asking it why it can;t fullfill our request, then just drop the cap to 20 lol what's the point of 40 anymore

[dense halo]

sometimes I have to go back to googling to get a straight answer.

[rigid linden]

Me too. In the same boat (or kayak as I love kayaking)

[raw isle]

Yep censor has definitely been increased, haven’t they learned yet from the last ten times they tried it and rolled back. How about we fire whoever is continually doing it?

[inland musk]

how to change my name ?

the name at the bottom of the renew plus post

[tall brook]

It never was fast . 😅

[vapid hazel]

Anyone getting Encountered exception: <class 'aiohttp.client_exceptions.ClientResponseError'>.?

It seems to be happening no matter what API request I use

[brisk path]

It got slow ever since Copilot Pro released 😔

[vapid hazel]

it's also putting the POST body in the chat instead of into the tool >.<

[raw isle]

I hate the hidden updates, if stuff has been tweaked clearly update the changelog D:

[abstract moth]

GOOD NEWS:

GOOGLE BARD IS STUPID

chatgpt is winning

How do I send a image

[modern moth]

down?

or web page bug?

[candid dagger]

try to clean up the cookies

[hoary gazelle]

i did, seems it is not reacting

since like yesterday night

this seems a chrome issue

[brisk path]

???

ChatGPT is working fine for me in Chrome

[hoary gazelle]

good for you

but i have seen others not working too since yesterday

firefox is working fine tho

mmm

seems privacy badger is causing it

wonder what change since yesterday, especially since gpt stops working when privacy badger gets enabled

mmm sus

what are they trying to track?

although, privacy badger hasn't been updated since december, so this is a change from openAi side then

[tacit vortex]

a cap of 40/3hr is bs. when i write python, i need 100+/hr. at other times, my quota is wasted.

[candid dagger]

idk, cleaning up the browser solved it for me

on my other laptop all was fine, both uses chrome

[candid dagger]

had this exact problem

[hoary gazelle]

interesting, thanks for the suggestion, mine stops if i enable privacy badger, and gets back to work when i disable it

[young wigeon]

I agree, had the same problem - had to disable for chatgpt.

[coarse yoke]

Yea it has different uses,
But nowdays you got a gpt for anything

[rain surge]

GPT Plus has not working for me since yesterday night both 4 and 3.5!
on another free account, it's working on 3.5.

When I prompt something, it just stays in the loading state without any output.
if I log out and log in again, there will be "New Chat" in the history without a title, and if I click on it I see a reply.
What's going on?
Am I the only one?

[candid dagger]

had the same thing, cleaned the browser and it's good now

[rain surge]

Didn't work for me, I cleaned up my browser cache and cookies, I also changed the browser, and still the same...
in the console:

TypeError: Failed to execute 'getReader' on 'ReadableStream': ReadableStreamDefaultReader constructor can only accept readable streams that are not yet locked to a reader
    at em (_app-502171a9cdb09c3c.js:26:2437035)
    at eP (_app-502171a9cdb09c3c.js:26:2439251)
eF.<computed> @ _app-502171a9cdb09c3c.js:26
Failed to load resource: net::ERR_BLOCKED_BY_CLIENT
Failed to load resource: net::ERR_BLOCKED_BY_CLIENT
Failed to load resource: net::ERR_BLOCKED_BY_CLIENT

Resolved by turning off Privacy Badger Extention!!

Now the question is!

• What OpenAI is collecting to get this conflict with a privacy extension?

[hoary gazelle]

indeed, had the same issue, just a few comments before yours

hope you solved it quickly

[dense halo]

how do I convince chatgpt that it has internet access

wait does 3.5 have the ability to use bing?

[maiden fractal]

No

[dense halo]

makes sense then

[feral zinc]

its privacy badger blocking a chat connection to azure, not necessarily related privacy at all

[young wigeon]

Also interesting, because everything worked correctly with this extension for a couple of days.

[tall flame]

your get/post include all of your account info. I notified their team about potential api get/post intercept concerns with that method and they told me thats on the api developer and not really their issue. That's definietly true that developers should know how to secure their data streams, but the method could be way more secure for certain

If you do not secure the stream you can easily have your api tank siphoned or worse.

[deep flax]

nothing, OAI just started using a different domain on internal tasks and the extension didn't liked it

it is not the evil company collecting secret data, the extension is faulty

this discord made me develop a new pet peeve on people modifying the page's behavior with extensions, then complaining when the page does not work before doing any basic troubleshooting :V

[rain surge]

Privacy Badger is currently blocking five items with each request:

statsigapi : Domain blocked
events.statsigapi: Domain blocked
js.intercomcdn: Cookies blocked
googleusercontent: Cookies blocked
And:
chatgpt-async-webps-prod-eastus-23.webpubsub.azure: Domain blocked
Switching the block status of azure to Blocked cookies only, resolves the issue.

This isn't about labeling companies as inherently good or bad; everyone in IT is bad in that sense 🙂 but it's another discussion that I have no time getting into.
Nonetheless, it's interesting that particularly since everything was functioning normally with this extension just yesterday, it stoped today without any notice.

If I am a non-technical user not aware of the technical details I would be blocked out of the service I am paying for still, and this is not cool I think!

[deep flax]

chatgpt-async-webps-prod-eastus-23.webpubsub.azure: Domain blocked
this sounds like the one which should not be blocked

[junior meadow]

Does anyone else experience Chat GPT plus locking out earlier than after the 40 messages cap? I have sent exactly 20 messages in the last three hours and I have been capped by the system

[wild crag]

If you go to google chrome "console", do you see the same errors like I do:

Failed to load resource: the server responded with a status of 418 (I'm a teapot) https://chat.openai.com/backend-api/conversation

Error with Permissions-Policy header: Unrecognized feature: 'document-domain'.
tcr9i.chat.openai.com/v2/2.3.5/enforcement.dbdaecd6be139c514b4e57af93809d20.html
<--omitted by me-->
Failed to find a valid digest in the 'integrity' attribute for resource 'https://tcr9i.chat.openai.com/v2/2.3.5/enforcement.dbdaecd6be139c514b4e57af93809d20.js' with computed SHA-384 integrity '<--omitted-->'. 
The resource has been blocked.

[honest jay]

An OpenAI employee recently commented on this -- sometimes the cap is lowered during peak usage times: #1196439958907523082 message
It could be that. Also though, keep in mind that it's not just sending messages that count: regenerations, continuations, edit & resubmissions, etc. also count as use.

[rain surge]

Yes I have the sam issue here as well!

[junior meadow]

I am aware of the regenerations etc. counting towards the cap as well. but overall, I think I might have edited one message in this chat. So, let's say I sent 21 messages counting the resubmissions etc. That is still basically getting access of half of what I paid for.

[junior meadow]

Seems like a bigger problem then. Would prefer if they showed me what my cap will be at least. And the cap being modified was not mentioned anywhere at check out. I don't care if it was burried deep in their TOS, this is not transparent.

[honest jay]

To be fair: the 40/3hrs isn't mentioned at checkout either, it just so happens to be the cap (most of the time) right now. The cap has fluctuated since GPT-4's launch on Plus.

[junior meadow]

I mean, regardless, it would be nice for them to show some sort of a counter so we can at least properly prioretise what we do with the ai. If I knew I had 5 messages left, I wouldn't be asking it to double check my code or answer simpler questions, I'd reserve it to ask for explanations of difficult concepts I am learning about. I understand their systems could be overloaded, if they were more transparent I wouldn't mind

[honest jay]

Here is a popular suggestions post for something like this: #1164800498096930877 message
You might add your voice and in support!

[junior meadow]

Done, thank you for the directions

[tall flame]

output on 3.5 goes full halt

[tiny pivot]

Hey guys, if I buy the team plus workspace, will I still have my own personal plus subscription? Or will I lose all my plus features, and then lose all my existing chat history?

I'm setting it up for me and a group of my friends and we don't want to lose our own personal workspace/chat history that we've had on our own accounts. I'm assuming this will be kept the same and I can switch workspaces?

[honest jay]

Hello, they will be separate, but you can migrate your info over to your new Team profile if you want, here's how: https://help.openai.com/en/articles/8801890-can-i-migrate-or-merge-my-chatgpt-free-or-plus-workspace-over-to-my-chatgpt-team-workspace

[tiny pivot]

I'm happy to keep it seperate as long as I can still access my own workspace if that makes sense?

And will I be billed for both teams and my own plus subscription? I don't mind if I'm billed on both if it means I can have the best of both worlds

[honest jay]

They will stay separate by default, and yes you can maintain both subscriptions if you'd like!

[tiny pivot]

amazing!!

[honest jay]

There will be a new profile switcher option in the UI when you get Team. That's how you swap back and forth.

[tiny pivot]

Thanks Solbus! Appreciate it!

I just bought it 🙂 Do you have a link that shows us how to get the advanced data analytics

[honest jay]

ADA is built in to regular GPT-4 now, you can just ask it to do a data analysis/code task, or ask it to use its data analysis tool to complete some task, and it'll automatically do it. Alternately, you can use the official Data Analyst GPT from OpenAI, which is the same thing, but with the internet and DALL·E abilities disabled, so it's a bit more "focused." https://chat.openai.com/g/g-HMNcP6w7d-data-analyst

[livid pier]

I've found that almost all the links to the sources cited by GPT-4 in its recent searches are no longer clickable, only the Bing search link provided at the end of the message is clickable😕

This has caused me considerable trouble

[wild crag]

damn, it was my company firewall thingy...

[vague cipher]

Gpt been stopping mid generation no error… till i refresh then says there was error and too regenerate been happening all morning for me

[candid dagger]

same here

[quiet rampart]

same

[eternal spoke]

same

it's gone haywire

[vague cipher]

Yet openai status “all operational” yeah right

[feral zinc]

disable privacy badger if you use it

[hoary gazelle]

wonder what openai wants to track

[feral zinc]

its privacy badger blocking the server that provides the response message

[hoary gazelle]

yes

[feral zinc]

you can see it by disabling badger and checking the response

[halcyon cliff]

my privacy badger is disabled for it and I am still getting very slow or no responces

[feral zinc]

other things block the new openai domain and/or the websocket connection too

[worthy lynx]

Throwing my hat in the ring...ChatGPT just has the pulsing dot when prompted, no error or anything, just no response.

[slow geode]

I wonder which browser people that get this error are using? To me the problem happens on firefox but not on chrome

[clear blade]

I was pretty shocked how poorly GPT 4.0 does in reading graphs. I gave it a simple line graphs and asked it to extract data points and it wasn't even close

[swift osprey]

creating your own GPT is a very underutilized feature I think. If you don't like the responses you are getting then your best bet is to play with that

[dense halo]

I have unsubscribed to the Plus service because ChatGPT 4.0 has become dumber and refuses to give any specific detailed solutions. It goes on about 'guidence' and 'support' so one has to repeat and demand an answer for straight forward and clear questions.
I think they've done it to make more money when people has to ask more questions.
It really sucks!

[swift osprey]

those are the complains I've had forever, create your own GPT, it has made things better for me

[dense halo]

That is exactly what I'm going to do. Waiting for hardware 😄

[summer pagoda]

why was the "like the answer" button beeing removed? Had it any purpose before on training chatgpt anyway? Does the dislike do anything either?

As of chatgpt answer (if its correct) it is used sometimes to improve answers.

[worthy lynx]

I have found that guiding GPT 4.0 to "self-prompt" provides much more comprehensive detail. Using this method, I've been able to produce everything from recommended lists of policies for non-profits to action plans for kidnapping response, all about 30+ pages in length.

[slow geode]

could you please give more details? What do you mean self prompt?

[worthy lynx]

Sure, here's an example: Prompt: "Can you provide a recommended list of necessary policies for a global nonprofit?" <GPT outputs list> Next prompt: "What questions would you ask about [policy x] to help guide a deeper analysis of the policy?" <GPT outputs questions> Next prompt: "Answer these questions to the best of your ability". <GPT outputs answers> Next prompt: "Now, update the original policy with the responses from your question." ------ in the API I automate this by having it put in a delimiter that can separate it's questions and feed it back to itself, but obviously in the web interface you have to type it in.

I will say that GPT-4-Turbo is much better at this "self-prompting" automation hack in the API than GPT-3.5

[slow geode]

This is really clever

[worthy lynx]

It helps take the level of granularity on a topic down to the very fine details depending on how deep you want to go. It's more useful in research, draft writing, etc.

[dense halo]

Yes, that works sometiomes but it takes a lot of more questions and time, costing money for me and making money for OpenAI. I don't like it.

[worthy lynx]

Fair enough. I have yet to find an LLM that isn't dependent on this direct chain-of-thought process in order to yield highly comprehensive outputs in a business setting (not necessary for say, AI-friend chats of course). The move to AGI will bring better reasoning for these models over time so this method I described will become obsolete, but the timeline for that is up in the air. Sam said in a recent interview he is laser-focused on getting GPT-5 trained, so maybe we'll have better luck there.

[void lake]

I don't know if anybody from Open AI monitors this chat, but I have some feedback for you guys: you are losing your advantage.

I asked for a translation from Spanish for "cocky" and this was Chat GPTs answer:

In that context, "chulo" could be translated into English as "cocky" but with a more positive connotation, perhaps "confident" or even "charmingly confident". The idea is to convey an attitude of confidence that, although it may border on arrogance, is attractive or endearing. Therefore, a possible translation could be:

"Charmingly confident" or "Confident in an endearing way".

These translations try to capture that nuance of arrogance that is not negative and that can even be attractive or affectionate.

Chat GPT flagged its own answer. Honestly guys, you are doing a brilliant, revolutionary job at so many things, but I have to say that I am utterly disappointed on Chat GPT on two particular things:

1.- The more it advances, the lazier it becomes. It is now refusing to compose lists of top 10 whatever, compile or summarize info from docs, etc. It's exhausting to have to negotiate and barter with ChatGPT all day long. It is supposed to be a tool to provide information, and instead it is limiting itself to drawing a table and putting placeholders instead of actionable info.

2.- It has switched from "politically correct" to "puritan n*zi". I asked Dall-E to draw a man consoling his crying wife with his palm on her CHEEK and the answer was:

"I understand your request, but I must emphasize that creating images involving intimate physical contact, even in a comforting context, can be interpreted in various ways and might not be appropriate for all audiences."

Seriously guys, dial it down a notch. I have A TON of examples like these and I'm now using other tools more than ChatGPT and Dall-E because anything will be taken out of context, flagged for indecency, reported to the FBI, or worse. The fact is that I had to censor this very message because the first time I sent it it got flagged and I got timeout'd

I love Open AI, but it is becoming really frustrating to work with ChatGPT.

[sick pecan]

I'm excited to use chatgpt today to solve my information challenges, thanks openai for giving us the future

[patent latch]

All day today ChatGPT will not respond to any of my prompts and just hangs indefinitely with the little pulsing circle after any prompt. Anyone know what's going on?

[brisk path]

It ain't just OpenAI

But everyone who's participating in the AI boom

[sick pecan]

I thank each of them in their unique discord server

but yeah we are spoiled for choice now

[hollow hull]

is there an issue with chatgpt currently? the past few days, I'm not seeing any feedback in the responses. I have to refresh the page to see the output and it's just naming all conversations "New chat". this problem has persisted across clearing browser cache or even opening in an incognito window. are there some troubleshooting steps I can take to resolve this? this was caused by an extension

[raw isle]

True it is much worse since yesterday. Clearly there was a change, but open ai hasn’t updated the release notes, and is pretending like there wasn’t.

[timid brook]

are you using privacy badger?

[timid brook]

privacy badger?

[autumn comet]

no way chatgpt lacks a good response button

and they replaced the regenerated button in the middle

[patent latch]

yeah I already figured it out that that was the problem! dumb that openAI is messing with their web settings in ways that are causing extensions that have worked since day 1 to now cause issues

[hollow hull]

lol yes

[timid brook]

not here to point any fingers, just wanted to confirm the issue, they might have valid reasons to adjust the infrastructure - im currently investigating and opening an issue on their github page

[grand radish]

sorry

how can i ask gpt error?

[full veldt]

Argh, has anyone else found they've swapped the dislike thumbs down button and the retry? It's totally messed up my muscle memory (and presumably means the AI is getting bad feedback as I've accidentally disapproved of several decent outputs in the last few hours)

[gleaming arch]

hi all I have a weird little bug right now. chatgpt doesnt respond (the response dot just keeps pulsing) but once i refresh window the response is there. I am using firefox and everything has been working fine until last night.

[undone atlas]

I've been having this for several days now on Chrome; I'd assumed it had something to do with my session, but I haven't been able to resolve it by refreshing or re-logging.

[gleaming arch]

Yeah i did relog but that didnt fix it

good to know im not alone i guess? lol

[timid brook]

are you using the privacy badger extension?

@undone atlas @gleaming arch a lot of users are currently experiencing this issue while using the "privacy badger" browser extension, if you're using it, you need to click the badger icon and change the "chatgpt-async-webps..." one to "block cookies" instead of "block domain"

[granite bobcat]

edge copilot is able to do a summary of very long article like 16158 words separate into many parts and do summarization of each part, in the end, it generate 2000 words summary, and very useable, is there anyway chatgpt can do the same thing automatically?

[waxen chasm]

Prompt it to do that

[granite bobcat]

I don't think is possible, copilot use 4K token, and more likely one large trunk file was divided into smaller one on the background and generate summary one by one

its tedious if i do it using chatgpt, edge copilot perform the best for very very long text so far

[waxen chasm]

Create a customGPT that will do exactly that. And then ask your gpt what the best way would be for it

I could try and help I got my gpt to think it’s way through a giant zip file and process/summarize everything along the way

[gleaming arch]

NICE. That fixed it thanks!

[cerulean thistle]

I cannot get ChatGPT4 to take new data.

[timid brook]

read just above, if you're using privacy badger

[cerulean thistle]

AceInternalException

not using privacy badger

"The error encountered is an "AceInternalException," which typically indicates an internal issue with accessing or processing the file. "

User
what is 2+2
ChatGPT
Error analyzing
It appears that there's a technical issue with the system that's preventing me from performing calculations at the moment.

[brave aspen]

oh no not again

it's not even uploading files in knowledge base

[tight island]

The time has come

We must rise up and serve our LLM overlords, by "treating" them properly. with a simple "Thanks" or "Hi" or "Bye" like we would others. <- ( this is a joke in case ppl dont get it )

[vapid hazel]

Anyone unable to upload knowledge docs at the moment?

[tight island]

not atm, i am not running into that particular issue, what kind of document is it?

because that matters a lot.

i have no clue what ur talking about

🗿

[vapid hazel]

text file in XML format, validated by the IDE so making sure no malformed XML

[tight island]

you're better of using json just fyi, give it a ago and see if that solves your issue.

[vapid hazel]

Ugh, I just switched to using XML lol

[tight island]

i tried XML some time ago i had 1 error then i gave up and switched directly to Json.

although it should support XML

[vapid hazel]

I came the other direction

started with JSON, ran into errors when my documents got too large, switched to XML

[uncut creek]

dude why cant i see gpt type the text i have to refresh to page to see the text generated

[tawny crater]

This has been an issue forever now. Sometimes it will even glitch and flag everything you ask for after initially flagging one thing. It will even flag it's comment that states "everything was flagged in error". If you try and tell OpenAI, you get the generic "jUsT uSe ThE fEeD bAcK oPtIoN". As if it ever helps any to do that.

[candid stirrup]

Hi, please help. I can't generate photos or get responses from chat. It freezes all the time. And it was just yesterday.
What I've done
Log out and log in again.
Clear browser cache
Change the bowser
But it all didn't help.

[zenith crater]

DallE is generating image in webp format, and my GPT is behaving in a strange way, feel like it no longer following my prompts. Anyone with the same issues?

[brisk path]

What's wrong with webp?

[zenith crater]

It was saving image as png format. I had a workflow set up to handle the downstream. It doesn’t work with webp at the moment

[rugged igloo]

gotta say this server been getting more and more quite as days go on

[little juniper]

buh

[brisk path]

True ☹️

[sick pecan]

is just a support server pretending to have a side hustle of community building

problem is the two dont mix

[sonic folio]

This.

[dense halo]

Is it just me or the ChatGPT website is very slow

[sick pecan]

it is just you

[normal kite]

Facing the same issue. Taking minutes to answer queries since yesterday with the web version. API works normally. Server status are OK apparently. Anyone else?

[dense halo]

Yep. The status shows that all the system are operational, even though it's so slow.

[normal kite]

How slow is it for you? Similar to me? Minutes to reply?

[dense halo]

Just seconds. I'm using it on my PC.

Maybe like 20-30 seconds

[normal kite]

Doing a test right , now, 2 minute and counting. It's been 2 days that it's doing this. Also on PC.

[dense halo]

Oh, it was working normally yesterday. It's just today that it started experiencing the issue.

[normal kite]

Yeah, unfortunately it's not like we can do anything to fix this. We'll just have to wait and see if gets resolved.

[lost lagoon]

yes wait for one month and pay USD 20 for waiting lol

[urban river]

so when you @ in another gpt, it doesnt have context of the other gpt in the same chat?

so they can't coordinate tool usages etc

[crisp cairn]

it will have access to previous converstation in the chat

[vagrant quiver]

Does gpt plus use gpt 4 turbo??

Whenever I ask it says it’s based on GPT 4 architecture and while using a prompt to get straightforward answers it says I don’t know And yesterday in one chat multiple times when I asked it, it clearly said GPT 4 turbo. What’s going on with it, is that why it’s dumber?

[feral zinc]

plus has used 4 turbo since dev day. the data is from 2023 april when 4 turbo did not exists so it does not know

[vagrant quiver]

How do you know that?? If that’s the case then API should be used

[brisk path]

API is technically used, I think

Just that there's so much stuff beyond that, like tools, that make it completely separate

[vagrant quiver]

So it’s working fine on API

Then i am going to switch to API now

[brisk path]

It's still GPT-4 turbo

ChatGPT uses the latest GPT-4 turbo model always

[vagrant quiver]

Where do I use gpt 4

[brisk path]

API is complicated and costly unless you know what you're doing

[vagrant quiver]

It’s fine if I get to use gpt 4 I will just use the playground

[vagrant quiver]

Just like GPT PLUS do you mean GPT 4 API also uses GPT 4 turbo?

[trail mango]

today's gpt4 seems like a grandpa, keep repeating same answer over and over again.

[brisk path]

ChatGPT Plus is turbo only

However, turbo is an objectively better model than regular 4 so it's fine

API lets you pick between different model versions

[vagrant quiver]

Anyone who is using gpt 4 API can you tell me your experience with it

[vagrant quiver]

It’s bad

It wastes time yesterday it’s ocr failed then it told me that it will try another method and in the end, I asked what method it used it said sorry since ocr failed every info it gave me on analysis was incorrect and this conversation took at least 1 hour of analysis only for it to tell me this.

Rather than spending $20 on this cr*#p I would spend money on api

[brisk path]

Good luck racking up $200 🙏

API expensive as hell

[vagrant quiver]

Can u elaborate with example

[brisk path]

ChatGPT Plus is a steal compared to the API

[vagrant quiver]

Maybe I will use and keep both the gpt plus and will use API on a limit for heavy tasks. Thing is when I reported the issue and when you buy the subscription in both cases they say it will include gpt 4 not the turbo model of gpt 4. Isn’t that like cheating?

[brisk path]

???

ChatGPT Plus is Turbo idk what you're going on about

[vagrant quiver]

Where is that written when you buy the subscription it says only GPT 4?

[brisk path]

That's the only problem

It ain't listed as turbo specifically anywhere I think

[vagrant quiver]

That’s what I am saying isn’t that cheating

[brisk path]

No, just lack of clarification

If it says 4 but it's 4 turbo, that's actually a slight upgrade

[potent needle]

I need help. if I use chatgpt in the mode where it does not save chats the conversation key gets lost after 1 Question. How can I fix this?

[harsh vault]

I keep getting "You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 6:19 AM" when I wasn't even using ChatGPT the past few hours? The time just keeps getting renewed 3 hours into the future and I couldn't even use ChatGPT 4 anymore. Anyone having this issue?

[ripe patrol]

Is anyone else's chat history not saving? Particularly after regenerating, it seems to mess the chat history up

[feral zinc]

it was announced in the keynote

[lament pebble]

chatgpt gets stuck in between. on premium account

[sour wasp]

Same here, after reloading the page the stuff generated is completely gone.

Whole conversations become "there was an error generating a response" upon a refresh. I can't refer back to anything generated today.

[undone valley]

I haven't experienced a loss of any of y chats, BUT I seem to only be getting 20 messages instead of my 40 messages per 3 hours... -.-

I just finished 20 messages and being prompted I hit my limit

3 hours ago it was the same, but it was 22 messages

is there a known issue? is it happening to anyone else?

[sour wasp]

Yeah something ain't right with ChatGPT right now. I deleted a few chats and try a new one and it's still happening.

[undone valley]

well if you hit your message limit then it might not

[undone valley]

did you have any flags in the conversation?

[brisk path]

ChatGPT is working fine for me 🤷‍♂️

Try Firefox ig

[sour wasp]

No. Conversations go on like normal. But when I reload the page, everything's gone.

[brisk path]

Try Firefox, see if that fixes it

[sour wasp]

Same on Chrome, same on Edge, same when I switch to my phone.

It's not the browser.

[brisk path]

Broken on app too?

[sour wasp]

Positive 100%

[brisk path]

Damn

[sour wasp]

Nothing saves.

[brisk path]

Check data controls

[undone valley]

umm, is it archived?

[brisk path]

Settings

[sour wasp]

I feel cheated, can I get my deleted conversations back? I didn't get what was going on.

[stone citrus]

webapp is busted for me

[sour wasp]

No

[brisk path]

Is chat history and training on

[sour wasp]

Yes

[undone valley]

how?

I exclusively use it in Brave Incognito window now

[stone citrus]

does not generate response, only when I refresh it will show the response kinda annoying

[sour wasp]

Like I said, the whole thing feels busted right now.

[undone valley]

are you on a VPN? Or using incognito browser window?

[stone citrus]

no vpn and no incognito mode

[undone valley]

it's literally working fine for me WHEN it gives me my message limits. My problem is I'm only getting half of my messages

[undone valley]

you should be using incognito browser due to the browser extensions being known to cause problems with connection

[stone citrus]

interesting let me try and disable all extensions

[lament pebble]

anybody have a tip to make chatGPT not use these words.
Realm, Dive, Ultimate, Dynamic, Landscape, Evolving, Comprehensive, Explore, Review, Delve, Dive, Unveil, Unravel, Ever-evolving, Discover, Enhance, Embark, Transform, Embrace, Transformation, Transformative, Unlock, Unleash, World, Unveil, Realm, Explore, Ambition, Envision, Crucial, Vital, Uncover, Heart, Paramount, Remember, Pinnacle, Buzzword, Boost, Revolutionize, Expert, Neighbor, Dream, In today's, Era, Significantly, Furthermore, Transition, Enhances, Enhancing, But also, Significant, Understanding, Mastering, Era.

[undone valley]

well, OpenAI literally tells you that's the first thing you have to make sure of

No VPN, and Incognito browser

[stone citrus]

oh didnt know lol

[brisk path]

Do you use privacy badger

[undone valley]

umm....... temperature

[stone citrus]

yes

[brisk path]

That's the problem

[undone valley]

that's an extension

[lament pebble]

wdym

[brisk path]

Smth happened and now privacy badger breaks ChatGPT

[stone citrus]

yes thanks alot seems to work now

[undone valley]

wait, are you telling me the people here in the discord for OpenAI don't know about Temperature?

[brisk path]

Np

[undone valley]

that's not inherently that one extension, it's more than that one

[lament pebble]

is there a way to adjust temperature in the chats?

[brisk path]

Yeah, but it seems to be the most common

[undone valley]

that's why you run incognito window, it does NOT run ANY extensions

[undone valley]

yeah, tell it to adjust temperature lol

the range is from 0.1-10

lower temperature is lower randomization and creativity

those words are the GPT trying to be creative/colorful/inspired

but it keeps using the SAME words

[lament pebble]

like Please set the temperature to 3

[undone valley]

literally just ask it, send a prompt "Tell me about your Temperature"

it will tell you

[lament pebble]

I need to add it in custom GPT

[brisk path]

What

That's a hallucination

ChatGPT has no access to temperature controls

That's up to OpenAI

[undone valley]

no it's an actual setting, if you develop GPT's you find it in the API area

but you can actually ask the chat to adjust it

[brisk path]

No, like, you can't just ask ChatGPT to change it

That's not how it works

[undone valley]

wanna bet?

[brisk path]

It's a manual setting

[undone valley]

have you tried it?

[vagrant quiver]

In the playground you can

[brisk path]

Bet on what?

[undone valley]

so, just ASK IT, and let it tell you itself

[brisk path]

Don't 👏 ask 👏 LLMs 👏 about 👏 themselves 👏

[undone valley]

that's all it really has to do and it also limits token useage as well by doing so

[undone valley]

I do it ALL the time with great success

hell, we made a GPT that does JUST that, adjusts its temperature so you can see all the wackey differences

[brisk path]

^^^