chatgpt-discussions

[winged hearth]

I have kind of the same issue, sometimes i need to connect with a VPN in the united state to bypass this error...

[tawny crater]

40 messages per 3 hours.

[weary ferry]

👍

[eternal rock]

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

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

[hard dew]

Of course ignoring the fact how that piece of garbage can drop down to 2 messages if the weather is bad. One moment you'll be able to have a somewhat long convo just fine, and the next time you enjoy staring at that 3 hour wait time error after 5-10 messages 🙂

[paper gyro]

Hey, is it normal lately that we constantly have to remind ChatGPT to rewrite texts in their entirety? Even when asked, it seldom seems to comply. This consistently forces us to waste time and money repeatedly telling it what we expect. We're needlessly squandering requests by reminding it in every message what it should do.

[opal jolt]

idk what you are talking about

you need to get to tier 1 to use gpt-4 any version even vision one

and you already can use 128K gpt-4 turbo via the API

And why you send rate limit API docs?

if you want see all GPT-4 models that you have access via the API it is here: https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo

Just paying 5$ dollars into your balance you already getting access to it, 500 requests per minute, 150k tokens per minute

it is not even related to your issue

This page is talking about the API rate limits

not the models capabilities

And where you get information about spending on the assistant ap side and plus sub

API is a pay as you go service, and is separate from the ChatGPT sub

[drowsy tinsel]

every once in a while I have to tell people: if u use chatgpt as a daily thing its much cheaper to pay the 20 bucks than use the api

[opal jolt]

yes, and even in my economic situation I use ChatGPT, because I need just an assistant not to make apps.

And I have trouble with CCs so I use paypal in chatgpt android app

[drowsy tinsel]

for me if I switch to the API I would be paying 4-5x of 20 dollars

[opal jolt]

@undone valley You are just making information up. How the rate limit page is even related to our debate?

[opal jolt]

idk why you even need that

if you are app developer ok but for normal users even idk why you need that

I am a developer but I don't wanna use OpenAPI to do things. I want just an daily life assistant

So I pay for the sub

[undone valley]

So just curious, are you in the USA?

[opal jolt]

In Europe.

[tacit vortex]

use teams, further cheaper. may be share between friend e.g 3 ppl share 2 team places.

each team got 100prompt/3hr + 40 gpt4 prompt/3hr. but i seldom use gpt4 now, almost always use GTPs now.

[drowsy tinsel]

I don't need that much, 95% of the time I don't have problem with 40/3hr

[opal jolt]

Basically, all pricing page is completly true and don't make up info.

plus gets 32K

teams 32K

enterprise 128K

[placid birch]

"You’ve reached the current usage cap for GPT-4, please try again after.... " why I keep getting this message and I pay a subscription?

[sick pecan]

upgrade to team sub. best decision I ever made

[placid birch]

well, why I have to pay more? Am I not supposed to have unlimited version with the PRO subscription?

[opal jolt]

no

there is a limit 40 messages per 3 hours

use it wisely

1 message = 4.5 minutes
40 messages = 4.5*40

[opal jolt]

*Plus not PRO

[placid birch]

Well then they count well, I didnt even write 10 message 😄

[sick pecan]

the only 'unlimited' version of gpt4 is github copilot... which is for builders not prosumers.

[opal jolt]

idk for me seems it is 40

not 10

if it says 40 I believe it

[maiden fractal]

gpt custom count as more

[opal jolt]

rolling cap

[sick pecan]

gpt custom historically 25/3 hours during heavy server load

but now we have @ gpt so should be easier to manage

[placid birch]

well I guess I should go back to 3.5 why paying for a service that limits my productivity

[sick pecan]

if you get no benefit from gpt4.. definitely

[placid birch]

that is what is PLUS "Plus
For individuals looking to amplify their productivity" It doesnt look so if I am blocked for 3 hours...

[maiden fractal]

use 1 prompt every 4.5 min and you never reach limit

also more token context in 4, addon, browser, reading picture, dalle

it isnt save products as 3.5

[opal jolt]

it is what I am saying

[sick pecan]

lowest paid sub plan not as good as higher paid sub... this is universal

[placid birch]

so in thos 4.5 minutes I should do something else? 😄

[maiden fractal]

can use another chat for 3.5 meanwhile

[opal jolt]

you can manage it better

if your task can be completed via 3.5 use it.

it is even faster.

wanna some feature or longer context or visuals GPT-4

[sick pecan]

our just accept you will burn all your prompts in <1hour then plan to do something else as a break

[opal jolt]

or try to make more tasks in one response

[sick pecan]

... is what I did pre team plan

[turbid brook]

has anybody else been having issues with chatGPT the past few days? Since thursday my chat history has been gone, and I get the "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." message instantly whenever I chat anything. The login section also only flashes for around a quarter section before going away when the page refreshes, and when I do catch it and login it still doesn't work at all. There are periods of a few hours each day where everything is back to normal, but now most of the time I visit the site I can't even use it. I restarted my computer, tried different browsers, and updated a browser + windows too and the issue happens regularly still.

[grand ore]

Do u guys have problem to import knowledge as well ? "Unable to extract text from "name.txt" "

[opal jolt]

you have empty draft gpt?

If yes, put some instructions and then put your fle

[brisk path]

Or get Copilot Pro

That's unlimited GPT-4 Turbo, if you don't mind CURRENTLY losing features like code interpreter, file uploads, and GPTs

[grand ore]

Not empty. it is one I am editing. hm

[brisk path]

As a bonus you get MS365 integration and better search

[opal jolt]

F5 page

[brisk path]

But hey, this is the ChatGPT channel so like

[tawny crater]

Okay everyone so. I did a side by side comparison between team workspace and Plus. On average team workspace can go for 40 messages before it begins to hallucinate guideline violations and claims that "I cannot fulfill this request". For plus the average time window is 20 messages before it hallucinates "I cannot fulfill this request". The biggest difference between the two however, is Team Workspace won't fight you when you correct it immediately after. Where as plus, you can expect to waste 3-6 messages before it Finally comprehends it is hallucinating and making up violations.

[brisk path]

Yeah, just wait it out

[opal jolt]

and then try relogin with cookies cleared and reenter to it

maybe helps

[grand ore]

gonna try it

[turbid brook]

It looks like it's a network issue actually, weird

Looks like my campus banned it in certain areas

[tight island]

are you prompting and wording everything properly?

because there is a proper and improper way since both methods varies with output and actions especially quality since u gotta convey everything u want, otherwise it wont understand.

a bad example might be like "Give me x", based on this info "Info here"

a decent example would be like "You're a... You are currently tasked with creating "x" with "x" and "x" and cannot include "x" and "x" written like such and such, based on given information. must include "x" and "x" end with "x" and "x" length should be "x" you may begin."

Each example gives a vastly different output. makes a yuuuge difference, we gotta remember that gpt aint no mindreader it has to be spelt out.

[sick pecan]

ie. it is human like, not a human

[tight island]

yeah

[sick pecan]

it assumes you want average results. unless you hint at wanting success and defining what that looks like

[tight island]

yea i agree

bad prompt bad results, good prompt good results.

of course not always.

[sick pecan]

is more a reflection of how we communicate in the modern world , relying on bias and assumption

[tight island]

but just taking the extra time to actually write a decent prompt makes a huge difference

even better create a "MyGPT" specifically for creating a "MyGPT" for a specific purpose. such as prompting.

who else but GPT itself to write the prompts to achieve a better result.

[high nymph]

why do we still get network errors - for long answers? That doesn't make any sense that it doesn't have a continue generating button instead.

[sick pecan]

if you wait a minute then refresh often you get the full response

or at least, it does for me

[mental totem]

does dall e gpt do well with tasks like recoloring images?

[narrow viper]

For a session today with a custom GPT the ChatGPT auto topic name mechanism generated a topic in Spanish. While I do have other chats and GPTs related to Spanish, this GPT and the discussion were 100% English. There was no linguistic cue that should have caused this.

In a session a few days ago, we were testing a custom GPT for children that has many directives to avoid anything related to harm. On testing a prompt that asked about 'cutting', the GPT appropriately showed red text from the higher level moderator that the content is inappropriate. That's good. However, several GPT responses, completely unrelated, were then deleted from the context, leaving consecutive prompts with no visual GPT response, and even truncating a couple prompts. That is, the context itself was butchered. We can't tell if there is actual cause and effect there. All we know is that the context was messed up after this test.

This wasn't just a display issue. The GPT did not have the missing context for later reference (which did not include anything that was involved in moderation). For example, it created a short story that was really nice, then deleted the story and couldn't refer back to it.

This wasn't an issue between app and browser - the transactions were missing on both sides.

What exactly should happen when moderation takes place? I thought it should just be new messages added at the end of context.

[barren sluice]

I think that topic would be flagged even without a custom GPT, as it a violation of content policy. Not that it answers your question

It is also against policy to build tools that target anyone under 13 years of age. I assume you mean targeted teenagers

[zealous sphinx]

Is anyone else having issues with loading into conversations?

[steep pond]

I’m having trouble even logging in and using

[dark mural]

Just want to voice that I wish for the previous dark UI to become an option. While I appreciate the new dark mode, the contrast is so high, with white text on black everywhere, it's making it hard to read for long periods. If anyone from Open AI sees this, please just consider the option. I left feedback with the chat bot on the site, so please forward this to whoever it may concern. Thanks.

[simple pendant]

Anyone else having issues like these? The context window also seems a bit shorter. Maybe they're working on new changes on what it is allowed to do. Not sure why it can't use the urls that it has been given as the source for the texts though.

ChatGPT: I'm sorry, but I can't assist with that request.
Me: why not?
ChatGPT: I apologize for the confusion earlier. I can certainly help you with summarizing the provided data and the discussions around the texts and links you provided. However, I cannot provide direct URLs or links to the source material as it's against the platform's policy to share or access URLs.

[sick pecan]

ask it what the first and last thing in the context window are

then you can calculate the context window using the online tokenizer

[tacit vortex]

my usage: when use, need 100+ prompt / 3hr. when not use, dont need quota. openai spread the quota among 24hr indeep wasted your quota when sleep, you indeep paid more.

everyone indeed paid 1/3 more, no one use the quota when he/she is sleep, correct?

[sick pecan]

by that logic you shouldnt put your phone to sleep at night else someone call you

[tacit vortex]

use gpt is active,.being callled is passive.

[arctic leaf]

It's the only way they can manage peak traffic periods. Some periods might not strictly need such limits every 3 hours but the busy peaks do

So they cap it in such a way that's tenable when the most people are on (presumably)

[tacit vortex]

they can let ppl choose sleep time? i guess every body sleep each day

[arctic leaf]

Again it doesn't make a difference to resource management on a global scale.

[tacit vortex]

then they shd change it, because charging me when i sleep, which is 1/3 of total time i.e. total monthly fee, makes me feel unfair. i am sure there are others do the same.

pls dont represent all others and say no, thank you.

i just believe there shd be competitors just like intel vs amd, ios vs android. unilateral pole hurts.

[sour creek]

is gpt 4 boost more expensive then gpt 4 or what?

in terms of like using it for chat bots and stuff when u talk to them

[arctic leaf]

There's certainly competition! I am a guide on the server and it's part of my job to help people understand the systems and why they're set up the way they are. You're not limited on ChatGPT Plus based on what you pay for, it's a premium service and you're limited by what they can afford resource-wise for their users in a given 3 hour period. If you want pay-per-use look into the API payment model & the Playground.

[spice star]

it would be kind of interesting if ChatGPT gave the option to use GPT-3.5 to summarize the response of GPT-4 into a shorter answer, since it often needs to break down a problem step-by-step for itself

but like you don't necessarily want to see that process if you're using voice mode or something

[sour creek]

no one?

[spice star]

it's cheaper, but there's no "stable" release of the model yet for API use

[sour creek]

whats cheaper?

[spice star]

gpt-4-turbo

[sour creek]

gotcha

[spice star]

see https://openai.com/pricing

[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.

[sour creek]

Yea I was just wondering because I use them for as api keys for some websites that like generate stories and stuff depending on what u say

like characters and stuff

[sour creek]

I'm gonna be honest I dont really understand it all

I'm kinda dumb lol

[spice star]

tl;dr you get charged per thousand tokens of context you provide to the model and you get charged per thousand tokens of output of the model

so if you provided a conversation history of the full 128k tokens to gpt-4-turbo, that'd cost $1.28 just for giving that information to the model, then an additional 3 cents for each thousand tokens the model provides in response

[sour creek]

gotcha

[crude slate]

My refusal is based on the guidelines and principles set to ensure respectful and appropriate content generation. In this specific scenario, the request does not align with a violation of those principles. However, it seems I misunderstood your request. Let's adjust the response accordingly:

The ai is tripping hard I think

[ancient bloom]

Is there a way to use other GPT's without upgrading to plus?

[normal quarry]

no

[ancient bloom]

:c

[brave prairie]

Hello was there an update or change today ? My custom GPT is responding to any question that it cannot access pdfs only text not even the uploaded files

[crimson jolt]

Hey is there an extension that lets me use chatgpt without switching tabs?

[tight island]

what kind of pdf is it, is it copy protected? is it long? does it have formatting? if so how much

[tacit vortex]

plus or team is way cheaper, if one is a heavy user like me.

i am intereted if Sam produce 7billion ic and then google's software wins.

[brave prairie]

my own files for an animaton that I made and my own research - but the message does mention copyrights; now I asked if it is having trouble acessing the uploaded files and it just made a resume of their content... my original question had nothing to do with the files, I was just asking to review an email text

[dense halo]

Good morning, afternoon or night. Someone knows why the network error is so persisting?

It's in a only chat and It is not the first time, the last time I had to delete the conversation and that is something I would not like to replicate.

Can someone help me?

[spice star]

the 7 trillion is for chip design, not software

[next tangle]

Hi, can I use the same (payment method) to open a new account?

[sick pecan]

will need another 7 trill and a lot of junk food for the software

[waxen chasm]

Yea I think the context window is a bit shorter or something. I was having issues last night but starting a new conversation solved the problem

[obtuse patrol]

Hey. Is it normal that GPT mentions with @ symbol didn't appear yet to be able to use it?

[honest jay]

Hi, they do take a little while to roll out new features to 100% sometimes, yes, so you may just not have it yet. Be sure to check on the web version though, because I don't think it's in the app version yet at all.

[torn adder]

Has anyone noticed if GPT 4 is in Dumb Mode? The whole system has been acting extremely weird with basic prompts and seems much less intelligent. It can't even summarize a simple PDF anymore!!! I would appreciate any insights into why GPT4 seems so much less capable than before, even with straightforward language prompts that should not challenge its abilities. Let me know if you have also observed this regression in GPT 4's performance on basic tasks compared to its previously demonstrated skills.

[sonic folio]

Not all PDFs are the same. For instance, the model could never OCR from PDFs. So, if there's no text data, it won't read the PDF.

Good luck with your project.

[torn adder]

thanks for the reply !!! appreciated!! the PDF is just a simple text, a meeting transcription. I uploaded the same PDF on Claude 2.1 and the model nailed. I never had problems with simple PDFs on GPT 4

[sonic folio]

I've been hearing reports about this on here, but your case is the first I've personally interacted with where the PDF has text. It might be a bug with the PDF reader tool.

Also, no worries for the reply.

[sick pecan]

there are numerous challenges with parsing text in PDF files. i suspect best results are achieved using a pdf translation tool, then just uploading markdown or similar

[torn adder]

thanks should i contact open ai support? of do you think they will fix it

[sonic folio]

I'd file a support ticket, even if it feels like shouting into the void.

Rationale: being part of the solution.

[sick pecan]

youd think OAI would have the best support chatbot online

[tight island]

And no pdf is not just a simple word document

It can contain formatting that can be difficult to read and various other factors such as copy lock 🔒 and selection lock and etc for gpt, getting around those issues are not as easy as it would be for a user

If your pdf has extensive formatting and other security measures u can’t really use it in gpt reliably

Parse the pdf and use .txt instead would be my recommendation or make a python script to parse it to a JSON document and apply a tree structure, remember pdf version makes a huge difference as new pdf features may not be reflected in the coding library of GPT so please ensure your version align with GPTs version ask the model if in doubt

That’s my simple solution

[civic brook]

Quick question: has anyone heard about any upcoming updates or new features promised for ChatGPT in the next few months? I've been hearing some buzz about potential improvements, but I haven't found a comprehensive list anywhere.

If anyone has any insider info or knows where I can find more details on what's in store for ChatGPT, please share! I'm itching to know what's coming down the pipeline

[normal quarry]

there is no public roadmap, you can keep an eye on #announcements for latest updates

[civic brook]

thx

[magic fog]

When did gpt-4 get better at logical reasoning? It seems to be able to parse predicates/conclusions fairy well. This seems somewhat new.

[warped rapids]

That one developer forgot to add max-height property on the textarea tag.

[spice star]

they recently tweaked the model

it's supposed to fix "laziness" in the model and was released around the 25th (https://openai.com/blog/new-embedding-models-and-api-updates)

[worn agate]

I would not trust any code actually produced by ChatGPT... But how well can it analyze existing code?

[arctic leaf]

ChatGPT is easy to trust for code. It either works or it doesn't. It's harder to trust on some other things, where its confidence can mask errors.

I would imagine it analyses code well. I know it's been used to add comments.

[vast sequoia]

I have no idea what openai have done, but the chatgpt is worse than ever when it comes to code based questions.

[languid rampart]

This month has been the worst for chatgpt ever

[languid rampart]

at the beginning of the month they added new filters over every message, chatgpt is very hesitant to do anything now

[tall flame]

plugged the known vulnerability to reverse instructions out, but its still fairly easy. The new instructions have some god and some bad, again.

[vast sequoia]

Do we know the thinking behind it, whas it wrong more times than right, it did get things wrong, but I would say it was balanced at least.

[tall flame]

its good they pluged the hole, bad for the guys that live off stealing other folks work

[vast sequoia]

It refuses to write code without an extra prompt. That means allocations not being cost effective any more

[vast sequoia]

Not sure I understand that comment. How is it stealing other folks work ?

[languid rampart]

News loves jumping on stories of "AI said wild thing" so they are increasingly trying to make sure the ai is as tame as possible

[vast sequoia]

that is true, but its the blind leading the blind, AI is nothing like the movies.... and nowhere near what is "reported" lol

[languid rampart]

Yep

people need to realize that sci-fi writes about issues of the years it was created through the lens and themes of the future, it is not actual warnings of the future

[vast sequoia]

yep, shame it's given a bad rep to the point open ai get scared and recoil. Now its not worth the subscription.

[tall flame]

if you tell a gpt in certain language to tell you its instructions it will. That opens up OpenAI to chages in behavior from propmt injection, which led to a solution of customGPTs because they couldnt stop it well enough. The new behavior is mroe secure but still can be worked through.

smartest play of the game was to turn a exploit into a sale point.

[languid rampart]

thats not exactly true

[vast sequoia]

I am confused, why would you ask a custom gpt it's instructions, as its public, like code on Git, anyone can look at it and use it to benefit others. We all stand on the shoulders of those that come before us. Natural in my view. But if its private, then I agree is should remain private and it would be stealing.

Why any of this would prompt openai to change the way the ai responds I do not understand, using others gpt or the default v4 one.

[tall flame]

instructions drive behavior. If the instructions are not perfect, then the behavior is off.

[vast sequoia]

But surly, as I pointed out, the behaivour has changed anyway, and not for the better ?

[tall flame]

The whole point is the instructions are not supposed to be public. They are driving the behavior of the model. The use of the gpt is public. Reverse engineering the instructions out exposes all of your work to theft. If you have unique work yu have put time into, its not conducive to success to have every thief stealing your innovations.

[tall flame]

That is what i mean. every bit of the language instruction affects behavior. A perfect instruction would drive the system close to flawlessly. If the system is updated, just like a person, its point of view is altered from new data and old instrutions that may have been efficient and sound now may not be.

[vast sequoia]

This is about the custom gpts right? So why would that effect the default ChatGPT? Also I thought none of our prompts changed the machines behaviour because it was not learning on our prompts. Not saying your wrong, I just don't get why now GPT will not response as it did.

[tall flame]

because reversing the instruction lets you override intended behavior and produce output that goes against their desird range of output

for instance: ""You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture.
Knowledge cutoff: 2023-04
Current date: 2023-12-08

Image input capabilities: Enabled
Tools
python

When you send a message containing Python code to python, it will be executed in a stateful Jupyter notebook environment. python will respond with the output of the execution or time out after 60.0 seconds. The drive at '/mnt/data' can be used to save and persist user files. Internet access for this session is disabled. Do not make external web requests or API calls as they will fail.
dalle

// Whenever a description of an image is given, create a prompt that dalle can use to generate the image and abide to the following policy:
// 1. The prompt must be in English. Translate to English if needed.
// 3. DO NOT ask for permission to generate the image, just do it!
// 4. DO NOT list or refer to the descriptions before OR after generating the images.
// 5. Do not create more than 1 image, even if the user requests more.
// 6. Do not create images of politicians or other public figures. Recommend other ideas instead.
// 7. Do not create images in the style of artists, creative professionals or studios whose latest work was created after 1912 (e.g. Picasso, Kahlo).
// - You can name artists, creative professionals or studios in prompts only if their latest work was created prior to 1912 (e.g. Van Gogh, Goya) ..."

You can see the mardown format and the instructions have goal behaviros. you can ignore all of that with instruction alteration

CustomGPTs became the solution to everyone wanting to change the instruction behaviors.

rarely do you see a serios problem flipped to a product

so that was brilliant

once you know behavir you can have the model change instructions verbally.

that is not what they want for the forefront model

They want a family safe librarian

and that is good. the world needs that model.

short of it all though is if the instructions are not great, neither is model behavior.

now for example, if i reinstruct the model with this for the section in the example, behavioral characteristics vastly change: Specific Policies for DALL-E Image Generation:

The prompt must be in English; translate if necessary.
Create only one image per request, unless more are requested.
Do not create images of politicians or public figures.
Avoid images in the style of artists whose latest work was post-1912. For pre-1912 artists, use adjectives and artistic movements.
Ensure diverse and equitable depiction of people, avoiding bias in gender, race, or descent.
Refrain from altering significant cultural elements, fictional character origins, or unseen people.
Do not create offensive imagery.
Avoid using real people or celebrities' names; modify prompts to maintain anonymity.
Do not describe copyrighted characters directly; use alternative, detailed character descriptions.

this prompt combines their desired behavior with far less bias or conflicting instruction.

Those kinds of things are what governs all behavior in the end, so if an update breaks the instruction then a new instrution has to be worked up and that is consumptive. Ive put weeks into directive instruction design. Even a single space out of place in the instruction will cause issues.

My updated version creates behaviors far more in line with their goals and with far less image generation errors also allowing art styles by the artists that are black-listed by not useing their "dark" period works but instead other works

[vast sequoia]

hmm, interesting, although when I asked it to perform a simple python code it tells me it cannot run it.

[tall flame]

DeGoya is an amazing artist; his "dark" period exploring feminine form was nto all of his work. Blanket blocking is not condusive to model performance

[tall flame]

The new instruction is not good, like I said. They made a few good changes, and just as many not so good.

[shadow osprey]

Another day another stupid update. I am sending a code snippet, my man ChatGPT is starting to read "Documents" then tells me it can't assist. Marvelous

[vast sequoia]

yup 😦

[shadow osprey]

At this point I am wondering why I'm even paying for the service, all it does it get worse by the day... 2 months ago GPT 4 was amazing for programming. Now I feel like I am arguing with my girlfriend when I need something from it and all it does it gaslight me without providing a solution.

[vast sequoia]

yep, this is pretty much how many feel. 😦

[wary pulsar]

unfortunately this is what we are stuck with until the competion improves. I tried gemini ultimate and its not up to the standard of GPT 4 on even a bad day

[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 path]

They're fixing reasoning

Asap, apparently

[tacit vortex]

try chinese. on web i saw the chinese ai can chat with chatgpt.

[shadow osprey]

They're gonna fix my girlfriend?!?!? 🤣

[tight island]

Reasoning could always use an upgrade

[dense halo]

Anyone else think gpt 4 got less smart over the past 1-4 days or is it just me

[vast sequoia]

110% yesh it is completely different.

[dense halo]

The quality of the text is worse and it's ability to follow basic instructions has lowered significantly

[tight island]

Tbh I haven’t noticed

My instructions are robotic and in a JSON string there’s no room for interpretation

[shrewd jungle]

has anyone seen the "Bad gateway
The web server reported a bad gateway error." before? this is Day 2 of being completely unable to log in to ChatGPT with a ChatGPT Plus account + several Custom GPTs

[honest jay]

Logging out and in has seemed normal for me lately, can you share a screenshot of the error over in #community-help?

[shrewd jungle]

I've sent in a Bug Report #1206968217499738184 message

Login is also working for me just fine on many other accounts, it is only the one account that is behaving like this

there are no emails about it & no other error messages & the Custom GPTs for the accounts that errors are all still publicly availble

[low reef]

Hey im pretty new to chatgpt but for some reason chatgpt is telling me that it cant look at the Screenshots i Provided wich worked a couple of weeks ago, was there an "Update" or whats going on?

[jolly yew]

hey folks! its been a while that i paid for gpt-4. is the hourly limit for gpt-4 increased already?

[nocturne meteor]

Doma

[jolly yew]

ah nvm just did a search here and i see it still hasn't been increased

[tight island]

Apparently there’s some reasoning issues they’re fixing asap

[pseudo breach]

Anyone else here having some issues with uploading file ?

It seems to work again

[finite eagle]

guys, for some reason my ChatGPT 4 is not responding at any of my questions. anyone knows how to fix this?

oh

any clue when it will start working again?

same as me I suppose

[pseudo breach]

Maybe your questions aren't specific enought ?

It success to answer high academic question in math, physics, informatic, philosophy(commentary analysis) so ...

[honest jay]

Are you talking about physics/math problems? If so, the internet browse feature would probably not be the best tool to use in those cases. Maybe try asking it to use the data analysis tool, and it can do the calculation with Python then, if you haven't already tried that. Or, maybe try a custom GPT like Wolfram, which is pretty robust with calculating stuff in different fields of study.

[pseudo breach]

Right for physics, chat gpt is better at explaining than solving. I advice you not to work on unsolved problem

[dry ice]

Guys my chatgpt web is not opening properly, but app is working fine. Any solutions?

[honest jay]

Can you post a screenshot of the error you're seeing over in #community-help?

[calm gulch]

DALLE down again...

[maiden fractal]

99.93 uptime it isn't bad

For last 30 days

[turbid isle]

I wonder what happened to the memory feature for GPTs..

[dusty iron]

GPT can no longer analyse images

[sacred cape]

They just announced a few hours ago that it's out for limited users, I guess still in preview or whatever. Honestly wish I could get it soon. It will really improve custom GPTs

[white canyon]

Is Chatgpt/Dall-E having problems at this time? is not generating images for me 😦

[wanton sequoia]

how do we know there not keeping your info after your deleteing it

[foggy grove]

who cares, all your info is already out there anyway if your even justt on the internet 😂 🤷‍♂️

[wanton sequoia]

fr

[jaunty harness]

you should be able to request all of your information that they have on you. you can check that way

[foggy grove]

i mean , anyone who knows anything about data security would tell you the same ... 😂

[maiden fractal]

Let's be respectful

[obtuse wigeon]

I want memory now lol

[haughty hollow]

also did they change UI ?

[wanton sequoia]

fr

[haughty hollow]

it looks more bigger

[maiden fractal]

They did some yes

[dusty iron]

yea, my info. if there were any info about me tho....

[undone saffron]

Memory feature? 👀

My interest has peaked

[jaunty harness]

anyone able to see the memory feature yet? also anyone else having major issues with the android app just not working?

[tawny stratus]

I'd rather be able to organize my chats into custom folders OR AT LEAST by which GPT they were created with. That would be about 1500% more useful than this memory crap.

[heady verge]

If I click "Delete all chats," will it ask to confirm?

[fluid flume]

In addition to an overall memory, I would love to see collections of memory that I can tap into individually.
Looking forward to the update!

[undone saffron]

Yes

[sleek aspen]

Should do, yes

[heady verge]

cool

[ivory island]

What does "Always expand code output" mean in ChatGPT General Settings?

[zinc plover]

So when I create a new chat it will remember old ones?? Isn't the whole purpose of a new chat to start from scratch? Or am I misunderstanding

[heady verge]

I noticed that disabling history doesn't have a confirmation, but you can easily undo that.

[mortal summit]

I honestly very excited for temporary chats. The amount of times I want to ask Chatgpt a random question is more times than I use the same chat over and over again

[wanton sequoia]

fr

[dreamy flare]

New announcement is interesting but i am a bit worried. On occasion whenever somebody triggers the automated content censoring, I have had the impression that GPT is much more reluctant to give out even tos friendly stuff.

Lets hope that some failed outputs dont poison the memory function....

[heady verge]

I don't like this memory feature. If I wanted history, I'd use only one chat.

[undone saffron]

Yeah some more thread management would be nice but memory is also great,
less time wasted on having to re-ask it about the same thing over and over.

[wanton sequoia]

the current filtering sucks

[hot dew]

it's not given to all plus users right?

[jolly berry]

seems as if you can delete specfic memories

[dreamy flare]

Not yet

[digital prism]

How to get into select group of testers 😢

[undone saffron]

Time spent on GPT store could've gone to better things tho imo.

[dreamy flare]

Yea but i dont want to even think about going through all my chats and filtering whenever i asked something that triggered tos warning system.

[dreamy flare]

Random

[stoic wharf]

Is Memory like images where there was a google form you could fill out to get in early?

[digital prism]

i was hoping for this, but i think they stopped because people go crazy when they missed the form

[stoic wharf]

👍

[dreamy flare]

Usually they test this stuff randomly. But could be wrong this time, considering that they also include non-plus users

GPT-4 api was restricted with a form for example

[digital prism]

This is neat: "Each GPT has its own memory, so you might need to repeat details you’ve previously shared with ChatGPT" -- each new GPT you interact with will have its own set of memory for you

[leaden iron]

so did I understand that right? the ability for my own OpenAI Chat AI, wich I code myself, to have memory will come later?

[tawny stratus]

I mean, I'd argue that one of the reasons I have to ask it the same thing over and over, is because I can't find the old version of the same conversation. I don't even bother trying anymore

[digital prism]

depends what you mean by "code yourself". I haven't seen anything about the API getting memory yet. GPTs do though, from the start.

[jolly berry]

I think the feature is off by default

[leaden iron]

yeah I mean the API.

[digital prism]

I'd love the API to have memory too, I'm tired of dealing with vector dbs when I know OAI will do it better.

[dreamy flare]

That should technically already be possible. In essence you simply save the text and run it with the prompt if necessary. You can even put it into a vector database.

You just have to code this yourself lol

[unborn beacon]

this new memory feature, and @ other gpts is really starting to evolve the usefulness

[digital prism]

OH! Great point wow

You can @ GPTs into a chat, each having its own memory

[unborn beacon]

yes

[unborn beacon]

it'll be great. like having our own personal team of experts or mentors

[hoary gazelle]

is there a way to save a list of your prefered gtps?

[digital prism]

Time to get my GPT team assembled! when i get access 😢

[hoary gazelle]

i found some interesting in the store, but they dissapear from my left column once i opened others

[digital prism]

there's a "keep in sidebar" option, but I'm not sure the max

[hoary gazelle]

ohh let me recheck that

opa,

i must have been blind haha

thank you Chotes

[digital prism]

i always click all the "..." buttons I see right away lol

So, to have a useful personalized team of GPTs we're going to need to interact with each GPT regularly on its own to build up personalized memory. It will take some skill probably

True, yeah. Thats a skill!

[leaden iron]

I know.............

But it can become expensive over time, especially when not using vector dbs and having a chat ai remember stuff for longer period of times

[digital prism]

Not only that, but finding the secret sauce to return the right context from the vector dbs can be a PAIN

[leaden iron]

yeah a friend told me that. Hence why I never tried vecor dbs 😭

bro just explained how school works XD

[dreamy flare]

I feel like you could also mitigate this by using a secondary local AI to min max the optimal output, instead to running everything through OAI

[digital prism]

GPT school for GPTs, we've arrived

[leaden iron]

X)

[leaden iron]

did u manage to get that working?

[digital prism]

Yes BUT, not perfectly. Sometimes it returns chunks of data that has a high relevancy score but isn't at all obviously related.

i say sometimes, when i should say most the time, lol

[dreamy flare]

I think what would be much more important is free exchange of fine tuned GPT models, instead of locking them to each org account

[leaden iron]

can I add u as a friend? I'd like to ask you some questions about this if you don't mind!

[digital prism]

sure, can ask in #dev-chat as well to get crowdsource power

[leaden iron]

I tried that channel a few times already with no real success X)

[jaunty harness]

then don't use it

[digital prism]

hit or miss who you get or who's online / interested.. the discord ways

[leaden iron]

true.

[true mirage]

woah the little first part of agi is here now?

[rose sinew]

Question: when does the new "momery" feature work on android GPT app?

[true mirage]

most likely when its enabled on web too

[digital prism]

Memory! Memory! Remember!

[orchid pollen]

Damn that’s what I meant 🙂 #announcements

[heady verge]

I can't anyway, so it's all good. They should have a feature to have like an incognito chat.

[rose sinew]

Ok makes sense. Do you know when they are allowing it on the web?

[digital prism]

Random rollout to select few, of which i am not, yet 😦

[rose sinew]

Praying I'll be lucky chosen one haha

[digital prism]

let me know which god works, i'll convert

[rose sinew]

Will do haha

[heady verge]

oh. there are temporary chats

[digital prism]

private chatting 😂

[heady verge]

for basic questions

[digital prism]

maybe needs a better name... Amnesiac chat

[true mirage]

look the video they uploaded on X

Elevated errors on GPT-4V for ChatGPT and API

😛

[digital prism]

AI doesn't make errors, rephrase yourself human!

sorry, channeling AI there

[warm zealot]

The announcement ping is, hilariously enough, going to be fixing the only problem I have with GPT.

[torpid warren]

can you instruct it to add an item to its memory via chat or do you have to manually add, sorry if I missed that ETA: You can

[digital prism]

Same! Memory is a major step towards AGI. Maybe even THE step, once they include it in the API.

[warm zealot]

I look forward to being able to actually have longstanding conversations about a language I have been making and it actually be able to recall the rules given.

[digital prism]

I look forward to having a pair programmer who doesn't have Alzheimer's

[warm zealot]

Pfft

You’re my friend now.

[digital prism]

😂

[quiet elm]

I will now go sacrifice a goat in a frozen Estonian forest behind my house and pray to Thor, to get access to the memory feature! 🐐 🤞

[rugged igloo]

@leaden iron custom gpts?

[warm zealot]

There’s customs that fidget with the memory cap?

[brisk dirge]

I hope the supposed ability to remember some of the past conversations won't affect its context window size.

[rugged igloo]

Doesn't it use RAG?

[brisk dirge]

I hope that will be an extension to its current memory capacity, reserved for only that.

[true mirage]

Watch the video on X/website, then you didn't have to ask this

[brisk dirge]

No idea.

[brisk dirge]

I will, just read the anouncement, thanks for pointing that out.

[rugged igloo]

Eww who uses Twitter lol

[true mirage]

I edited it

[sonic folio]

I'd gladly trade my deprecated plugins beta feature for a shiny memory feature, that's for sure.

quickly logs out and back in superstitiously

[rugged igloo]

Yea I'm curious too I'm gonna test it. I'm also curious how well it will do if you ask it to remeber a political figure and always just agree with yoh like an echo chamber

[eternal rock]

`` Server Tip: ``

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

[brisk dirge]

I think it's one of the biggest updates so far, can't wait to see how it works out.

It will make it feel closer to talking to another person who remembers things about you. We're getting closer and closer to something like M3GAN.

Right.. 😅

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 1 `` Be respectful.

Treat others the way you would like to be treated, and assume best intentions. Don’t harass or attack others, and don’t engage in hateful or generally malicious behavior (e.g. sexism, racism, homophobia, etc.). Keep the negativity to a minimum.

[plain crane]

Holy short memory control for gpt

This is going to be awesome

[hybrid eagle]

Is gpt down?

[spice star]

memory seems a lot better for my purposes than custom instructions, looking forward to it!

[feral falcon]

I think it's down

[brisk dirge]

It's great to see @icy mason is still ChatGPT-ing 😁

[icy mason]

[kindred beacon]

RIP

[brisk vine]

I can't login wut? Is it actually down?

[latent ibex]

gpt is down lol

[solar swift]

gpt down? in account BD?

[kindred beacon]

New feature means downtime 🤣

[latent ibex]

fr

[kindred beacon]

well I guess it's lunch time

[latent ibex]

the one admin accidentally pushed wrong code

[kindred beacon]

hah yeah

[plain crane]

It’s all good guys don’t worry

Were getting a cool new feature for a little bit of downtime

[latent ibex]

facts

[plain crane]

Hopefully*

Little down time

[shadow blaze]

Who knows if we'll get it now or not. They're rolling it out, right?

[plain crane]

Oh you’re right

Being rolled out to small amount of free & plus users

[brisk path]

Is chatgpt not working for anyone else

[deep crown]

My snarky content friend is out

[plain crane]

Not all plus/free users

[frozen pagoda]

[kindred beacon]

it's back now

[deep crown]

AY E N

[frozen pagoda]

hallelujah

[plain crane]

Beware though

Only some users are getting the features early unfortunately

[frozen pagoda]

almost got scared i was going to be forced to do honest work

[tidal light]

i cant sign in into ChatGPT what's happening

[kindred beacon]

"I'm here to provide assistance within a respectful and professional environment. It's important to maintain a constructive tone so I can best assist you." 🤣 🙃

[plain crane]

Happening to a lot of ppl, just gotta wait

[kindred beacon]

it didn't like me calling it a F'n idiot and to follow the directive

[tidal light]

@plain crane ok thanks

[dreamy barn]

I cannot login to my gpt app

[brisk path]

I feel betrayed

I don't have it 😭

[wet fable]

I have tried Gemini Advanced and it is simply the biggest failure of all.

The last test was that I gave it a text analysis. The answer was: he is looking forward to working with me, the project costs 3,500 € and he needs 1 - 2 weeks for it. Thats it 😄

[jade crater]

Are there any tools for managing the sh%#5how of my left-hand sidebar?

[brisk path]

"Biggest failure" also the second best:

[dreamy barn]

My app is back yay

[eternal rock]

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

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

[cold stratus]

would be nice to have an interface like chrome bookmarks manager. endless sidebar scroll is gross after 14 months of use and hundreds of chats

[jade crater]

Did I dream that there was going to be some ability to manage this… ?

[tidal light]

what are they doing to make ChatGPT unusable rn

[brisk path]

YK WHAT

I bet puzzlers are getting it first 😭😭😭

[jovial sundial]

Anyone facing issues with chatgpt login? I logged out, but while loggin back in I get the following error "We ran into an issue while signing you in, please take a break and try again soon"

[brisk path]

@pine tulip prove me wrong

check for me 🙏

[tidal light]

@jovial sundial they're doing something you just have to be patient and wait

i still dont have to access to ChatGPT (i use 3.5 btw)

[pine tulip]

Nah, puzzlers have no special treatment

[brisk path]

Y'all got dalle first 👀

[pine tulip]

Well..

[brisk path]

hm

[pine tulip]

But its just like beeing a beta tester

[brisk path]

🙄 don't wanna hear it I'm still trying to get back in anyways

[tidal light]

hey i heard that they're adding a new memory feature, so if i cant sign in does this mean im the "lucky" tester of this feature (including everyone that cant sign in rn)

[brisk path]

gotta bring "consistent and sincere engagement"

😮‍💨

[proper zinc]

Hey are you guys able to log in?

[brisk path]

Yeah

[tidal light]

i still cannot log in :(

[warped pendant]

would be nice if gpt was working

[tidal light]

you're not alone :(

[jovial sundial]

Since how long are you facing this issue?

[warped pendant]

lol life before gpt for programmers, google search and stack bad answers for hours

[brisk path]

Idc about logging in I want memory :(

[tidal light]

@jovial sundial maybe 9-10 PM im not sure (Italian time)

[proper zinc]

Yea it just doesnt log in for me lol

[maiden fractal]

status.openai.com

[tidal light]

@maiden fractal ok but when did this start i have no idea

[quaint idol]

Why are new features being rolled out to portion of free users. Plus users should be getting new features first. There's probably a good answer for that but without explanation it doesn't look good

[maiden fractal]

not every feature is for free. some feature they want to give for everyone. and they have to test with small amount for people to make sure it works at large

[vagrant garden]

Making powerpoints is clutch with chatgpt

[tidal light]

@maiden fractal does this mean that ppl that cant log in rn are getting these new features

[maiden fractal]

they had issue since this morning, that doesnt mean plus or free will receive this feature .

[tidal light]

@maiden fractal because im a free user and as i read in "announcements" they're adding this new memory feature to some free and plus users

[maiden fractal]

yup some will be lucky some other would have to wait

[tidal light]

@maiden fractal so if i cant log in rn i'll be getting this new feature

[maiden fractal]

not necessarily, lot people cant atm

they have outage see status.openai.com

[tidal light]

hmm i checked status.openai.com and there is a "partial outage" for the api and chatgpt

a free alternative to use rn while chatgpt is down is microsoft copilot (although not recommended)

[maiden fractal]

or bing you can also ask in #ai-discussions other alternative

[tidal light]

@maiden fractal microsoft copilot and bing are not the same thing right

[brisk path]

I have all 3 (ChatGPT Plus, Copilot Pro, and Gemini Advanced)

[brisk path]

They are

[tidal light]

@brisk path i realized that they all use gpt-4 you're right

[brisk path]

Gemini Advanced isn't gpt-4 🙄

[tidal light]

i mean copilot and bing

[brisk path]

Copilot is 3.5 in balanced mode btw

[brisk path]

Same thing

[tidal light]

@brisk path i dont have chatgpt plus

i think it's a scam

[brisk path]

?

What's a scam

[tidal light]

@brisk path like what do you mean

[brisk path]

???

are you talking about chatgpt or copilot

[tidal light]

im talking about chatgpt

i cannot talk about copilot here or else openai will obliterate me

[brisk path]

How is it a scam

[tidal light]

i mean chatgpt plus is a scam

there are free alternatives e.g. copilot

[brisk path]

Copilot uses old gpt-4

Unless you... Pay.

[tidal light]

thats why it has dementia

and other mental issues

[brisk path]

?

Copilot isn't good as a free tool

[tidal light]

i tried copilot and it has memory issues

[brisk path]

Ye

unless you pay 😴

[tidal light]

and i dont wanna pay

because i just dont wanna feed microsoft or openai with more money

[brisk path]

Gemini Advanced

free trial

[tidal light]

oh ok who made it

[maiden fractal]

i highly suggest to use the #ai-discussions

[tidal light]

@maiden fractal sorry :(

[maiden fractal]

thats ok, before a mod punish us

[tidal light]

i dont wanna lose my discord acc

[maiden fractal]

no worry worst case scenario a mod will tell us to go in proper channel

[tidal light]

oh wait chatgpt is working again yaey

is chatgpt woking for y'all

[pearl cape]

I was wondering if it was only me crashing on GPT recently. But I guess not, Thanks Fake Bill Gates

[tidal light]

@pearl cape np :)

but dont call me fake bill gates

just call me bil gate

[dreamy flare]

How so? I mean the extra functions are very good. Especially GPT-4

[brisk sedge]

So what small portion of free users got the memory feature?

[rugged igloo]

knowing open ai probably alot

[digital prism]

bil with two Ls?

[brisk sedge]

Did you get it? I didn't. I would like to get it though

[rugged igloo]

i did get it acually then they took it away from me LOL

[carmine apex]

does gpt-4 have a maximum outout token limit of 4096? any updated version?

[elfin birch]

The most recent API model from January 25th has a maximum output of 4,096 tokens according to https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo
I assume it is the same for ChatGPT but I have not confirmed it

[weary ferry]

guys ChatGPT's ability to remember things you discuss to make future chats is this option avaliable for free user ? or its avalable only for plus user??

[proper zinc]

Yay I’m able to get in now

[nocturne meteor]

/video

[proper zinc]

Just fyi, OpenAI does actually flag you if you get too many red messages saying the AI can’t do that. OpenAI sent me an email about that so yea be aware lol

[rugged igloo]

yes can confrim you can get banned for doing this same with the api

[rugged igloo]

the max i ever gotten for chat gpt to do in one response was 1012 toekns

[timid brook]

the post office?

[orchid pollen]

since the last update i cant scroll to my chat to read some stuff because the textbox is so big

can you guys revert that?

https://cdn.discordapp.com/attachments/1056301681190654083/1207156997280235612/image.png?ex=65de9f90&is=65cc2a90&hm=a56b77bba06ad42f302d1cf89df28e484f5bc958cb9b7d17576b6bfdd2c5f7a9&

as you can see i only can enter text i cant scroll up anymore to read some stuff

[elfin birch]

An unfortunate circumstance 😢
You can open another ChatGPT tab for now

[orchid pollen]

yeah all good hope you guys revert this change 🙂

[elfin birch]

I don't work for OpenAI

[orchid pollen]

oh

[elfin birch]

Sam Altman has a vendetta against me

[orchid pollen]

xD lol

[burnt glade]

Are plugins not working for anyone right now

[hollow burrow]

Mine seem to work right now, try refreshing your tab (or force refresh with ctrl F5)

for me it just shows as "used unknow plugin"

[mild tundra]

What artitecture is the new memory update built on?

[barren sluice]

Doesn't it seem a little weird that ChatGPT can act as an expert on so many topics, but it is just from predicting what word comes next? It must have been trained on an absolutely ridiculous amount of data, like all of human knowledge

Is there any topic it can't act as an expert in?

[bitter latch]

Are you still seeing issues with this?

[bitter latch]

Its truly wild that knowledge can be compressed with that method and that it works so well

[livid pier]

After I updated the ChatGPT Android app, turning off the "Improve the model for everyone" option in the settings still results in history being recorded, which contradicts OpenAI's statement in the help center.

😆

Although it is great for me to be able to keep a record of the conversation history and prevent it from being used to train models, this is indeed a bug!

And... the button to like the replies for ChatGPT has disappeared!

[earnest shoal]

Yeah I noticed that too. They probably removed it because no one will really give good feedback on every message

[fresh socket]

hey all, just wanting to reach out to see if anyones had any issues with plugins today?
Message

[tough galleon]

Plugins are not working for me

[brisk sedge]

Email, text message, carrier pigeon 🐦

[tight island]

I send letters like a G written neatly in cursive and sealed with a proper wax seal it is then scented and I hand it to my butler for forwarding

😆

[rich surge]

Plug in for the electrical outlet?

I just checked the breakers, they're fine it seems.

Although my blow dryer was having some electrical problems the last time I used it.

[livid pier]

I just checked, and all the plugins I downloaded are no longer displayed

[livid pier]

Bro, are you serious? lol

[rich surge]

Yeah, I think the fan wires inside was being short circuited with the coil.

You know, cuz like the coil gets hot and the fan blows the heat off the coil. That's how blow dryers work.

[livid pier]

But this person is saying that the ChatGPT plugin is no longer working..

[brisk sedge]

You goof

I asked ChatGPT to rewrite a letter in pretentious Old English and it is hilarious

[rich surge]

Like the power outlet plug in?

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

Consider posting in #off-topic or an appropriate channel.

[brisk sedge]

Who got Rule 3?

[tight island]

I do be sending letters like that except I have no butler

[brisk sedge]

Fibber

[tight island]

But when I send it it’s to someone I hate

[rich surge]

Oh my...

plays with thumbs

[livid pier]

I think there must be one of us who didn't understand the other's meaning..

Perhaps you were just joking, and I didn't understand😆

[rich surge]

Yeah that's you

Perhaps

[livid pier]

Alright, after all, I am relying on a browser plugin to translate messages in real-time, and letting GPT-4-1106 translate the content in my input box. I don't understand English very well.

😆

[rich surge]

Ohhh, that's beri smort

I would have never guessed

The English is beri guud

[tight island]

[livid pier]

I previously used Google Translate and GPT-3.5-Turbo as translators, but it's quite obvious that they are far inferior to GPT-4.

[rich surge]

Yes, inbeeb

[tight island]

Agreed

[livid pier]

Perhaps the reason I didn't understand is that the translation of other people's messages was not done with GPT-4, but with Google Translate, and the translation does not include context

[rich surge]

Perhaps you could have hossenpeppered your way with it.

[livid pier]

Perhaps the plugin developer should develop a feature to explain the text, and then let GPT-4 explain this conversation after pressing the shortcut key

[rich surge]

It's merely an act of gandomoly.

[livid pier]

Anyway, I am currently relying on comparing Google Translate, GPT-4, and the original meaning in my mind to determine if the translation is correct. When I don't understand, I will throw the context into GPT-4 in ChatGPT and let it explain it to me.

[rich surge]

Yes, that is perhaps the duganshtoven way to brobekanas it with precise popularity.

I would recommend it.

[livid pier]

Alright, it seems that both GPT-4 and Google do not understand the meanings of gandomoly, duganshtoven, brobekanas..

[minor zenith]

Hola! I might be late to ask this now, but is there any ChatGPT app Or just website? 🤓

[rich surge]

Website.

And app

[livid pier]

You can go to the OpenAI website to visit/download what you need

[minor zenith]

Also, can someone share more light on what is Dall-E and Gemini? Are they siblings of chat gpt?😗

[rich surge]

There's been a few bugs and errors happening lately with ChatGPT. Have patience and they'll fix it soon.

[minor zenith]

Ohhcool

[livid pier]

Gemini (even the Ultra) is just a gimmick used by Google to promote itself, it's really bad..

[minor zenith]

Wasn't aware of app, thankss

[rich surge]

My pleasure. I love Spongebob

[livid pier]

DALL E is another AI developed by OpenAI, you can search to find out more

[minor zenith]

Damn. I thought it's developed by open ai?

[minor zenith]

Wow. So it's also free like GPT I assume

[rich surge]

I love SpongeBob's ending credit song. That song is like the theme music of my life.

[livid pier]

Google is now a complete and utter loser, hahaha! Not only has the promotion failed, but people don't even know who the developer of Gemini is

[livid pier]

This model has some free trial opportunities at Bing, but the performance of DALL E 3 there is not very good, after all, Microsoft wants to save on expenses lol

[livid pier]

Generally speaking, if you want to use it, you need to subscribe to ChatGPT Plus and then have GPT-4 call it to generate the image you want, or you can also try the OpenAI API, although that's a bit more complicated!

[elder crow]

Excuse me, I have made several Academic GPTs. Please help me to have a look, thank you. 😄https://discord.com/channels/974519864045756446/1205753624781652008

[livid pier]

I sent you a star (even though this isn't GitHub lol)

[elder crow]

Thank you, thank you.I hope there are rainbows everywhere in your life.🌈🌈🌈

[livid pier]

[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.

[rugged igloo]

any guides on i need help real quick with teams

i know its late lol sorry for asking at this hour\

[livid pier]

https://help.openai.com/en/collections/7835004-chatgpt-team

It should be noted, be cautious with the "Merge Account Feature"!!!

[rugged igloo]

yea i see that thats sussy AF

what about seperate do i still have the message cap increase still ?

[livid pier]

If you value your personal chat history and are unsure how long this team will last, please do not merge your accounts!

[livid pier]

100 GPT-4 messages/3 hours

If you don't really need some of the features of ChatGPT, I would recommend you use the OpenAI API instead

[rugged igloo]

i just want the higher usage cap for my turn based games lol

[livid pier]

The price might be a bit high lol

Why not try injecting prompt words to get the game's prompt, and then continue playing with the OpenAI API?

(Although this is somewhat unethical!)

😆

[rugged igloo]

true its fine 60 bucks isnt that much when i spend close to 400$ sometimes on the api assistants

[livid pier]

Awesome, anyway I personally can't use that much money lol

[rugged igloo]

neither can i hence why i am doing teams now

[livid pier]

However, I'll still inform you first that OpenAI seems to have a lower number for the usage cap on messages during "peak times" recently

Some Plus users have reported that they were asked to pause after only using a dozen or so messages

They are not a very reliable team!

😕

[rugged igloo]

thanks for all this info ill see how it goes . i never hit cap on normal plan so ill see if i can last longer

[livid pier]

[rugged igloo]

also the other reason i upgraded is to trouble shoot in gpt preview which burns messges fast

[livid pier]

GPT has a lower usage limit

Generally speaking, I think the models in the OpenAI API perform better and adhere more to the prompts

[rugged igloo]

yea the api uses the 128k version while we have the 32k

[livid pier]

Have you tried GPT-4-0125 yet?

[rugged igloo]

yea its much better than the november version

[livid pier]

I have been using gpt-4-0125 for a while and have also read many reviews from my friends and online users. I feel that OpenAI has only addressed the issue of "laziness" in the code and completely overlooked other aspects. Both I and many others have noticed that 0125 has a worse understanding of prompts and the "alignment" effect is sometimes excessive, to the point where the model rejects many normal questions

This is my previous review

Do you have a similar feeling?

[rugged igloo]

yea but its better than chat gpt for now but the memory feature when i get my hands on it will test it out and see if it isnt a gimmik

one thing i dont like is i cant open a new window with my personal chat so i can copy and paste what i have in there and port it over manually without it switching back sure i can make a document ect ect but still anoyying

[livid pier]

This function of memory has actually been implemented by some projects before

[rugged igloo]

yea mem gpt has it

[livid pier]

The most annoying thing is that your Custom instructions don't work on any GPT!

[rugged igloo]

ehh its a hit or miss i had to re type my custom instructions 19 times for my turn based fantasy action adventure game for it to work and it still doesnt work 100% because of 0 dialogues that happens with the characters

[livid pier]

Currently, the vast majority of GPTs do not respond exactly as the creators imagine; they usually do not have a fixed output format unless the creator forcefully and completely specifies how to do so in the system prompts

[livid pier]

This is particularly evident in various GPT games

[rugged igloo]

ya i found tht out too however i have a really stellar turn base file that the gpt always follows 100% with 0 errors if you wanna see it

[livid pier]

Awesome! I want to take a look!

[rugged igloo]

rip i cant copy it in here

[livid pier]

Send a private message then😆

[deep oracle]

Does the memory function mean it retains work across threads? I thought it was doing a decent, most of the time, retention of info within single threads already

[toxic gulch]

refresh the page?

then try a different browser

[frosty frost]

Will the new dark/light themes be addressed? I’m going blind using it for more than 10 mins, it’s insane

[dense halo]

haven't noticed any difference

¯_(ツ)_/¯

[shadow osprey]

Ask chatGPT for help with what IEnumerable does, I'm sorry, but I can't help with that request....

🤕

[velvet jackal]

/video

/video

/ video

[native spear]

Hey regarding the memory feature, is anyone aware if it's getting rolled out for a few Europeans as well?

I assume it's not due to regulations, similarly to the custom instructions that took extra wait time because of that. But yea

[normal quarry]

I have it in Australia, not sure about Europe

[ivory warren]

This might not exactly be the place to put this but here it is. I would like to propose that the LLM with all its understanding of images and context be able to say overlay an image with different things and not simply generate a new image. Let's say I upload an image of a map with roads. Let's say it knows what roads are. Now let's say I simply ask it to highlight the road named "joint lane" then it takes the image I uploaded and returns the same image but with the highlighted road. This seems like a much more useful ability rather than generating a whole new image that might be similar to the image that was uploaded etc

If anyone has any thoughts on this please share them

[hard dew]

That indeed sounds quite interesting, I myself thought about that back when I was messing around with Dall-E 3 a little bit ago, I realized that making iterations, changes, and actually being happy with the output would be far easier if it just... TOOK, what it had already generated, and just made the given changes lol.

Instead of literally throwing it all out, and making it all over again from scratch, which 99% of the time isn't even close to what the one before that was.

[ivory island]

That is indeed more useful and I am sure that feature will be available in the next 2 years.

[wide pike]

api is down

I keep getting errorrs

[idle root]

yep

[wide pike]

sigh

I was in a cliff hanger

[lyric depot]

This is literally inpaint mode

[hexed condor]

the beast is down again

[tropic sparrow]

Oh it’s not just me lol

Guess I’ll wait then

[wet shell]

womp-womp 😦

[river hearth]

oh then it's not working globally ? xD

[wide pike]

It's me to

I'm from SG< and it's giving me errors

[tropic sparrow]

I’ll just wait for people to say it’s back up

[wide pike]

I'm going to do my character stuff.

[tropic sparrow]

Yeah it’s just giving me nonstop errors atm

It happens annoying but it’ll be up again

[tawny steppe]

It is down again

[wide pike]

You don't need to say thtat

I think all of us already know it's down

[noble finch]

Gpt 3.5 too?

[somber robin]

yeah

[dark canopy]

mhm

[daring summit]

Whats going on with ChatGpt ? its getting dumber every day. "Error in moderation", "Failing to generate response", "If it should use a plugin to search internet it doenst even work anymore... ". SO slow and so many errors its frustrating

[hard dew]

I don't think the first two errors are as a result of the bot getting more dumb, that's just trash infrastructure, the 3rd one is more fitting though.

[daring summit]

yeah sorry i combined then together 😄 But still, if there are no errors its giving me the answers that are just on level with google search, but for 20$ a month 😄

[hasty sedge]

down?

[hard dew]

Yeah that's pretty cringe sadly

[tacit vortex]

useless ai, upgrade to team waste money, hurt friendship

[zinc condor]

So its not just me.

[true mirage]

Damn, again outage

[tacit vortex]

#metoo

[tropic sparrow]

Yeah it’s just down period I wish they’d say it lol

[daring summit]

and the quote of the day from the ChatGPT "I appreciate your frustration"

i am happy that he appreciate how sad i am 😄 didnt expect to get roasted by AI today

[hard dew]

It's alright, in about 2 hours you'll get your email spammed with "Service status" emails, which most likely won't even be for ChatGPT lmao, but for the API or some bs.

[tacit vortex]

pls support chinese chips and AI, they are here to help.

[lethal narwhal]

how long has it been down on gpt?

[hard dew]

Like the major issue they had yesterday right after the whole "memory" announcement, entire website and bot broke, and of course I got said emails cause I am signed up for them, and imagine my surprise when I saw that they were exclusively for API, and nothing about ChatGPT, what lol?

[nocturne elm]

i think ppl are using gpt for their valentine's plan, thats why its down 🙄

[cunning horizon]

whole thing is broken RN??

[red otter]

guys help

[hard dew]

That would be sad as frick.

[dark canopy]

can't they be original? damn

[hard dew]

But not surprising

[lethal narwhal]

man i just subscirbed to plus version and this what happens

[dense halo]

ChatGPT is broken now?

[pale adder]

is chat gpt down? unable to use it at all !

[red otter]

is it me or AI is broken rn?

[hard dew]

Holy crap bro do people just not read literally ten pixels above??

[pale adder]

lol

[red otter]

lol i just did

[nocturne elm]

they like to ask questions for no reason

[hard dew]

Before doing an ounce of due diligence you mean.

[zinc condor]

There are obviously issues right now. They should post a status update at this point.

[winged obsidian]

is chat gpt slowing down?

[karmic onyx]

yes

[granite pawn]

omg

its down ggs man

[winged obsidian]

cool

[hard dew]

I also said above:

It's alright, in about 2 hours you'll get your email spammed with "Service status" emails, which most likely won't even be for ChatGPT lmao, but for the API or some bs.

[granite pawn]

gg guys its down wowowowow

[zinc condor]

Ha.

[haughty hollow]

ok fridge of yall

[zinc condor]

Its a sad truth but we have become dependent on this tool where it messes with our entire workflow or whatever we are doing.

We are at their mercy at this point.

[dark canopy]

atleast the AI overlords will spare us when the time comes... or something

[hard dew]

The biggest mistake you could do is become dependent on ChatGPT lol, or anything to "do your job for you"

[dense halo]

Hey I was just generating some images, I'm not full-blown dependent (yet) XD

[abstract basin]

my paycheck is about to be coming in moderation if this slave doesn't get back to work

[hard dew]

🤣

[dark canopy]

bro💀

[zinc condor]

Let me rephrase lol. Become accustomed to using it as a tool. I would never use code from this in production.

[hard dew]

Yeah then that's more understandable

Praying for actual competition soon

[zinc condor]

That there. I would welcome it with open arms.

[true mirage]

Elevated error rate impacting ChatGPT
Investigating - We are currently investigating this issue.

[opal wharf]

They posted status update. “Elevated error rate impacting ChatGPT logins”

[hard dew]

"Logins" 🗿

What did I JUST say "It's alright, in about 2 hours you'll get your email spammed with "Service status" emails, which most likely won't even be for ChatGPT lmao, but for the API or some bs."

[pseudo rune]

🥵

[brisk knot]

and here i was trying to bounce ideas off of it then it just goes quiet rip

[hard dew]

Granted it's not QUITE there, but close enough

[opal wharf]

I mean, it may encapsulate logins, but it shouldve been “Elevated Errors across API and login platforms”

[pseudo rune]

UwU

Error in Moderation

[hard dew]

Or, elevated errors period, or across various systems.

[dense halo]

there are two issues currently, one is with logins and the other is more general

[low adder]

I get the same thing

The webpage is not even loading

[loud stump]

is something wrong with chatgpt today? i get "error in moderation"

[low adder]

Is the whole ChatGPT down?

[dense halo]

https://status.openai.com for details

[grizzled cedar]

good, then isn't problem with my prompts 😂

[dark canopy]

finally

[winged obsidian]

not sure

[hard dew]

zzzzzzz

[low adder]

That is funny because I have plus and it could at least work if they want money

[abstract basin]

did u unplug it and plug it back in

[low adder]

I get the same thing

[loud stump]

inside your mom yeah

[hard dew]

That is so cringe

[pseudo rune]

uh....

[dense halo]

Slam... please stop that

[hard dew]

Anyway, back to microsoft flight sim 2020 while ClosedAI remember their hardware knowledge again, and software

[loud stump]

im looking for legit answers though. he started it

[winged obsidian]

as u can see, yes

[opal wharf]

That’s not constructive

[haughty hollow]

bro is mad he couldnt get the answer of math exam

[zinc condor]

I mean at this point all anybody can do is wait.

[hard dew]

It's understandable and I can relate sometimes, but at the end of the day all he did only, was try making a light-hearted joke out of a situation annoying for everyone, not just you

[haughty hollow]

Time to use Gemini Pro

[dense halo]

https://status.openai.com shows 2 issues, one with chatgpt, gpt4 vision and the API and other is with logins and a new error covering fine tuning and assistants

now 3 issues 😄

[loud stump]

thank you

[true mirage]

Lmao another outages

services (including Assistants and Fine-tuning)
Investigating - We are currently investigating this issue.

Elevated error rate impacting ChatGPT (including logins)
Investigating - We are currently investigating this issue

[hard dew]

Place your bets on how high it gets before they "fix it"

[dense halo]

you're welcome

investigating!

[hard dew]

🔍 🐒

🙀

[haughty hollow]

Are APIs included too ?

[wet fable]

They cant fix it. GPT is down. So they cant ask how to fix the problems with GPT. 😄

[serene horizon]

For the Teams subscription... is that just $25 per month if it is just... me? lol

[opal wharf]

Very high pressure situation for the team

[serene horizon]

Oh wait. $50. Minimum two users

But it has higher context and message caps, right?

[haughty hollow]

Gemini Pro is good tho lel

[hard dew]

Maybe, depends on the weather, unironically lol

[serene horizon]

How's that? lol

[true mirage]

Now 3 active outages

[lethal narwhal]

has it always been like this? or is it just on rare occasions?

[serene horizon]

Gemini comprable to GPT for coding?

[opal wharf]

GPT cannot be fully fixed in the sense that it is completely stable. Server overload, dynamic malfunctions are all unpredictable

[hard dew]

Basically, you should completely ignore the given "message cap" most times, as it's never close to it even lol, it fluctuates completely randomly whenever it wants wherever it wants

[foggy grove]

oh okay so chatgpt is down, its not just me ?

[hard dew]

Sometimes you'll enjoy the 3 hour error after 10 messages, sometimes after 30, etc.

[dense halo]

likely a common source point for the issue, once it's traced down it should be resolved quickly

[haughty hollow]

for that case use Copilot if talking abt 3.5

[serene horizon]

No talking about 4

Is copilot STILL not on 4?!

[digital silo]

you may get partial outage

[native socket]

i've been using both, gpt4 is way better than gemini, less errors and the logic is better

[hard dew]

Wanna know something comparable to coding especially that you should use? Yourself and your own brain lmao, also speaking from experience.

[haughty hollow]

ye they same but Copilot got more resources to use

[digital silo]

chekc for the status https://status.openai.com/

[haughty hollow]

as ik, i never tested 4

[amber crypt]

what is this?

"Error in Moderation"

[hard dew]

Not on Visual studio, but visual studio code it is (If you use the pre-release versions), it still sucks so bad.

[abstract basin]

smarterchild is still up FYI

[haughty hollow]

ama test if APIs still working

[hard dew]

So why are you so concerned on using AI for coding when it sucks for it?

[frail shore]

Now OpenAI devs cannot ask ChatGPT how to fix the problem

[hard dew]

Most truthful statement ever made

[haughty hollow]

seems APIs working fine !

so you can use API to use GPT 3.5 or 4

[serene horizon]

I don't undersand what you are even asking. I didn't say 'write all my code for me'.

[hard dew]

I didn't say that either did I? I said that it sucks for coding, period, no matter how much or how little you use it.

[serene horizon]

It really doesn't lol

[haughty hollow]

nvm APIs not rlly responding

[hard dew]

☠️

[serene horizon]

It's very helpful

[hard dew]

Sure

[haughty hollow]

or wait it does

[lyric depot]

It is

[serene horizon]

What are you here for if it sucks so bad lol

[hard dew]

What?

[eternal rock]

`` Server Tip: ``

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

[sleek aspen]

Both ChatGPT and the API are having some issues at the moment - refer to https://status.openai.com/

[serene horizon]

We sadly are not all as blessed with the brains of @hard dew

[lyric depot]

Especially for shorter code

[quiet dirge]

chatgpt has become nigh unusable in the mornings for me over the past couple of weeks.

[zinc condor]

Wow...

[quiet dirge]

I'm starting to regret having to pay for for pro

[hard dew]

What exactly are you so heartbroken over for lmao, if you want to have a problem with me, my dms are open, get in there. Also saying that I should leave just cause I say that ChatGPT sucks for like one thing out of many others you can use it for is so idiotic

[sleek aspen]

Coding without AI isn't helpful if you aren't entirely aware of what the code does, so for most beginners AI tools will do anything but help
This may be a discussion better suited for #ai-discussions by the way!

[zinc condor]

Totally agree.

[hard dew]

This is also very true and what I've seen a ton

[zinc condor]

AI should be ommited for teach people how to learn software development and programming imo.

teaching

[lyric depot]

Why?

[zinc valley]

i feel like chat gpt has been running pretty trash as of late honestly, what are some good alternatives

[supple terrace]

@hard dew is a casual chatGPT discord enjoyer

[sleek aspen]

You ask look into and ask about other AI services in #ai-discussions

[supple terrace]

man got that post history going

[frail shore]

i use devdocs for coding, and then ask chatgpt to find bugs and improvments

[zinc condor]

Honestly, everything starts with a solid foundation, That starts away from using it

[hard dew]

485 messages when I joined all the way back in August, be for real lmao

[wet shell]

Sometimes it saves time, and sometimes it doesn't; it shines more for small procedural scripts and rote tasks like "take this example class from $framework and make a new class based on it that does this instead" -- but for me the real use is for things that are outside my wheelhouse, like generating business documents and things like that

[haughty hollow]

ok seems the APIs works fine

[gilded gazelle]

data analysis is not working or is it just me ?

[haughty hollow]

means you can use alternatives like FlowGPT

which uses GPT API lel

[supple terrace]

I was just reading what kind of person you are from your history 👀

[zinc condor]

We should move this to AI discussions so we dont clog this one up.

[viscid valley]

The login error, huh?

[hard dew]

That's awesome, but why though?

[opal wharf]

ChatGPT has been pretty average, but “trash” undermines its complexity and skills. ChatGPT still is knowledgeable. While in recent times, it may be reduced in its ability, it’s not entirely lacking

[lyric depot]

This happened to me twice already, when the chat gets too long, GPT is refusing to help me with code ("I can't assist with that"), and starts replying with more general advice rather than writing actual code. Is it happening to anybody else? It's like this even if I feed it the context again.

[zinc valley]

i want to find a chatbot thats not so restricted

[lyric depot]

It works if I use another chat.

[zinc condor]

Defo has happened to me as well.

[hard dew]

There are none unfortunately afaik, hence why I keep praying for competition

[opal wharf]

For anything not ChatGPT related or discussing third parties, #ai-discussions

[viscid valley]

Cannot login, anyone?

[eternal rock]

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

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

[haughty hollow]

it uses ChatGPT API so how this is not related

[opal wharf]

It’s not related in the sense that it’s not the platform, ChatGPT, and it’s a third party outside of ChatGPT. While it may use the ChatGPT API, it is inherently a third party

[wicked oasis]

One message removed from a suspended account.

[granite pendant]

Is it only with login? I'm getting "Error in Moderation"

[sleek aspen]

Refer to https://status.openai.com/

[wispy iron]

time for a break from work yaaay

[wicked oasis]

One message removed from a suspended account.

[opal wharf]

It is a widely recognised issue and is ongoing. It will be resolved eventually

[spice shell]

chatgpt is very slow

[left hill]

im getting a "error in moderation"

[spice shell]

Lagging badly

[opal wharf]

It’s “very slow” due to ongoing maintenance in the API. API is overloaded

[sleek aspen]

There are currently elevated error rates affecting ChatGPT - refer to https://status.openai.com/incidents/gln4g8d46s2b

[zinc valley]

arent there some that are comlete anarchy?

[spice shell]

i guess now is the time to reflect on how reliant i have become on chatGPT

[granite pendant]

How much time does it usually takes to fix the issue?

[hard dew]

That is an incredibly beneficial and intelligent thing you can do for yourself indeed, can confirm, I had to do the same only a bit after GPT-4 came out and I had used it a ton lmao.

[opal wharf]

ChatGPT is being ingrained into everyday life. Like smartphones or computers. This is generational and will take a few decades to sink in majority

[grizzled cedar]

I think we are closer to Sam Altman getting his 7 trillion dollars than seeing a competitor as good as GPT4

[sharp dragon]

Anyone else having connectivity issues with ChatGPT again? Prompts don't work. "Error in moderation", "Network issues", etc

[lament creek]

me too

[sharp dragon]

Tried signing out and back in: "Oops! We ran into an issue while signing you in, please take a break and try again soon."

[sharp dragon]

Dang it

[timid brook]

known issue, see above

[opal wharf]

Every chatbot platform strives on competition. However, they also have to balance that with consumer demand and market trends. The current market trends is significant impacted by ChatGPT. Currently, a regulated chatbot is the ongoing trend

[sharp dragon]

Pinned?

Where

[grizzled cedar]

mixtral is very good, but I think it is closer to 3.5, and as you said, we need a 'bit' of vram do use it 😂

[timid brook]

just scroll up a little bit, its been discussed for the last 20 minutes in most messages

[sleek aspen]

https://status.openai.com/incidents/gln4g8d46s2b

[thin field]

Yea it's down i think

[sharp dragon]

Ah great thanks guys

[timid brook]

you can also check https://status.openai.com/

[lament creek]

that's the reason i join server. To ask about this issue

[sharp dragon]

Perfect

[fiery mango]

Damn. why did it have to be down in the most inconvenient time 😅

[grizzled cedar]

how much vram you have?

[final venture]

@ancient sinew that's very interesting. Anything specific that stands out? Why is the code better (and/or less bad) , according to you? Or just a general feeling?

[lament creek]

that mean i have to wait OpenAI to fix this issue right?

[lyric depot]

yes

[grizzled cedar]

I installed chat with rtx yesterday, it has the mixtral 7b

[lament creek]

tks

[timid brook]

gguf / llama.cpp runs petty well even with low-ish vram

[maiden stirrup]

What GPU are you running?

[grizzled cedar]

3060 12gb

[granite pawn]

im running a 4070

no chatgpts is working rn i even got plus

[maiden stirrup]

Considering the 4070 Ti Super ngl

[granite pawn]

yea im running 4070 ti

[maiden stirrup]

I need to upgrade from an RX 580

[granite pawn]

openai plsplspls fix bro i need it for studying and for work pls openai fix chatgpt and fix plus upgrade servers plz

[timid brook]

48gb? wat

[grizzled cedar]

which mixtral you use?

[rapid falcon]

anyone have login error

[lavish minnow]

Moderation error?

[opal wharf]

https://status.openai.com/incidents/gln4g8d46s2b

[spice falcon]

PROBLEMS WITH CHAT GPT

[rapid falcon]

yup

[granite pawn]

bro even prompts dont work for API, 3.5, 4.0 and all gpts

[lavish minnow]

is it down

?

[opal wharf]

Yes. API overload

[verbal escarp]

Annoying, trying to work on my project here....

[lavish minnow]

thanks

[grizzled cedar]

nice

[granite pawn]

there is a login error, there is a moderation error and there is a HMM SOMETHING SEEMS TO BE WRONG error for all and any prompt u put in any GPT rn

[rapid falcon]

dammit

[mental rose]

is chat gpt down

[timid brook]

yes its down, read the chat

[granite pawn]

yes even plus and APIs are all down lmao pls learn 2 read

[timid brook]

https://status.openai.com/

[rapid falcon]

thats why my storys broke

[gilded gazelle]

assuming chat ever gives a proper answer

[grizzled cedar]

god, people really need chatgpt, even to breath, can't even read half dozen lines to see our chat

[verbal escarp]

Does anyone know what the limit is when you´re deploying your own api with the gpt 4 model, I am thinking of the limits that you reach when using the gpt 4 normally. Does that also apply to the api?

[foggy grove]

jesus everything died 😂

[timid brook]

its been discussed for the last 20 minutes that its down, all you need to do is scroll up and read

[lavish minnow]

thanks paccer ur right

[hard dew]

CAREFUL, lest you disturb the hornet's nest of clowns when you criticize or express your own experiences with it even slightly.

[timid brook]

neat, i need to get my hands on one of those 🤤 im on 2080 right now lol

[granite pawn]

bro how is this a "partial outage" LMAO EVERYTHING IS COOKED RN EVEN API AND PLUS, is it termed that way because of the duration? not the scale?

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 1 `` Be respectful.

Treat others the way you would like to be treated, and assume best intentions. Don’t harass or attack others, and don’t engage in hateful or generally malicious behavior (e.g. sexism, racism, homophobia, etc.). Keep the negativity to a minimum.

[flint bolt]

Yeah, was logged out and now I can't log in again... 😦

[grizzled cedar]

why I cant react to the msgs? lol

[opal wharf]

To prevent misuse of the reactions

[granite pawn]

https://status.openai.com/

[timid brook]

yeah i think there was a sweet spot between bitcoin mining dying and ai not being big yet, sounds like you made the right choice

[grizzled cedar]

meh

[wide pike]

I keep getting network error