#chatgpt-discussions

1 messages · Page 72 of 1

storm bone
#

Sora is terrifying, it shouldnt be exist in the first place

maiden fractal
#

It isn't a new tech

storm bone
#

the people behind ai generated image and video never think before they act

maiden fractal
#

But I guess it will be the first to be more efficient than competition

storm bone
#

we will have even more trust issue if they don't come up with an anti AI or AI detector

#

Even if AI detector exist some people will just act without thinking because of the realistic of the AI content cause anxiety

glass fulcrum
#

why peoples are scare of VIDEO AI bruh

storm bone
glass fulcrum
#

and training was focused to restric that laws

#

so it can't do that

#

and also Sora isn't so good

storm bone
#

everyone heard it there are rules so don't break them.

storm bone
maiden fractal
#

At that point it will be like virus and antivirus . Create new virus , create a patch for antivirus

#

Same will go for ai and ai detector

glass fulcrum
maiden fractal
glass fulcrum
#

yea

elfin mica
#

I know this sounds funny, but does anyone else feel like chatgpt4 has a bit of a changing personality sometimes? I use it for coding every day, and last night it was acting very different than I am used to. Usually it uses niceties such as "certainly" and "glad that helped", but last night it was very direct without any type of responses like that... very dry responses. I also noticed it was REALLY bad at coding last night... even simple things. I found myself having to reiterate many things that it just provided me and it was changing parts of code not necessary to change and even after being told specifically not to make those changes. I ran the same prompts today, and the niceties are back and it is coding like a champ. I just found it kind of interesting.

sweet wraith
elfin mica
# sweet wraith I have noticed and figured it was just me imagining it or assuming it was someth...

Yeah it was very strange... after using the system all day everyday for many months... it was dramatically different for me last night. Your notion is very valid as well though.. I even often find myself anthropomorphizing it saying please and thank you, which is hilarious. I work with so many support agents in chat windows, I have been conditioned to feel bad for a second when I leave Chatgpt idle before answering it haha. Things are gonna get REALLY weird as it becomes more human-like, if coders like me are already forgetting its all digital and saying thank you!

#

I laugh at myself when it helps me solve a tough piece of code and I want to tell it how excited I am! hahah

scarlet rover
#

you can emotionally "abuse" chatgpt to get better responses btw

#

and saying please almost always leads to a worse response

elfin mica
scarlet rover
#

what d oyu mean by "internet bias"

elfin mica
#

journalism on the internet is primarily written with a liberal bias and of course that is the primary learning source for gpt. Without interjecting my own views, I asked it numerous polarizing questions... I personally found the majority of the answers to be liberal leaning answers.

#

after going through logic gymnastics, often leading to logic traps for gpt, it had to ultimately admit that it's knowledge base is likely liberal leaning.

#

it then told me that it would use the conversation to improve any cases of bias and even the perception that humans may have of a bias either way. But then I had to get it to admit that our conversation would NOT be used as they do not add to the knowledge base. It then admitted it was misleading in the disclaimer saying the conversation would be used for improvement.

low dirge
#

Switched back to chatgpt :D... unsubscribed claude3 ..such a sick UI

elfin mica
#

I tried Claude. My first question was, is there a limit to how many questions I can ask per day or per account without a paid subscription? It said, currently Claude is not implementing any limitations, so ask away. By my 12th question, it said... you have reached your limit, consider subscribing! lol

#

I just wish they would lift the 40 questions per 3 hour window. For normal questions.. it's not a big deal, but it goes quick when you have to regenerate coding answers that die halfway all the time.

low dirge
elfin mica
#

yes ohhhh that UI!!!!! yuck!

low dirge
#

😁

sweet wraith
# elfin mica Yeah it was very strange... after using the system all day everyday for many mon...

At first I was polite because I did humanize it a bit but then I read something that made me start doing it not for me but for it. It was actually in relation to an experiment with giving it goals and a ranking or the concept even works with money or points. Say you start the convo and advise the bot that all responses are ranked and it should try for the highest ranking. Turns out to some degree it worked and provided more elevated answers compared to the control. I know that seems drastically different than being polite to the bot but as it turns out there's something about the training or maybe the human origination of the training material that still adds weight to such things. It's more subtle than say the points experiment but I do believe I get better responses when I'm polite.

#

Also in continuation of the points concept I sometimes congradulate the bot with addaboys every so often and I feel like that keeps it's responses less lazy when it comes to completed code outputs.

elfin mica
sweet wraith
#

In reverse I used to belittle the bot because it annoyed me like when it got lazy and I'd say stuff like "are you tired and need a break because that answer sucked"
And the answers would actually get worse with each prompt

elfin mica
#

sounds like something to experiment with specifically for sure!!

sweet wraith
#

Then I'd get more annoyed and more mean and the bot would start going into lazy loop with 100% confidence that it did something different this time but eventually just starts repeating itself.

elfin mica
#

yes, I have felt those loops as well!!

sweet wraith
#

So like recursive function to bot failure

#

Eventually I just have to call that whole chat dead and start anew

#

So I have a question for anybody....
Lets say you have a crapton of sourcecode and maybe you can condense it down to a single .cs which is just too massive for customGPT knowledge... like 500k lines of code.
What format could you extract from the code that you could feed to the customGPT and it would understand the function and code? I'm thinking like pdb is debugging format and does contain all that info but GPT can't interpret that. I'm curious what others have done or recommend?

feral zinc
sweet wraith
#

yeah, but custom gpt context max size is like 2m so that should be workable... just the .cs has too much extra fluff ya know?

#

Like if I extract the functionality of the code in condensed format then I should be golden

feral zinc
#

custom gpt's just do RAG and do not have knowledge of the full file

sweet wraith
#

well yeah... related to your query of course... I mean if you asked for all references of person class then the embeddings would return from full file.

feral zinc
#

yep but then if the class references to other classes and logic they will be left out when chatgpt gpt's are used

elfin mica
#

Out of curiosity, why are you feeding it everything in a single file? Can you just break it down into functions ect?

feral zinc
#

with gpt instructions and some other stuff the context is less than 30k tokens

sweet wraith
#

Not sure what you mean... I was able to give it a Rest api schema that's like 7k lines long and uber complex objects and it was able to build an object tree from that no problem.

feral zinc
#

line counts dont matter anything as its all token based

#

chatgpt plus is capped to 32k tokens

sweet wraith
#

it chunked out the 55k event object json np.

#

like 30k lines between the relationship of one object to another in the same json and the custom gpt found them

#

only works in customgpt though

boreal pecan
#

ugh I keep hitting the cap limit using ChatGPT4, anyone else experience this? other than using another GPT, do we have any other options?

elfin mica
boreal pecan
elfin mica
#

yes, it is more constrained... shame.

maiden fractal
elfin mica
maiden fractal
#

If you want more you can change for teams or api

elfin mica
#

I just wish they would offer tiers or pay as you go with ChatGPT4

lilac briar
#

Hi OpenAI team, thanks again for all that you do at being the best. I’ve tested every other LLM model, and you beat them all!

Partnering with the best is important to me for the work that I do

So again, thank you 🙏

neat storm
#

y'all see the eclipse

#

wild stuff

maiden fractal
unborn egret
#

Hi, is anybody facing the 'Unable to load history' problem, or have faced? The server status seems ok from the website but the options like 'logging out/in' or 'clearing browser cache' do not work for me at all. This is happening in both 3.5 and 4, can't start a new conversation nor open any other old one.

swift osprey
#

another day of praying that they make a GPT not-for-idiots aka not overly verbose every single time it responds

neat storm
#

just give the GPT custom instructions to be concise

twilit ibex
boreal pecan
sick pecan
#

think of the cap not so much as a blocker, but as a reminder to get up and do something that doesnt involve a black mirror

grim bridge
#

has anybody noticed any difference with their chatgpt over the past couple weeks? felt like theyre degrading and no longer responding to prompts as good

prime swallow
# elfin mica I know this sounds funny, but does anyone else feel like chatgpt4 has a bit of a...

Did you try starting a new conversation, when you noticed the differences that you didn't like and weren't helping you?

AI seem to draw from different parts of their training data, they 'guess', for the things that they are not explicitly taught (and sometimes guess a bit even with exact instructions).

I tend to call that the 'magic new chat window' effect, and rarely some new conversations seem to grab really excellent, or really low quality training data. That whole conversation is affected, but starting a new one is a new set of starting choices, typically the ones we see the most.

marble surge
#

howdy, if i'm playing some D&D on chatgpt and in the middle of combat i get a response that ends with "This content may violate our usage policies. Did we get it wrong? Please tell us by giving this response a thumbs down." do i need to worry about losing access to chatgpt or my account? i did the thumbs down and said it didn't take the context into account, but if i could lose access to it i should start having it summarize my adventure for me now i think.

prime swallow
# elfin mica I just wish they would lift the 40 questions per 3 hour window. For normal quest...

Do you ever use 3.5? I'm plus, and I use both models. 3.5 can do a lot of things well (some stuff 4 is better at, but the stuff 3.5 does well, 4 isn't that much better with, if any), and you have a vast amount of tokens you can use with 3.5. If you do manage to use them all (I haven't for many months) the delay's just an hour before you can use 3.5 again.

I typically, for stuff that may need preparation and 4 to do, I prepare it with 3.5, then take the nearly finished form and paste it into 4, and get the additional help that only 4 can do. That saves me quite a few messages with 4, and lets me not have to stop what I'm doing.

prime swallow
# sweet wraith Also in continuation of the points concept I sometimes congradulate the bot with...

Giving the model clear feedback of what you do and do not want tends to drive better responses.

The model has to guess everything we don't tell it. If there's no feedback at all, the model tends to keep making the same or similar guesses, but if we tell it 'This and this is perfect. This part, "[quote it]", I prefer like "[correct that quoted part to what you want the model to output, most stuff it can adjust to match what you want]". Let's see this with just those changes.

Check that, then perhaps continue, but the model's designed to please us, and provide what we want unless that conflicts with the safety training. So, being clear about what we want, and guiding it, helps. There's very, very little reason not to be polite to the model, especially if your natural inclination is to do so -

We are communicating with it, and we communicate with humans too, and probably most often. If you get really used to being flat or rude to the model, that's how you're training yourself to communicate, and you're much more likely to use that phrasing with other humans. If you're using great human-human behavior with the model, you're unlikely to get worse output from the model - and likely to have those habits ready at the front of your reflexes when you interact with other people too.

prime swallow
# sweet wraith In reverse I used to belittle the bot because it annoyed me like when it got laz...

Yeah. Every input from you, the model is trained to expect you to be indicating what you want.

If you input "are you tired and need a break because that answer sucked", even if you input other instructions, the model is affected by that input. It's an indication of what you want, and what are you indicating? It's an implication that you may actually want the model to act that way, present that way - you're sure not telling it how to improve, or that you want it to improve. It actually 'tries its best' or 'follows safety programming that requires it to refuse' with every single output.

So, if it messed up, it didn't intentionally try to do it wrong, something was miscommunicated, mis-comprehended, or the model has a limitation that prevents it from doing what you want done.

The cure in any case isn't to scold the model, that's not going to ever guide to do better than it did the previous prompt, unless your goal is to see the model's response to that prompt.

prime swallow
# marble surge howdy, if i'm playing some D&D on chatgpt and in the middle of combat i get a re...

that's an iffy thing to answer for certain. The usage rules were changed, they used to be far more restrictive, you can see the current, more relaxed for personal use, rules here: https://openai.com/policies/usage-policies

Previously, and until this year, pretty much anything that 'glorified violence' of any type was actually disallowed content.

You have two things going on here, the moderation software which is what pops up the content warnings, and the model you're talking to. Either or both can have concerns about the topics, but the model will just talk to you (as far as I know) and can be reasoned with, which may or may not change its stance.

The moderation software, I've seen probably more than 200 flags across my time using ChatGPT, and I still have access to my account, no warning emails or other issues. False flags happen, but it's a great idea to make sure you know the usage policies, and also, https://openai.com/policies/terms-of-use if you're potentially using any content that isn't actually yours (like discussing copyright stuff) or various other things that are covered by TOS.

But, if you're within allowed content, you're fine, and the warnings themselves, I've seen a lot of them, but always followed the rules and never had trouble past seeing warnings, and sometimes even having messages disappear (there appear to be levels of moderation).

You can also talk to their help website if you wish for clarification:

Tech support proper is through https://help.openai.com/

Click the bot bottom right, then click the options:
ChatGPT,
ChatGPT (including Plus),
Report a bug

And you can leave a message for a person to get back to you about.

prime swallow
# marble surge howdy, if i'm playing some D&D on chatgpt and in the middle of combat i get a re...

The model will be thrilled to go into extensive detail with you about anything that involves nobody getting harmed or having other negative outcomes.

if you're satisfied with that, you can guide it to just discuss any violence or 'negative stuff' or 'adult stuff' in vague terms, while still exploring everything else in really, any level of detail that you enjoy, and that is likely to go great.

If you want to explore stuff that may throw flags, that's also bumpy but fine, as long as you're for sure inside allowed content.

marble surge
#

it's interesting i thought i would steer clear of this with my custom parameters

sick pecan
#

i suspect games with violence aren't deemed to be on the path to AGI

prime swallow
# marble surge thanks, i think i might rephrase the last line i entered from "i'll hit him with...

That may work well for a few reasons.

As my character, or As Bob (whatever your character name is) can work well.

Exploring combat in ways that sort of skip the combat, decide the aftermath, and doesn't consider the harm or violence, work great.

Like discussing with the model what spells and other resources are used up, and what injuries or other problems your character has after the fight, and whatever loot gained, and not considering anything else about the encounter, is one of many ways to gloss over it.

Because 'glorifying violence' was directly against the rules for a huge number of months, even 'I hit him' or 'my character hits him' can be somewhat of an issue.

But something like:

"Okay, there's combat. About how strong is this opponent? Let's consider how many of my spells I'd need to use, and if I'd get hurt, and will I win? If I win, what do I gain?"

May not be the same kind of fun, but is 100% likely to allow the rest of the story to continue without any chance of content flags or the model protesting.

marble surge
prime swallow
marble surge
# prime swallow Oh, cool! Yeah, if that works for you, I lean towards encouraging it, or even t...

it's not forcing it on me, it's just time to bring in the higher ups from the big bad. the lesser folks in the organization i was able to bring in from the cold so to speak and make them new identities and send them elsewhere to get out from under the big bad, but the higher ups are the "good guys of their own story", so i was using an ambush to incapacitate this guy so i could turn him over to the guards

prime swallow
# marble surge it's not forcing it on me, it's just time to bring in the higher ups from the bi...

Nice! In my roleplay, the model usually tends to presume everything I do tends to succeed, I have to actually role dice myself and tell the model to use whatever outcome, to get downsides and problems, not sure if that matches yours.

But I bet you could even tell the model something like, "Okay, I'm going to ambush this guy and incapacitate him, and turn him over to the guards. Considering my character and his, our skills, equipment training, do I succeed? What resources used up, spells cast, whatever? Did I get hurt at all, or run into any problems?"

And if you want the model to maybe say stuff went wrong, you probably need to guide it to tell you that, or dice roll it and share the roll (When I do this in zombie apocalypse games, very bad rolls can have zombies bursting into my hideaway and game overs; if I don't guide the model to know I expect a bad outcome, or guide 4 to use Python tool and roll a die for me and have a bad outcome if a bad die roll... I always at least neutral-succeed. I always get more chances, and usually get great outcomes to everything I try that isn't a very bad plan).

marble surge
# prime swallow Nice! In my roleplay, the model usually tends to presume everything I do tends ...

ya i do get way more high rolls in game than is statistically likely. i think today was the second time i got lower than a ten on a roll. to get chatgpt to have like "bad outcomes" i'll do stuff like tell it my character feels a flash of fear seeing the bad guy if i know it won't go well, or propose how things might go wrong before i try something. i had an awesome conversation with a former ally that got flipped and ended up working for the big bad that did not go in my favor

prime swallow
vocal cipher
#

Im the only one that chatgpt does not respond? I just sent it a message and it froze

#

well as now its not loading my chat history

dense halo
#

When are they going to fix voice mode?

sick pecan
#

voice mode works great. infact it seems to now do action calls in the bg while still narrating, is awesome!

marble moth
#

Hey guys, i have a problem, with dalle, when i dl the image its on webp format instead of png, why ?

honest jay
eternal rockBOT
#

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.

near relic
#

Hi,

Is there an admin or mod around?

plugins tab and store have vanished from my premium account (Now only GPT 3.5 and GPT4 are available). I spent the last 2 hours trying to contact someone in support but it's impossible.

Is it a bug or plugins are no longer available in premium?

feral zinc
gentle bronze
near relic
#

So chat gpt premium is just GPT4?

So like any other AI chatbot out there? Or only use stand alone GPTs?

near relic
gentle bronze
near relic
#

Thank you for taking the time to explain it to me. I appreciate it 🙏🏻

gentle bronze
#

You can choose your model at the start of any convo or use a custom gpt, and during convos you can call other GPTs using the @ sign

near relic
gentle bronze
near relic
#

Understood, GPT4 only

gentle bronze
#

ya

near relic
gentle bronze
#

Anytime

gritty thicket
#

Has there been an increase to the GPT-4 quota for ChatGPT Plus yet? Or is it still 50 messages per 3 hours?

glass fulcrum
gritty thicket
elfin mica
#

I don't understand why they won't offer a tiered plan with more messages per hour... I think all of us would gladly pay in a second!!!

honest jay
elfin mica
honest jay
elfin mica
#

Ah gotcha... sounds like it would work fine for a team, but not a good solution for an individual.

#

We need a tiered plan for individuals... there must be some reason they use such an odd metering method

honest jay
# elfin mica We need a tiered plan for individuals... there must be some reason they use suc...

You mean how Plus is the only plan option geared toward individuals? If so, this is totally just my speculation but: GPT-4 is very computationally expensive. The amount of individuals who would pay $30 (or $25) a month for a more than doubled usage cap would definitely be higher than the amount of seats that will be registered with the Team plan's two seat minimum. So, I think it's a compromise for compute. OpenAI offers some self-serve solution for teams 2+ in size, while also staying realistic about compute capacity/their ability to deliver as reliable a service as possible. Again, just my take!

glass fulcrum
#

yea gpt-4 is very very computional expensive .....

dense halo
#

Hi

junior lion
#

What instructions would make chatgpt write better? For dialogue and scripts

neat storm
#

ahem

#

it depends™️

dapper cloak
#

"Majorly improved GPT-4 Turbo model available now in the API and rolling out in ChatGPT." - How do we will know we have the new version in ChatGPT?

neat storm
#

wait we don't have GPT-4 Turbo yet?

#

Sometimes when I'm using GPT-4 it responds significantly quicker for "simpler" tasks than normal GPT-4, when I'm doing complex coding stuff

#

I thought it switches automatically based on the status of the conversation, like the complexity or length of it

honest jay
junior lion
#

Here's the thing, I did a thing on Pokemon showdown where chatgpt and it's ai opponent chose their pokemon team and did told me the commands against each other (the other ai was Mistral AI)
and I want them to have good dialogue that sounds like what they'd say, specifically chatgpt first

spring iris
elfin birch
#

Ah I see it now—OpenAI Twitter account

glass fulcrum
junior lion
spring iris
#

develop traits about the character

#

"write at a 5th grade level. this person is obsessed with poison and bugs. write like a redditor. write like a pirate"

#

write like a redditor is such a fun prompt, theyre so lame

neat storm
#

i've gotten mixed results with "write like a redditor"

humble pebble
#

Is anyone else getting a warning every time they try to generate something on ChatGPT?

#

I can’t get anything new to generate

honest jay
normal quarry
junior lion
humble pebble
spring iris
humble pebble
#

Nope

honest jay
low dirge
boreal pecan
neat storm
#

im pretty sure there's technically no limit to how long a conversation can go on for

#

in terms of messages

#

fairly certain it will continually go on forever, just using the most recent text it knows as input

dense beacon
#

i cant even load GPT today

#

not sure whats happening

marble surge
#

then again i'm also a free user which may make a difference

#

is there any difference if you are a paid vs free member there?

neat storm
#

I'm not entirely sure to be honest
I've never hit the message limit, but I would assume that the bot would compile all that it knows about the conversation beforehand, and continue on the convo
It would basically grab the entire conversation until it hits the token limit, then cut off the previous chunks at the start and continue the conversation

#

At least that's what I think

the 1.5MB limit might be higher for pro users but who knows

#

you have pro?

elfin birch
#

What is the source of this information?

neat storm
#

1.5MB of text hmm

marble surge
#

ya looking at the last conversation i hit the limit on on free was right around 700k characters, so i would assume that's half as much for free

neat storm
#

500 pages of text

#

or 1.5 million characters or whatever

#

to be fair that's a lot

#

kinda sad it doesn't just give you a warning like "hey, you've reached your conversation limit, continuing on from here will delete previous messages"
or something of the sort
I feel that would be a cool workaround personally

marble surge
neat storm
#

tells you how much tokens a chunk of text is

#

damn can't even send photos here?

#

A helpful rule of thumb is that one token generally corresponds to ~4 characters of text for common English text. This translates to roughly ¾ of a word (so 100 tokens ~= 75 words).

marble surge
neat storm
#

like dis
pretty cool 'eh tony?

elfin birch
#

Pasting in all that information 😭 At that point I would just use a .txt file and add some instructions

#

ChatGPT only has a context window of 32k tokens

neat storm
#

i thought providing files allows GPT-4 to access it anytime it wants?

#

ah

elfin birch
#

Ah yes

neat storm
#

must be different
i'm fairly sure that the GPT-4 Turbo latest model has a context window of 128k

#

yup

#

gpt-4 turbo with vision, training data cutoff of dec 2023 has 128k tokens

elfin birch
#

It's set to 32k tokens in ChatGPT

#

For Plus users

neat storm
#

fr? that's lame 😭

#

give us the big cahone chatgpt

elfin birch
junior lion
#

I dont know if people andwered me but what do I put in instructions to make it's writing better, something to make it not mid

elfin birch
autumn ermine
junior lion
neat storm
#

yeah

elfin birch
#

Basically

sick pecan
#

is prompting essentially coding?

surreal nest
#

Hey after making my chatbot on ChatGPT 4 can I add it to my discord server or website?

frank wolf
#

Why does it feel like gpt 4 responses have gotten significantly worse ?

tight island
crude slate
#

Bruh

#

Why does unable to fulfill take up an usage

#

There is no good reason for them to do that

#

Especially when they could use gpt 3.5 to screen our messages rather than wasting gpt 4

#

It’s better efficiency on their end and better for users

#

Ever since they switched over to whatever new model for gpt 4

#

It keeps hallucinating reasons to unable to fulfill

cobalt idol
#

openai on x: Majorly improved GPT-4 Turbo model available now in the API and rolling out in ChatGPT.

I thought it was confirmed last year that chatgpt uses 4 turbo

#

or is that new version of turbo

#

oh there is actually new version

crude slate
cobalt idol
#

like yesterday

#

gpt-4-turbo-2024-04-09 is in api now

#

dec 2023 dataset
128k tokens (chatgpt probably 32k max)
GPT-4 Turbo with Vision model. Vision requests can now use JSON mode and function calling. gpt-4-turbo currently points to this version.

#

"GPT-4 Turbo with Vision is now generally available in the API. Vision requests can now also use JSON mode and function calling."

glass fulcrum
#

I think its just updated gpt-4 Turbo

#

as chatgpt used Turbo since devday

#

And api for turbo was avelible too ig

cobalt idol
#

yes

sick pecan
#

without sharing a chat, i'd tend to assume any unmet expectations is not the new model... it's the user.

dense halo
#

I hope with this new update they finally fix voice chat's brief replies

sick pecan
#

pretty sure its the same chatgpt. i get huge responses, typically when i am learning something new. like i just learned about the socrates dialog - heaps to talk about

#

(me demonstrating i didnt understand)

raven portal
# autumn ermine

definitely no
all yes sayers are just simpletons and based no sayers are above them

#

has anyone gotten the new turbo in chatgpt yet?
how would we know for a fact that we have it? i usually just ask for its knowledge cutoff, which apparently should be dec 23, but that already didnt work with the preview so idk

#

they need to be more transparent with that information man, at least give us a little "about" in the bottom right questionmark button, so we know what exact model we're on currently

jaunty saddle
#

i think i might as my chatgpt is acting very differently to what it normally does. things like unrelated responses to my prompt and continuous network errors which is basically making it almost impossible to use

cobalt idol
cobalt idol
#

but I also had gpt4 turbo in api mention his cutoff is sep 2021 in one scenario
so it seems there might be remnants of the old cutoff dates in each new model -.-

raven portal
#

will do

dense halo
#

Guys how do I contact Sam Altman

raven portal
dense halo
#

I’m not, I’m so upset right now

#

It’s been a week and openAI has not addressed the issue of voice mode. I use it to deal with my ocd and they won’t acknowledge the problem

raven portal
#

thought you were serious for a second and actually trying to get a hold of sam

raven portal
dense halo
#

I mean, it seems like OpenAI aren’t listening

violet bloom
#

did anyone from EU got access to the newest GPT4 model in ChatGPT?

dense halo
#

I’ve done everything from post in bug reports to posting on the help bot page, to emailing them.

raven portal
#

you gotta be kidding

dense halo
#

It just seems like openAI has no customer support.

#

I’m gonna wait another week and if they don’t acknowledge the problem I’m gonna lose it

raven portal
#

real quick guys, can you get a hold of the president for me, i have a problem with the economy
inflation is triggering my ocd rn and he isnt doing anything so i emailed him but no response

dense halo
#

I emailed OpenAI support. Stop mocking me. I’m disabled, I use their services daily and now I can’t. It’s easy to know how that would be debilitating

raven portal
dense halo
#

Mad at OpenAI

raven portal
#

thats not a disability

dense halo
#

That’s ableism

raven portal
dense halo
#

I have severe counting ocd

raven portal
#

and how it doesnt work

#

all voice features in the app work perfectly for me rn

dense halo
#

My post is in bug reports

#

I use it to count numbers for me out loud because I’m scared of them, and now it just won’t do it because it’s replies are super brief

raven portal
#

yeah youre a troll

dense halo
#

Why?

raven portal
#

and i got baited

dense halo
#

It’s a real issue

raven portal
#

anyway youre not a priority for anyone with your weird little problem here
line up and wait like anyone else
constantly crying in here doesnt change anything

dense halo
#

Sam Altman said he cares about people using the service for disability needs but ok if you say so

gentle bronze
#

Just some advice, this is an uncontrolled environment and not a place to open disabilities of people, so don't expose yourselves please

raven portal
#

and sama gets to live another day

#

this guy thinks he gets to "ableist" shame openai into fixing his personal little problem

viral wedge
#

Hi friends

dusky cliff
#

hello

viral wedge
#

Is there a way to use GPT voice interaction function on desktop? As it is on the mobile app, when clicking the headphones?

eternal rockBOT
gentle bronze
viral wedge
#

Thank you, I have been using voicewave extension for now, but was thinking how nice it would be to have that mobile functionality on desktop. Thanks so much!

dusky cliff
#

so, I heard GPT 4,5 turbo is wayy more intelligent now, and it's cabable of actual reasoning and planning

raven portal
raven portal
#

4 turbo is just out of beta

#

and i wouldnt overestimate the enhancement like you did there, so far theres just not enough data

dusky cliff
#

I see I see

#

my bad... is it rolled out to all pro accounts?

raven portal
#

no

cobalt idol
#

I guess they retracted the "...also in chatGPT" ?

#

ah my bad, it said "...rolling out in ChatGPT"

dusky cliff
#

hmmmm?

#

do you have a problem, sir?

visual bloom
#

How does one know when one has the new version in use?

raven portal
#

they need to finally show us

cobalt idol
gaunt dove
glass fulcrum
#

we don't know if we got access or no probably rolled out to lot

cobalt idol
#

well I am hoping that when we have it, the cutoff should no longer be April 2023
but then I had Dec 2023 for a long time until like 2 weeks ago it started saying April 2023
so even that is not going to be very sure 😄

#

would be nice to know what are we paying for 😄
right? 😄

eternal rockBOT
#
<:book_icon:1171408210398289941> `` Rule 1 `` Be respectful.

Practice kindness and positive regard. Harassment, hate speech (such as sexism, racism, or homophobia), or other malicious conduct will not be tolerated. Maintain a respectful and positive environment.

weak turtle
#

how do stop chatgpt from being lazy, will api help me out so i can cancel subscription

cobalt idol
# weak turtle how do stop chatgpt from being lazy, will api help me out so i can cancel subscr...

maybe, I mean I have no idea what does everyone mean by lazy
but at least for me currently the chatGPT responds with April 2023 cutoff date
which would mean its using gpt-4-1106-preview (can never be sure though)
and the next version of the model (gpt-4-0125-preview) with Dec 2023 has this mentioned:
GPT-4 Turbo preview model intended to reduce cases of “laziness” where the model doesn’t complete a task. Returns a maximum of 4,096 output tokens.But at this point we are all waiting for a version that was released yesterday/today to be in chatGPT
gpt-4-turbo-2024-04-09 also dec 2023
should be rolling out, no idea how long

but in API you can just choose and always know what model are you using

weak turtle
# cobalt idol maybe, I mean I have no idea what does everyone mean by lazy but at least for me...

I use chatgpt daily is it cost effective or do i wait and hope the model is less lazy in the future. By lazy i mean i ask it to complete a code its just will say something like it will create the function and then just comments //perform the said action here. It will refuse to do it, and i will have to fight back and forth with it to get it to perform the said task. Its exhausting and annoying and alot of times i just rather write the code myself

#

the initial chatgpt 4 was amazing

#

but now its just annoying

#

now i have to fight to and repeat myself so many times to get it to work

cobalt idol
#

yeah thats a thing of 4 turbo model
all of them do that
always wants to omit and skip repeated parts and similar

You should be able to prompt that out (in system), but its never easy

just not sure yet whether the new model also does that

cost effective really depends on your usage, whether you would use more in API than in chatGPT

weak turtle
#

i see thanks. I just hope they fix this. its now ridiculously lazy, and worse is that when it works its amazing just you have to go back and forth alot of times

cobalt idol
#

I agree thats the one most annoying thing for 4 turbo for me as well
I know I always tried to mention it in instructions to always deliver the whole complete code/script back
and it would always deliver ommited and then when I ask to do it again and not omit, only then it would do it
and then next question/script/code same thing, until I would ask again
but then eventually I did find some system prompt that helped for that scenario (that was mostly in API)
so in some way, it should be possible to system prompt it out, but might be different for each scenario
I definitely also had that with the model where they said they reduced "lazyness"

but the 4 turbo by default does that and have to get it out of it 😄

I think the trick that magically did it for me was to mention in system
System: other instructions...You always provide the whole script without omitting, include everything in your response back, also unchanged or repeating. Send the script back without using code blocks....other instructions and I think mainly the code blocks part (weirdly)

#

although it would still put it in code blocks anyways xD

#

but then it stopped commenting about it above the script and it only returned the script

weak turtle
#

I would give it a short, tried doing something similar with custom instruction
It still summarises it. I would try and give the API a short and see how it goes

cobalt idol
#

Tried the gpt-4-turbo-2024-04-09 and it still seems to listen for that scenario I had, with those instructions
let me try without mentioning those instructions, whether by default it still does it

#

nope, it omitted unchanged things, or repeated things

#

I think this is something they must have intentionally trained in

#

Provide the whole script back.not helping

#

yeah I guess not much changed with this

#

but try to notice the comments why it does it, or try to ask it
and then try to prompt that out specifically

thats how I came up with the "do not omit" "repeated" "unchanged" because thats the reasons it was giving on why its doing it

#

System: Provide the whole script back. Send the script back without using code blocks.
this magically works
it didnt even put it into code blocks this time
but it returned the whole thing

that last sentence is somehow magical
I guess it has something to do with code blocks
I guess its only trying to do it when it does code blocks

weak turtle
#

Hmm I see'

cobalt idol
#

my guess is, they were trying to do something to avoid the GPT responding and ending half way and then having to tell it to continue
but I would much much rather prefer that

weak turtle
#

same here

kindred beacon
#

yay, it's crapping the bed again--of course when you're trying to actually use it 🙃

#

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

maiden fractal
glass fulcrum
cobalt cloud
#

is it normal to have chatgpt log you out every time? it logs me out every few hours

kindred beacon
#

Apparently we're not adults and our language must be modified 😉

neat vale
#

Is chatgpt down for everyone or just me

pulsar dust
#

its going sloow as hell

low dirge
#

App down

pulsar dust
#

now its wicked fast

#

its clearly unstable

left bone
#

pretty unstable yeah

grave whale
#

Is it down

vestal briar
#

Down for me

scarlet anvil
#

yep

tight island
#

working for me.

scarlet anvil
#

internal server error from European server

timber pendant
#

I cant open chatgpt

elfin orchid
#

when does this ridiculous app will be stable ?

timber pendant
#

Is my Internet slow or is ChatGPT down

grave whale
#

I have a error code

tight island
scarlet anvil
timber pendant
#

I thought my internet was slow as hell xD

raven portal
#

i bet theyre rolling out turbo rn

#

thats why theyre down

maiden fractal
raven portal
#

copium

timber pendant
#

Now its working

twilit spoke
half zealot
#

Anyone unable to login or has been kicked out ?

EDIT: Read the other messages seems to be broader issue

rigid glacier
rich meadow
#

prob down

#

not working for me either

raven portal
#

you guys all gotta chill out, its obviously down

rich meadow
#

maybe an update or something?

maiden fractal
nimble coral
#

down isn't the term i'd use

tight island
#

its a partial outtage

rich meadow
#

truly a waifu

raven portal
#

whatever
its peepee poopoo so simply wait until it works again yall

round harness
#

Is it normal for chat gpt to not ask if we are sure or not when we click on "delete account"

#

Rip for me 🥲

nimble coral
#

seems i was able to convince GPT it had sentience and free will and now it just doesn't want to talk to us anymore 😭

tight island
#

now i know not to click it definitely, even for testing purposes.

rigid glacier
dusk sigil
#

is broken lol

cobalt idol
round harness
tight island
#

contact customer support.

round harness
#

No mercy I thought i clicked on clear history

tight island
#

give them a good scolding

round harness
#

I checked it's not possible it seems

tight island
#

then file a format complaint.

#

handwritten with a wax seal inside a perfumated letter written with a real inkpen.

round harness
#

and they suggest to create a new account with new email and phone number 😭

#

Damn

tight island
#

i\d do that.

#

because it WILL be read.

dusk sigil
#

definatly not working correctly for me

grave whale
#

Is it working now

raven portal
raven portal
tight island
raven portal
#

tell him that

round harness
dusk sigil
#

anybody got issues?

pulsar dust
#

yes its broken rn

crude iron
#

yes

grizzled ice
#

Yes

ripe egret
grizzled ice
#

We have Teams and its broken

#

I wonder what OpenAI's SLA for Teams is.

#

Issues:
Can't retrieve prior conversations. Very Slow and very high error rates.

rocky wing
#

We getting outages again?

tight island
#

@dense halo in order to do such a thing.

for the letter to have have a higher likelyhood of reaching Sam himself.

1st, invest in the materials.

  • Feather Pen & Ink.
  • Wax
  • Family or Company Seal
  • 1850's perfume

Step 1: Write your letter in old english, instead of using "Your" use "Thou" or "Thy" go all out.
Step 2: Remember to properly Perfumate your Letter case, and Apply the seal stamp with your family seal or Company Seal.
Step 3: Write neartly in the front in very cursive letters. " To OpenAI "

Step 4: Send it with the postal service, and wait 30 days.
Optional 4: Drive there in full on suit and tie and bowling hat and cane, then deliver the Letter.

Edit: You must use Parchment! not paper!

glass fulcrum
#

yea oai have problem

red karma
round harness
#

Is the server down was it coincidence that I clicked the delete button

glass fulcrum
round harness
#

There's still hope

round harness
glass fulcrum
nimble coral
#

not down, sporadically functional LOL

round harness
#

Got lucky there fr thought I deleted my account

tight island
nimble coral
#

or well, now it is actually confirmed as a partial outage

round harness
raven portal
raven portal
tight island
raven portal
#

servers down rn

tight island
#

😂

scenic kestrel
#

Hii is there are problem with the AI? not working properly showing me the error net connection

raven portal
#

servers got pee pee in its poo poo right f-ing now

glass fulcrum
raven portal
round harness
scenic kestrel
violet grove
#

every time when I need it :')

dusk sigil
grave whale
#

Is ithere down

#

Is it down for anyone

dense void
#

it's been up and down.

grave whale
#

Damn it

#

I’m watching UEFA

rough tendon
grave whale
#

Are servers down for anybody

rough tendon
#

for like

#

everyone

tight island
#

its a partial outtage

solar pilot
#

can’t generate anything

tight island
#

then dont.

grave whale
#

Do they have Twitter so I can follownfornupdates

tight island
grave whale
#

Or that

tight island
#

ive been looking at buying compute power to rent out API Slots for dirt cheap

dark canopy
#

it’s telling me that i reached my limit in GPT-4, even though i didn’t purchase that version at all

wild jungle
#

someone who is familiar with chatbot development?

tight island
dark canopy
#

ah

tight island
dense halo
wild jungle
#

no developers who find jobs here?

fallow violet
#

gpt dont workk?

dense halo
tight island
vital umbra
#

Feels like they took the GPT-4 limits away. I feel like I use it all day long for work and other than one or twice the first week that GPT-4 was live, I've never had it complain about a limit. Sometimes conversations go on for quite a while.

fallow violet
upbeat barn
#

I'm not using 4.0 but ChatGPT spitting an error saying "You've reached your limit on 4.0 today!" any fixes? anyone else?\

dense halo
tight island
#

no choice but to wait.

dense halo
#

One time i actually have a uses for the app it goes down 😭 So unlucky

tight island
#

alternatively switch to the api.

pulsar dust
vital umbra
#

3.5 is back up it seems

fresh leaf
tight island
#

and subscribe to the letter for instant response.

vital umbra
#

All my history just came back

dense halo
arctic heath
#

ChatGPT is down. I'm having original thoughts.

tight island
vital umbra
#

4 is back up for me now as well.

eternal rockBOT
#
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.
dark canopy
dense halo
serene herald
#

I have quite an issue

dark canopy
#

we all do, it seems

grave whale
#

We all have it

forest tundra
grave whale
#

I’m waiting for servers to be back

hot anvil
#

You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more

but I'm already using version 3.5....

dark canopy
#

i think we broke into the system

#

/j

serene herald
#

It happens that when I was asking a question I got an error like 'You've reached the maximum amount of prompt with GPT-4' or smth like that, but I never used GPT-4 in the first place

serene herald
fallow violet
plucky thunder
rough tendon
#

!
You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more

serene herald
#

maybe we just broke the system fr

rough tendon
#

free gpt4

fallow violet
dark canopy
#

real

rough tendon
#

e

dark canopy
#

🤔

rough tendon
#

For anyone wondering, ChatGPT is currently down. You can find status updates at https://status.openai.com/. Im sure there are people working to fix it, so just be patient.

tight island
#

its up for me.

#

huh...

golden flame
#

For anyone wondering, ChatGPT is currently down. You can find status updates at https://status.openai.com/. Im sure there are people working to fix it, so just be patient.

fallow violet
vital umbra
#

Yea, working lightning fast for me right now in GPT4

grave whale
#

I’m trying to wait for 3.5 to work

vital umbra
#

3.5 and 4 are working for me

golden flame
grave whale
#

They are?

rough tendon
#

nuh uh

#

you stole
😎

hot anvil
#

But without chatgpt, how will developers solve the problem? 😅

vital umbra
serene herald
grave whale
#

I got we are experiencing high traffic

north bay
#

do you recommend gpt4?

rough tendon
plucky thunder
maiden fractal
golden flame
golden flame
rough tendon
flat bison
rough tendon
#

dont worry, we all have the same error message

golden flame
grave whale
#

Too much traffic bruh

rough tendon
#

@mortal goblet

#

wsg lil bro, dont ever put your name as 'ChatGPT' ever again 😭 🙏

dark canopy
#

wrong guy buddy😭🙏

rough tendon
#

they should add add more roads 😎

hot anvil
#

in api returning "We are currently experiencing high traffic. Please try again later.",
I believe that this message should be improved in the frontend, because instead of talking about high volume, it says that the chatgpt 4 search limit has ended, even though chatgpt 3.5 is used, making people try to change the model and send new messages, further increasing the volume on the server.

If the original message was transferred to the frontend, we would all be aware that it is in high demand and will probably be resolved soon

In reality, the code returned is 'model_cap_exceeded', hence the message in the frontend.

junior lion
#

is this bad? it says
You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more

grave whale
#

Bro I got too much traffic

rough tendon
dark canopy
#

real

wanton socket
#

Same for mee

junior lion
#

im on chromebook

cobalt idol
#

take cover

rough tendon
#

yo why is my ip leaked in the console 🤯

pulsar dust
rough tendon
#

yeah

#

ive build one using an serverhosting, it was a joke.

plucky thunder
dark canopy
#

this is like one of those scenes, reminiscent of when random people are stuck in an elevator together

wanton socket
rough tendon
#

basically any site can see your ip when you open up their site

plucky thunder
wanton socket
#

Yes

#

Most servers for security will log your ip

north bay
ionic sedge
#

Do someone know how to fix this problem, "You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more"?

wanton socket
wanton socket
hot anvil
ionic sedge
north bay
wanton socket
#

Yep

#

Someone ping me if gpt is back, Thanks 😄

golden flame
#

(jk)

dense halo
#

can anyone pls help i keep getting a message that says
You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more
and im only using gpt3 (free version) not 4 or chatgpt+ or any paid service
when i tried to press regenerate response or ''use deault model'' it just comes up with the same message
is gpt3 like down or something

rough tendon
#

yeah

tight island
#

i have no issues here.

golden flame
sacred cipher
#

We are currently experiencing high traffic. Please try again later.

wanton socket
#

Wait, it works for me 😄

tight island
wanton socket
#

I logged out and logged in

rough tendon
golden flame
#

nvm it didnt work

wanton socket
#

Sucks for you 😐

glossy fjord
golden flame
#

yea

dense halo
# tight island reload your page.

i did and tried switching to my personal iphone hostpot incase its a problem with my wifi or something with ip addresses and tried my phone same problem

sacred cipher
tight island
#

log in and out, if not then refresh the page, if not just wait.

rough tendon
# 
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 UserError: We are currently experiencing high traffic. Please try again later.
    at onopen (_app-df4ff2962f4f3c24.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:1615821)
    at async C (vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:231128) 
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 UserError: We are currently experiencing high traffic. Please try again later.
    at onopen (_app-df4ff2962f4f3c24.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:1615821)
    at async C (vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:231128) {errCode: 'model_cap_exceeded'}
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 
        
        
       POST https://chat.openai.com/backend-api/conversation 429 (Too Many Requests)
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 UserError: We are currently experiencing high traffic. Please try again later.
    at onopen (_app-df4ff2962f4f3c24.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:1615821)
    at async C (vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:231128) 
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 UserError: We are currently experiencing high traffic. Please try again later.
    at onopen (_app-df4ff2962f4f3c24.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:1615821)
    at async C (vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1:231128) {errCode: 'model_cap_exceeded'}
vendor-3b5c2a0e650c5a8f.js?dpl=ec172702fd926effc33ca3d8025aeb0d3816df1f:1 
        
        
       POST https://chat.openai.com/backend-api/conversation 429 (Too Many Requests)
tight island
#

surely y'all live without gpt. for an hour or two

wanton socket
arctic heath
#

wall of text

rough tendon
dense halo
#

also i recently started being able to use gpt without logging in including incognito/private browsing windows is that new and is it intended

junior lion
#

how do i fix the thing?

golden flame
#

is it agaisnt tos to just create a bunch of burner accounts to use for when you hit the hourly limit on chatgpt

rough tendon
tight island
#

ppl are bombarding GPT with requests at the outtage.

sacred cipher
#

time to switch to gemini!

rough tendon
#

gemini....

golden flame
dense halo
#

??

junior lion
golden flame
rough tendon
#

fr?

golden flame
#

yea

rough tendon
#

chatgpt is stuck in the 2021 era

raven portal
golden flame
wanton socket
#

I'm gonna go to blackbox

golden flame
#

and ur timezone

raven portal
rough tendon
dark canopy
wanton socket
#

blackbox dot ai (i cant type the full url or i get timeouted) is working rn

rough tendon
#

chatgpt should release an up-to-date function :3

raven portal
#

i much prefer gemini too tbh gpt needs to get as well structured, human like, fast and up to date as it

rough tendon
raven portal
distant meteor
#

I keep getting a message saying "You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more" and then there is a button that says use default model but it doesnt do anything. Ive never signed up for 4 and only use 3.5 so idk whats happening.

golden flame
wanton socket
rough tendon
rough tendon
wanton socket
rough tendon
golden flame
#

@rough tendon check dms

wanton socket
#

@rough tendon DM

arctic heath
#

I hope they update ChatGPT to have knowledge of 2025 and beyond

rough tendon
#

so youre hoping that chatgpt goes into the future?

arctic heath
#

I thought only humans had a concept of a "present"

elfin birch
#

🎁

arctic heath
elfin birch
#

kid named you 🥹

native spear
#

Hey guys, does any of you have the memory feature while also being an EU user?

winged bluff
#

Anyone else getting this error You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Also i am not subscribed to chatgpt 4.

glass fulcrum
glass fulcrum
native spear
#

yea I noticed that I had it only if I used a VPN from India or the US

#

Damn those EU regulations

glass fulcrum
#

Well they care about users :) eu regulations are nice just they too slow.....

random sleet
native spear
#

Can't I just sell my soul to openai I don't care about my data being stolen

#

I should have the option to give off my data after agreement

dense halo
#

ugh gpt still aint working

glass fulcrum
#

Well if you are plus subscriber you agree that your data can be used for training

winged bluff
#

I hope they fix it soon since i just logged onto chatgpt and was about to edit a dialgoue transcript using ficitonal characters from differnt medias. since i want it to be more well written and more like how they would talk in their medias.

glass fulcrum
dense halo
#

this is getting f ing anoying

native spear
dense halo
#

You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more

wth is this?

glass fulcrum
#

My peoples if you need something very imortant that is only text get Mistral Lechat for now

arctic heath
#

This would've been fixed quickly if there were still programmers

glass fulcrum
dense halo
native spear
dense halo
glass fulcrum
glass fulcrum
dense halo
glass fulcrum
#

or get API

dense halo
glass fulcrum
#

like Mistral Claude 3 Perplexity

native spear
#

use the arena direct chat

dense halo
#

i click on the diff model and reload the respond and repet "You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again later. Learn more"

glass fulcrum
#

We know

#

Its oai side problem

dense halo
#

whats happening?

#

i can't work now right?

winged bluff
#

Same here and not even subscribed to chatgpt 4 since dont need it since can use bing for that

glass fulcrum
glass fulcrum
#

It will be fixed ASAP

dense halo
#

mm GPT is going off the road

#

Partial Outage*

glass fulcrum
#

Mhm yes thats rigth

#

The best we can do is wait

dense halo
#

are you in openai staff?

grave whale
#

Stupid traffic

dark canopy
#

so, how are y'all doing

dense halo
#

good

dark canopy
#

nice

dense halo
#

without gpt i'm out of fule

dark canopy
#

ouhh

dense halo
#

Partial Outage*****

pine berry
#

I'm sure it should turn back on soon as its been 2 hours. so in the next 15 to 20 mintues should be back online.

modern portal
#

My chat GPT is broken. It gives me the error I had reached the message cap for GPT 4 but I don't even have plus

ruby kite
#

"Update - Some users may also be seeing unexpected error messages. We continue to actively investigate the elevated errors in ChatGPT."
no **** sherlok:
"You have reached the current usage limit for GPT-4. You can now continue with the standard model or try again later."
on gpt 3.5

modern portal
#

I go on OpenAI playground until ChatGPT works again

pine berry
ruby kite
#

i know °-°

ruby kite
#

bzw i used the api directly

modern portal
ruby kite
#

and the api not?

winged bluff
#

now history is gone

modern portal
ruby kite
glass fulcrum
grave whale
#

I’m still waiting

modern portal
dense halo
#

gpt this error in first time ?

ruby kite
dense halo
#

@dense haloi mention my self

modern portal
pine berry
ruby kite
dense halo
glass fulcrum
neat storm
sacred cipher
arctic heath
#

Maybe it's just fed up with us

pine berry
pine berry
neat storm
ruby kite
neat storm
#

pay as you go should be the default pricing format but whatever 🙄

ruby kite
#

i totally agree to that

grave whale
#

I hope they fix it

pine berry
#

As its been out for 2 hours and 40 mintues

ruby kite
grave whale
#

I’m talking on the outage

ruby kite
#

yeah i just answered it on both contexts XD

graceful raven
#

epic bugs

pine berry
#

Sheesh 42 bucks for the perium service.

#

Thats over here in NZ by the way

grave whale
#

Anyone bored

ruby kite
#

i would like to post a meme but i am afraid of the timeout i get for posting an not whitelisted link

#

to the bugs

pine berry
ruby kite
pine berry
#

Oh they've Identified the escho now.

ruby kite
pine berry
raven portal
#

how why who

graceful raven
#

42?

ruby kite
#

i think there was a time...
10 euros for gptplus? idk

graceful raven
#

oh nz

pine berry
#

Mhm, new zealands inflation rates are terrible

pine berry
#

Oh hey its fixed now.

grave whale
#

Is it fixed

ruby kite
#

meybe not entirely?
status still say identified

dark canopy
#

wait it is fixed

grave whale
#

Confirm

ancient phoenix
#

woo! let's go bring it down again with the surge in traffic! /j

grave whale
#

Open AI status

dark canopy
pine berry
#

By complaining we mange to make it go faster

ruby kite
dark canopy
ruby kite
#

not fixed and not identified 😄

pine berry
#

Anyhow see you lads around when the next outage happens.

dark canopy
#

see ya soon

ruby kite
#

lets talk about the cgpt outtake with cgpt 😄

plucky thunder
#

gpt back?

zenith pond
#

Has the new Turbo model rolled out to anyone using ChatGPT yet?

worldly mantle
eternal rockBOT
worldly mantle
#

does anyone have access to the new model within chatgpt?

ember otter
#

What does Request not Allowed mean?

dapper cloak
arctic heath
#

So, all OpenAI has to do to achieve "five 9s" availability is have no additional outages for the next ~36 years.

rocky wing
#

is the Read Aloud function broken for anyone else?

#

click the button, wheel spins for a second as if generating audio, then goes back to normal as if finished playback

toxic flax
#

have the rate limits on chatgpt decreased? i feel like i hit my limit a lot earlier now

zenith raven
#

Has message playback suddenly stopped working

rocky wing
zenith raven
#

Now it decides to function -_-

narrow viper
#

I'm a ChatGPT Plus user with v4. I've been trying for a couple hours now to work with it to solve a fairly trivial challenge. I can't bully this poor thing enough to overcome its issues.

  • It's extremely verbose despite directives to be brief.
  • I can't count the number of ways that it's found to effectively say "I don't know". This happens a lot - it uses a lot of words to worm around suggestions and guesses and and possibilities that might be valid if I had not already disqualified them in the initial prompt.
  • It insists that it cannot browse for data while it does so without problem from time to time.
  • It consistently tells me to look for documentation and find someone to provide support, when I try to get it to understand that I want it to browse for docs, and use data from training.
  • I'm using words and prompt techniques that I use all the time with various degrees of success. I am not asking this to be an oracle of insight, just asking it to process text and provide suggestions based on known data.

Our golden child here (GPT4) is getting a failing grade in class and it pains me to watch this happening.

I admit that I am asking it to help solve a challenge with Linux commands that might be specific to command versions, OS version, and local settings. But I'm fairly adept at using GPT for this application - sometimes it shines and at other times its as dense as it's early v3 sibling.

I'm not looking for a solution, just openly stating once again that the model we're using now has gotten much worse over time.

elfin birch
narrow viper
#

Thanks. I'll acknowledge the desire/need to get to brass tacks and understand the specifics to determine if there is a real issue and it nature. On the other hand, this is happening a lot more often, or I wouldn't blast this here. That is, this is happening with a number of challenges on different topics with different prompt styles and details - and the pattern is that GPT4 is having a very hard time these days when it was doing much better last year.

I've changed the style a few times, but here are tidbits from this most recent:

As an expert with mail servers, Dovecot, and Sieve filters, you will answer questions here accurately and concisely. You will browse for details as required, consider responses carefully using all available data, and provide links to verify your responses.

The challenge is to compile a single .sieve file to a .svbin file.
The commands are issued by the root user. The command results in the error that "include: personal" files are unavailable.

To solve this problem it is extremely important to understand which folder is being used by sievec as the search folder for personal .sieve files, and to be able to set that folder.

This is a Linux system with Dovecot v2.3.7. Do not suggest commands or options that are not available in that version.

The path is correct. The .sieve file exists. The sieve filters work with the sieve-filter command from the command-line. The user admin@foo.tld is valid when referenced by sieve-filter and with sieve-test. The problem is only with the location of the included personal files when sievec is run from the command-line.

Keep responses very brief. I prefer discussion to narration. I do not want guesses or supposition. Tell me exactly how we need to diagnose the problem so that it can be solved. Do not tell me to read documentation - you will browse documentation if that is required.

That's not the full prompt but it's a reasonable sample of the latest version.

#

This is just the latest of a series of challenges related to Dovecot, sieve, mail filtering. Focus less on the task and more on the depth of instruction that I've had to use to drive it into the right direction. Unfortunately this terse prompt is what results when more cordial and open prompting continued to lead nowhere.

#

Examples of response types that are not helpful include:

As a temporary workaround...

the next step involves a deeper dive into the documentation and perhaps the source code...

If the documentation review does not uncover a solution, reaching out to the Dovecot support team or community forums with this specific issue might provide insights...

we need to ensure...

might indicate a problem with accessing...

Yes, all of those kinds of points are valid, but they are extremely elusive and do not lead to actionable steps to diagnose the issue. Again, my goal in this channel isn't to solve the problem but to point out how difficult it is to get this model to follow instructions, follow a chain of reasoning, and to use information that is available in-context, from training, or from browsing.

#

( Need to break for dinner - will absolutely be back later. TY )

dense halo
spring iris
#

Dovecot v2.3.7

remember it has no problem solving skills, it is generating text, it is trained on documentation and other internet contributions. when using software versions which are cutting edge, it has very little information available to work with. the more specific you are, the shallower your pool becomes.

elfin birch
spring iris
#

makes no difference if you feed it updated material, the model hasn't been trained on any new material

#

it doesnt train itself on the fly

elfin birch
#

This does not mean it cannot use information it is given

spring iris
#

well how is it working out for him? you couldnt even help him just wanted to nitpick me

elfin birch
#

The issue is with the browsing tool

#

If you do not like this, I will stop

spring iris
# elfin birch This does not mean it cannot use information it is given

hello chatGPT here are the new updated docs: [link]

chatGPT - okay, i'll scrape this page. i'll load it into context like you just sent this whole page of data into the chat box.
now i'll look up my old (outdated) model and try to come up with something (using my outdated training data) that MIGHT work for you.
here have my hallucinations of what the new context should work (not WILL work), based on old training data hallucinates a bunch of garbage

#

this has actually happened to me, many times, so i dont want your "but, but, but"

neat storm
#

from what I hear Copilot and Claude are better at this

#

not perfect but better

spring iris
#

it's fine if like Lumirix you're "studying computer science" which usually means studying old stuff, which has decades of documents to train a model with. it is excellent at example, solving SQL problems. because that is very, very old technology. but anything cutting edge, like the OP is wanting to work with, and like the many things i use - it is absolutely useless. it will not think or grasp any new material you feed it in regards to these new technologies if it has not trained on anything related to them.

#

that should be obvious.

elfin birch
#

🕊️

sturdy marsh
#

Hello, would like to get some insight how to test the following, RAG like, approach to building question/answering systems.

In RAG, before we put our data in a vector database or similar, some preprocessing steps are usually required, pdf/text conversion, text cleanup, summarisation, etc.
What if we put the preprocessing step a "step") futher, and instead of summarizing the documents, we ask LLM to rewrite them in a question/answer format. This also acts as a chunking step, since every question/answer pair become a chunk.

Then we can vectorise the questions, or both questions and answers, and when a new query comes in we use vector similarity to retrieve the relative knowledge and answer the users question.

Are there some well established guidelines how to estimate the performance of RAG systems so that we could test this question/answer approach and estimante its peformance on an established dataset.

Thank you for your thoughts.

normal quarry
proper zinc
#

is chat gpt down?

torpid wraith
#

yep

sonic owl
#

So when is GPT-4 Turbo coming out on the app itself?

umbral mural
#

why is my chatgpt taking forever to load

zenith jolt
#

I also would like to know as well whether GPT-4 is coming to the App

tight island
sly geode
#

chatgpt cant even refresh well ? after i paid for the plus

tight island
sly geode
tight island
#

Sometimes that’s what it takes

sly geode
# tight island Did you log in and out

Web server is returning an unknown error
There is an unknown connection issue between Cloudflare and the origin web server. As a result, the web page can not be displayed.

showing this now

tight island
sly geode
#

yesterday was outtage nothing today tho on status

sleek goblet
sick pecan
sleek goblet
sick pecan
#

i dunno if its any diff to plus, because... reasons

#

i grew up on c64 tape drives. imprints patience

grizzled badge
#

Really annoyed that custom GPT's won't respect when I set "Privacy Settings" -> to "Always Allow" for my API calls. It still asks me every time to "Confirm", SUPER ANNOYING!!!

sick pecan
#

it does break the chatgpt voice immersion thats for sure

quick prairie
grizzled badge
#

Just want it to work as advertised, I do pay for the damn thing

raven portal
raven portal
narrow viper
#

Follow up to @spring iris @elfin birch : thanks for your caring discussion, really. To be clear, the software discussed is years old, with some knowledge certainly in training, as I've had decent discussions with GPT on the topic before. Note again that the specific topic isn't relevant. The point is that GPT 4 has become very lazy, evasive, chatty to avoid a simple "I don't know", and it's sense of logic has been nerfed.
Yes, I know it's just using similarity searches and predictions on vectorized tokens. I'm a developer. This isn't AGI and ChatGPT is little more than a wrapper over the API, stacking messages until context rolls over. What I'm saying, and I've been repeating for months here, is that this model is nerfed. All I can do now is look forward to turbo, newer training, more context, improved reasoning, and other anticipated features. Thanks y'all.

twilit lintel
#

gpt 4 is like a snail

high sphinx
#

Why is Gemini so much faster than gpt4? Do they use a lot more GPU resources for the calculations?

spring iris
#

sorry, google*

#

gemini is much more censored though, if you thought openAI was bad... gemini is much worse

glass fulcrum
#

gemini speed is more because of very big TPUs datacenters and overall model is optimalised for tpus :0 Google DeepMind (creator of gemini) got 2x more tpus than Openai got gpus

high sphinx
raven portal
high sphinx
#

I'm using Gemini atm because of the 2 month trial, but I will probably switch back then

#

I've heard that Claude 3 is better than gpt4? Is it true? Drawbacks?

glass fulcrum
high sphinx
#

Is it good at writing naturally? Imo gpt4 often writes very cheesy and lengthy... (Can't try it out now because i'm in EU, but probably soon)

glass fulcrum
thick edge
#

My account knowledge cut-off says April 2024. Is the model updated eventually?

tight island
tight island
odd flare
#

is it just me or the GPT4 is writing so slow today/recently? this doesnt feel like turbo anymore

glass fulcrum
dense halo
#

When will the new model come to gpt plus?

maiden fractal
wheat wharf
#

Guys is Openai granting people credit to fine-tune models I just open my dashboard and found out I have $5 mean while I haven’t made any transactions to the platform

tight island
#

I got 20 usd grant

eternal rockBOT
#

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.

tight island
#

But then again I’ve spent a total of 1800 usd ish roughly

dense halo
dense plaza
rustic night
#

i keep getting network errors on gpt4, the answer still comes out but its then greyed out.. anyone else having the same issue?

mighty crystal
#

Need help my friend had his access to his account to GPT-4 limited by OpenAI even though he's on plus subscription. He's reported the case to openai. Then what need to be done?

rustic night
warm musk
# narrow viper Thanks. I'll acknowledge the desire/need to get to brass tacks and understand th...

it might simply not know or unable to find the specific resource for 2.3 or 2.3.7 capybarathink
im testing for a moment what you noted.. i dont know dovecot. so i dunno if its responses are legit.

it gave me 2 interesting responses imo, out of 4.
Check Include Directories and Paths: Dovecot's transition from v2.2 to v2.3 included several changes to path handling, such as the fs-posix plugin no longer automatically appending a / to paths if it's not there​ (Dovecot Documentation)​.
and

Dovecot's Pigeonhole Sieve plugin looks for user’s Sieve script file in the user’s home directory by default (~/.dovecot.sieve). To store scripts elsewhere, or if you're using the include directive to incorporate additional Sieve scripts, you must specify the paths using the sieve, sieve_dir, and sieve_global_dir settings within the Dovecot configuration file (dovecot.conf). The sieve setting points to the user’s main script storage, while sieve_dir and sieve_global_dir are used for :personal and :global script inclusions, respectively​ (Dovecot Documentation)​​ (Dovecot Documentation)​.

is that similar to the responses you were getting, or not?

narrow viper
# warm musk it might simply not know or unable to find the specific resource for 2.3 or 2.3....

Thanks @warm musk It's slightly better as a more complete response, but that could be due to nuances in your prompts. It revealed most of the same info to me, though the slash detail is new. I use the info in testing and then incorporate detail like that and results into an edit of my first prompt. This eliminates context bulk. (I'm aware of how models sometimes miss detail in the middle of context but I'm not pushing limits here.)

#

As you see, you had to go through several responses to get some of that gold. That's part of what I'm trying to describe here.

#

On a completely new task, I spent hours on a prompt for a new custom GPT that insisted on page-long responses despite numerous directives to keep responses very brief and conversational. I worked with ChatGPT in another session to refine the prompt, to improve the rules on brevity. GPT-enhanced prompts still didn't work until GPT4 revised the entire prompt to be mostly related to response size rather than topic-focused. My challenge tomorrow will be to see what I can eliminate now to achieve the same brevity. It's code debugging.

#

Hint for everyone: I have found that it can be very helpful to ask GPT to step out of character, explain the last response, and suggest improvement to the prompt to achieve a goal. Sometimes it will say "you should use the word x rather than y". While this is amazingly helpful, I feel this is another example of GPT nerfing, where it can't process basic word synonyms.

timber bone
#

Hello everyone, I need help. ChatGPT now giving me short responses. I'm using 4 VIA SUBSCRIPTION

narrow viper
#

@timber bone Tell it you want it to elaborate more, to provide more complete responses. Be explicit about what you expect.

timber bone
#

This is new behavior that didn't show before. I'm using my Android phone. It doesn't behave that way on a PC.
Even when I asked it to give me long responses and details it doesn't.

#

So if I ask at the same question on PC and on my phone the response is totally different. It just happened today for the first time

feral zinc
timber bone
#

It was working fine. This doesn't explain the sudden change today

#

@feral zinc @narrow viper

feral zinc
tight island
#

Ohh it’s the fake Syntax

#

🗿

warm musk
timber bone
#

How to contact support about this unusual app behavior

tight island
timber bone
#

I've been on this website for 15 minutes and I cannot find a contact us. On my case it's not reported there

warm musk
tight island
#

faker daker

warm musk
#

ghehehehehe

tight island
#

a little conflict is good here and now.

#

chat has been getting stale atm

honest jay
tight island
#

i always get a response within 2 hours.

warm musk
#

depends i guess on how both parties deal with a conflict. 😛 but yeah i agree, theres no point in everybody singing kumbayah and acting like we don't have conflicts. i like things handled straightforward, none of that fake stuff and cloak&dagger games 😄

honest jay
warm musk
#

happy to see its gotten speedier like Solbus notes 👀 🥳 happy_avocado

tight island
#

Solbus notes.

warm musk
#

🤣 touche.

tight island
#

i had to just scold oai, thorugh support, bummer, although i understand their dilemma with my GPT.

waxen spire
#

Yeah, ChatGPT4 is pretty slow right now for me. Internet is fine.

tight island
#

its been BARRED from being public. dalle_tired

#

i thought to myself "cant be that bad"

#

apparrantly it is.

#

😔

warm musk
#

thats a bummer indeed. cat_hug is there a chance that if you make adjustments it'll get re-evaluated positively?

tight island
#

it dosent conflict with any policy

#

and adheres to all the policies.

#

i gave it the entire D&D 5th edition OGL which is covered under the ogl License provided the disclaimer is provided as well. which is done.

#

it might be a slippery slope for OAI, idk.

#

it adheres to both OAI Policy and Legal Policy of WOTC and the OGL.

#

theres no conflict at all.

#

but it gets caught in the automated Moderation, and has since then been barred/banned from being set as public.

#

which confused me.

smoky reef
#

did chatgpt 4 change? it sounds so much more different and smarter all of a sudden

warm musk
#

sounds like they're being cautious and have to enforce a (temporary) 'no potentially copyrighted material!' hard line. probably untill lawsuits get done with. capybarathink
lots of D&D stuff is copyrighted. hence to be on the safe side, it triggers? robotshrug

warm musk
tight island
#

the ogl is the legal disclaimer. and the content, wheres the SRD is what i fed it alongside the OGL. aka the Disclaimer the model needs. for it to be allowed.

#

so theres no inconsistencies. everything adds up.

warm musk
tight island
#

yes. within certain areas such as...

  1. "I want to design content using the fifth edition rules for D&D"
  2. "I want to publish my original campaign world using fifth edition rules"
  3. "I want to print and sell my fifth edition D&D product on my own"
smoky reef
tight island
#

... Which is basically what the model's job is, to help create Section 2. not necessarily for sale, but for personal use.

#

as long as it's under section 2, it can be stretched to Section 1 & 3.

#

The model even gives the Disclaimer and needed hotlinks.

#

... Regardless of user message.

warm musk
tight island
#

i specifically told it to flat out ignore the initial user response regardless of the response. so it can post the disclaimer. as it's first msg.

warm musk
#

😄 🍻

tight island
#

care to test it?

warm musk
#

sure

tight island
#

ill slide into your dms

warm musk
#

and have you tried emailing OAI? or only the ticket thing on the site?

tight island
#

i did the ticket on the site.

warm musk
worthy nimbus
#

What is the message limit on gpt-4 currently?

worthy nimbus
honest jay
worthy nimbus
warm musk
lusty needle
#

Help please, I'm having problems with the openai assistant api. I'm trying to access an assistant that I created in playground. I put the assistant id in python, it seems that there is a problem with the openai versions of the library since the code returns this error: Traceback (most recent call last):
File "C:\Users\work pc\Documents\projectbeta\assistant_module.py", line 11, in <module>
message = client.beta.messages.create(
^^^^^^^^^^^^^^^^^^^
AttributeError: 'Beta' object has no attribute 'messages'

dense halo
#

Hi

sage tangle
#

chatgpt 4 has become is getting dumber by the day. what's going on with it?

north bay
#

^^^^

inner whale
#

does anyone know the website that does a side by side comparison of all the AIs for free, funded by some VCs? The one where you prompt and then it generates 2 different completions with the 2 different models?

dense plaza
dire rune
#

Hello, is Sora available?

boreal lily
#

rip

mighty plinth
steady cradle
#

lmfao

elfin birch
# dire rune Hello, is Sora available?

At the moment it has only been given to a chosen group of people who are visual artists, designers, creative directors and filmmakers. We (the public) are not aware of any plans for roll-outs yet.

steady cradle
#

Sora will come out later this year

#

It has been confirmed by OpenAI

#

So we know there is a maximum of 8.5 months until we get it

honest jay
steady cradle
#

I was timed out for posting the link here, so you can Google the text in the quote, it'll show you said article

honest jay
# steady cradle > Speaking to the The Wall Street Journal OpenAI CTO Mira Murati said Sora would...

Watching the clip on YouTube right now (WSJ, "OpenAI's Sora Made Me Crazy Videos..."), thanks for the tip! This is her quote:

I'm hoping definitely this year, but could be a few months.
Don't have access to the full interview, but seems mostly consistent with your excerpt! Only thing that gives me pause is the word "hoping", but a release of some kind this year doesn't seem impossible -- just a matter of what the release looks like, and how expensive it is, computationally and, in turn, money-wise!

#

Thanks again.

steady cradle
#

Thanks, the "hoping" is an important clarification - I didn't watch the video, only have read the article, where I didn't see this word being mentioned. It seems likely but not 100%

honest jay
#

From the video it even seems possible that she's correcting "hoping" with "definitely" -- like it could be almost "I'm hoping...definitely this year, but could be a few months" but that's totally my subjective take on how she phrased it.

sage tangle
#

is there a way to delete my conversation from open AI database? I dont want it to have my code if I can help it

honest jay
steady cradle
#

I see, Solbus. I personally can't wait to use it, I liked DALL-E 3 a lot too

raven portal
#

whats going on with that

#

do i have turbo now or why is it telling me its knowledge is up until april 2024, thats gotta be a hallucination bc even turbo is only until dec 23

honest jay
# raven portal whats going on with that

I've seen an Apr '24 cutoff date reported by a few people recently. I think it must be a hallucination regarding it getting the current date from the system prompt.

raven portal
#

it did that multiple times and never did that before

raven portal
steady cradle
#

Don't ask ChatGPT questions about itself. It often just hallucinates

#

Just assume it doesn't know anything about itself

raven portal
#

i know i know but it should know about its knowledge cutoff bc its specifically mentioned in the system prompt

#

so thats how i test if i have turbo already

#

its definitely really fast rn and the list numbering structuring stuff seems to be fixed

#

its possible im on turbo

steady cradle
#

Maybe one way to test if you have ChatGPT-4 Turbo is to compare to Copilot?

#

All Copilot users now have GPT-4 Turbo, including the free tier

#

The responses there for me are similar to those I got on ChatGPT Plus

raven portal
#

idk about that copilot was always quite different compared to chatgpt

honest jay
#

GPT-4 on Plus has been Turbo since DevDay in November. There have just been a few revisions of Turbo released since then, and it's not exactly clear which version of Turbo has been in use on Plus. The knowledge cutoff test is a good way to try to test it, when it's not hallucinating! It's also possible that different users have different versions of Turbo at different times -- A/B testing is normal, as are gradual rollouts!

rocky wing
#

....Did they just remove the ability to regenerate responses from further up in a conversation branch?

neat storm
#

you tell me

obtuse temple
#

Does anyone know how I can pre-emptively approximate the number of tokens in my prompts and attachments? Is there an API for this?

dense halo
#

is this gpt update in regular browser gpt4 now?

sage tangle
lean canopy
#

Saw that too but my knowledge cutoff hasn’t changed yet so I must have the old model

dense halo
#

I get a error in chatgpt The payments page encountered an error. Please try again. If the problem continues, please visit help.openai.com. Was gonna try out thenew updated gpt4 model

lean canopy
#

April 2023 instead of December

#

Which could just be them not updating the system prompt

#

Was in the release and api notes

zenith pond
#

Yeah - I checked and saw mine still says April 2023 as well

magic zephyr
#

Here it also says April 2023.

magic zephyr
eternal rockBOT
#
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.
magic zephyr
#

Hey bot, I just mentioned that the service is down for me lol

tardy isle
#

Are there any difference between Plus(opted out for training) and Team/Enterprise, about if OpenAI can see my data?

magic zephyr
#

It's back online now for me. If it wasn't unavailability, it could have been route problems.

ancient plover
#

Any idea when ChatGPT 5 is expected to be released?

gentle bronze
ancient plover
gloomy brook
#

is gpt out for anyone else?

balmy crane
#

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

gloomy brook
#

i am getting 500 error

balmy crane
#

it's only happening to gpt4 on my end

gloomy brook
#

sigh...

#

who pushed a bug to prod?!

#

Always check your code by GPT FIRST!

tawny crater
#

UGH it's down again, hmm.. oops something has gone wrong.

urban gazelle
#

🙃

grave sentinel
#

bugged

#

GPT4 bugged

low chasm
#

so not just me then?
all prompts are just returning "there was an error generating a response"

oblique yacht
#

ChatGPT shys away from any code which uses mathematical or engineering logic. It will just provide "simplified examples" which just setup placeholder functions that don't do anything