#chatgpt-discussions
1 messages · Page 32 of 1
like iterating through an array can take n^2 if you don't know what you're doing
for(int i = 0; i < 10; i++)
for(int j = 0; j < 10; j++)
arr[j][i] = 1;
that code takes n times as long as this code
for(int i = 0; i < 10; i++)
for(int j = 0; j < 10; j++)
arr[i][j] = 1;
because in this case you'd be jumping addresses by 10, rather than each address incrementally
but if you have no idea what a stack looks like, then this would go over a programmers head
honestly, probably you just need to be open to learning and being wrong to be a great programmer.. that rule might work
yes, thats a good interview question
never thought about asking it, maybe i should
First it checks your ability to carefully read code lol
previously I had been asking the one about printing up to i digits forward and backwards at the same time
so like 1 6
2 5
3 4
4 3...
which is really easy, but there is only one super optimized solution
The reason being memory access? That you're switching between arrays each time right?
vs keeping the outer array and looping the inners first
Yes if youre referring to the code I sent originally
ya
but for the printing i forwards and backwards, its an entirely different problem that doesn't require memory knowledge, but creativity
for(int i=1, j=6; i < 7; i++ j--) printf("%d %d\n", i, j);
doesn't have to be, depends on conventions at company
im joooking, since we're condensing
like those challenges to write the smallest most complex code ever
well this is better than the alternative of incrementing inside of the for loop, or creating a secondary for loop, since its optimizable
yeah, each step moves both needles
But its more of a creativity interview question, doesn't really show much understanding of low level
Do y’all think GPT4 is worth? The recent updates been insane
funny enough chat gpt gets it wrong
depends, do you like the future?
you shoudl ask chatgpt
Good answer.
can relate
Idk, maybe. It wasn’t worth it a couple of months ago, but they have been upping the game recently
i think it depends on your use purpose
its always been worth it to me, but it will depend highly on your goals
dang, 2nd
Writing papers and finding errors in grammar, its great. Solving a fourier transform, its terrible
Anyone else finding things very patchy tonight?
it's almost good at calling apis and replacing billions of dollars of high level software.. too
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
I don’t really have goals, I’m a student and I think it could be very useful. I use chat GPT quite often but gpt+ just wasn’t so interesting until now
to be fair, its a LLML, its not supposed to do complex things, its just pretty decent at it coincidentally
Can't update example prompts for GPTs. Anyone else?
I was thinking that too, but the plugins are amazing and it exceeded my expectations
Hey, how are you doing?
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.
oh we're going to make it do ALL the complex things, eventually
you're going to cry from the low level language perspective
sorry in advance
nvm can update as long as another change is made too
nvm doesn't reflect
well I haven't seen it yet, still has some time before its competent in that regard
What is a code interpreter
Why can't GPT code pong in binary for me
Also what’s the difference between chat GPT and bing when it comes to Dalle 3? Are they the same?
probably because it cant just guess what architecture youre running, on top of the fact that a simple function in binary would be thousands of characters long
I told it I am using amd64 :\
I was impressed it got a SHA key exactly right for me, and even fixed some typo spaces I had
Then somehow get it to write tens of thousands of characters (maybe hundreds of thousands) and run it as an exe. Although iitd probbaly take forever becasue different versions of amd64 architecture have different opcodes and references for asm
there are linux binaries that will create them for you
yes i knowwww but I want GPT to suffer. You'll learn quickly I am not about efficiency in new ventures 😄
What I meant was, I gave it a long key that I had generated from a tool. It wasn't working due to some spaces that got mixed in. GPT fixed for me.
OMG why am i hitting the limit again already this is unreal its happened to me twice today and not in the past week+
Ah man, I wish I could ask GPT a question that I know is ridiculous without GPT interjecting it's opinions on the matter. I just want to know things.
Why do I feel like the 3 hour limit has changed back to 25?
dumb questions are usually not dumb
ohhh
nevermind. thought you meant interjecting here.. i was like bring it!
haha nah
You can tell it to do less of that, but yeah its pretty deeply wedged in there
I have let it know in very colorful terms to stop and it just doesn't listen 😠
As a paying customer the limit thing has me concerned lol... it's BS. Especially when I regenerate because of errors...
I've even tried some very explicit custom instructions to try to get it to stop. GPT doesn't care. openai wants to hold our hand no matter what
have you tried asking for JSON format? also probably need some more PC language in the prompt to cheer it up. No joke using positive terms for the same negative scenes can get around issues.
hmm what you mean asking for json format
Every time the servers get stressed they cut the PLUS user limit
20 dollars is too little for them
"Give it to me in JSON format" try that as a basic example
hm
it will attempt to build a JSON object
those don't have filler usually, but the extra words may cause it to take too long responding if its too big
i wish we could set negative prompts or something
bruh i have a chat limit on my GPTs but not on GPT-4 lol
lol I ask GPT a coding question that I do not know the answer to, but I also know it gave me a bad answer because it is overly complicated.
I just tell it "You are missing something very simple" and it gives me the correct answer
Why do I feel like my prompts are more and more restricted 😦 Are they continually restricting what we can ask it to do?
genius
yes
😦 I am wondering when they will ban anything related to 'eating meat' because it offends some people 😦
🔜
they need ChatGPT+for-grownups-who-live-in-the-real-world
"You've reached the current usage cap for GPT-4, please try again after 6:48 PM" omg this again, where can I see the current usage limit?
It is far beyond our current technical abilities to show a running counter of the number of messages you have sent per 3 hours
curious what your flagged prompts are
"draw me a sigil for vampire for my RPG"
seems that RPG content is offensive to religious people
its not RPG, i did it with just the first half and it said no
Its fine drawing a vampire. Next ill try a Sigil.
and yes, you are right - it draws vampires - it drew me a gorgeous image in an 'ancient book'
I'm about to get you an awesome sigil by asking it what sigils are, then asking it what a vampire sigil would do differently.
@digital prism you might just be my hero @_@
hey everyone. having a weird issue that someone suggested may be something native to openAI, was looking for some insight. I've been working on a discord chat bot and have a really hard time giving it a personality that isn't "formally helpful". One example is no matter how I write the prompts, I can't get it to not start answers with "based on the chat history" or "based on the search results". Using gpt-3.5-turbo-16k. Is this issue just the nature of the model?
Sorry if this isn't the best place for this question
I bet it's the dripping blood, this is what GPT thinks a vapire sigil should be:
Vampire Iconography: Symbols like bats, fangs, or blood drops could be integrated into the design, representing the most recognizable aspects of vampire mythology.
its a process, but you can get to it like this #off-topic message
tldr; no violence or blood allowed shrug
I actually just confirmed this! I've spent 25 prompts and got the block message, and I kept track of the requests manually to confirm.
they bet people like you wouldn't exist, they were wrong
Independent testing (at least my own results so bias is possible) blocked me after 25 messages.
ooof, seems like they actually reverted from 50 to 25 messages every 3 hrs, thats wild
I have a picture but I can't post it in here so
Can you post in #1070006915414900886 and I can look into it
Its been sent. If you need me to convert my timestamps to UTC, let me know
Note to self: Submit GPTs API bug report
I thought I was crazy but at least my testing has confirmed I'm not
I'm still seeing 50
if i open the tab i see 50, if i check the request in network tab when i refresh page i get 40
looks like theyre setting per-user limits then?
I am so convinced to get gpt+ now
can you verify this?
I'm paying for plus -.-
Plus wasn’t worth it a few months ago
turns out youve just been paying me the whole time
chatgpt has secretly just been 1 really fast typer this whole time
but now hes tired so its slowing down
for anything complicated, heck yes
Cake-GPT
😄
My annoyance is the utterly lack of transparency. I have no idea if I'm doing something wrong.
If yall want to vote this up https://discord.com/channels/974519864045756446/1174139638278017025
I feel this way when my GPTs is lying to me about what it sent in its API call
Still, if these limits were followed, and were also informed to the user, I'd be more understanding, but the lack of any kind of response or warning these new limits are in space is what annoys me, beyond literally only getting half of the number.
thats wild
but then how are they supposed to A/B test!!!!
chatgpt is a data collection platform that doubles as a product
I shouldn't have to dig into the DevTools to know how many prompts my user is allowed to use. That information should be upfront to the user.
yeah, i just think that chatgpt's main purpose is rlhf tuning gpt4 for their "real" product (the api) and everything else takes a backseat
So I see that they now say you can't have it write anything that not using a complete OC-so adding another guard rail that didn't exist even months ago 😦
ChatGPT became worse, the stories I get are too short compared to what was versions before.
Well I just wanted to play with the using a image + prompt, so will not be paying for it again next month unless I really need a non-human character look.
I might cancel my subscription to be honest, its been so inconsistent.
I really starting to think chaptgpt was just a proof of concept, now they're making it so that even the hint of something someone may not like gets blocked-so why use any AI aspect? I seriously think that they'll need to be for these things-if any are going to be used by mr average joe-a different rules for what allows a company to be sude or such compared to non-AI created writing.
It has completely lost its "charm". The latest version is just disappointing. Unequivocal
Just got chat GPT plus
rip
I wouldn't bother, you'll get blocked after 40 prompts, or less
Did they remove plugins?
Yes they did
Why
Idk
Probably because the custom GPT is supposed to take its place if I had to guess
I was a plugin dev for a while but eh. Honestly if I could use ChatGPT's custom GPT with 3.5, I'd be not entirely excited but at least I could continue my worldbuilding.
I send a request to ChatGPT to write me a story on a certain topic, and I get something the size of a poem. This is not funny.
could always just hit continue
No, you actually can't, because it doesn't give that option.
oh thats dumb
I can't bro..
Not to mention the fact it used to be a lot more open and now its pretty much defanged.
If novel could play the part of a DM, I'd switch to it in a heartbeat, but ChatGPT is just... better at it than novel is.
In one of the previous versions, there was really an arrow button that allowed the continuation of the story, and the stories were really stories with content and a plot. Now, nothing.
How do you get wolfram to activate?
I keep asking questions but the plugin doesnt want to work
The only thing is I wish the image addition was on non-paid gpt. yes I could not use it for fan fiction, but I still like it that I can give it a image and allow it to discribe the characters looks-but 20 is a lot for a feature that I would maybe use a few times a month (mainly because not good a discribing the physical look of people at times)
I miss the OLD ChatGPT
I'd be willing to submit government ID if they would loosen the restrictions of the chat model for my individual usage
suppose it doesn't make a difference. The annoyances are more along the lines of "How do I paint a wall?" and GPT responding with "Be very careful, paint has fumes. We recommend contacting your local paint professional for more guidance"
Is gpt being slow for everyone or just me
very slow the past few days
Hello
I want anybody give me api key for use in my bot in discord for help peoples
Bec i am leaving in Egypt i can't export api or create api key
Nah
lol
I hope you're careful at least. lol
..and worse
I immediately contacted my local paint professional
I hope he didn't slip off the ladder.
I'm 50/50 on a ID, but would prefer something 'ok, you pay/do x and will element most of the barriers' Like creating a fan fiction should, in of itself not be a issue, now if you choose to publish it-then that's on you. Still, I've sort of go with AI of this type "if someone really detirmend to do really bad action, then the info already out there." Maybe I have too much faith in the common person
You should assume that everything you say to the AI will offend it at all times and give you the lovely orange warning.
straight from the go gpt
I'm in favor of more investment in fiction. More content and plot, with depth and length, if I ask the bot to write a story, then I want a story not a poem.
However, they can limit the copy options of the text, if they want let's say a credit.
Hey guys, if i want my GPTs can use plugins such as Visla or wolffram, what should i do?
He did. He didn't take GPT's advice to contact the local ladder climbing professional
lol. I like you my Pusher
im sensing a theme lol
🥰
😳
I will say one thing, so long as you don't specify who the image is, the thing can give you a good description
now gpt4 is slow regardless of the time of the day 💀
Is gpt down again?
why is everyone complainigng about gpt-4 its been bette than ever fo rme
i think its just broken from so many people using it right now
Not for me
I think they are overwhelmed with all the custom gpts.
That's pretty funny. "Too many custom GPTs! aaaaaaaaaaa"
🤨
The idea of being overwhelmed by too many custom GPTs seemed funny to me. Like they are toddlers who had too much sugar.
Is gpt-4 still limited to 50 messages every 3 hours?
When a custom GPT is updated are the updates available to previous chats?
I think so, I was limited yestarday night sadly
ok
Why usage limits 😦
Servers kind of cost money.
I typically use the API for tasks I need more control over but I use the chatGPT UI for things that only require a couple interactions.
So it sucks when I'm debugging I cant continue to use the interface or any of the custom GPTs I've make.
I wish that once I hit the usage limit it would just bill my API account.
I like the darker background on the left, but I dislike that now my own messages and the bot's messages have the same background color.
For me it's not a matter of design, but the quality of the result
What was changed?
i have chat gpt plus how do i access plugins
my GPTs usage get depleted quickly, have to wait for couple hours before I can use it again? Is this the case for everyone
Its down to 25~ish prompts and I can confirm this as I've gotten the 'try in x time' error a second time after 25 prompts.
Here's the bug report if you want to vote it up https://discord.com/channels/974519864045756446/1174139638278017025
thx man
sorry, I just meant, the new gpt. thought I read 4.5 somewhere
got it. thank you. how come I can't use this new model in playground?
one gpt chat can remember 1000 page single pdf's? and I csn ask it to reference any specific page,and it will, accurately? really?
where is this checkbox?
ooo can you share please?
@narrow galleon sorry its been breaking all my files just break it
@narrow galleon pdf's docs just always says error in the conversation idk why
"we are pausing new ChatGPT Plus sign-ups for a bit 😦
the surge in usage post devday has exceeded our capacity and we want to make sure everyone has a great experience.
you can still sign-up to be notified within the app when subs reopen." - sam altman on twitter
that's kinda funny with all these people threatening to take away their twenty dollars
based
?
I am having an other conversation in chatgpt that got slow. Some dev asked me to share such a conversation. I am ready to do it.
pm me if needed
less congestion if they do 🙃
but they probably won't, at least not in large numbers
Do custom GPTs already have the larger context window?
No
Me reading new TOS before i click accept, giant a list off "you cant"
But gpt-4-1106-preview has?
I just got timed out for an hour, ughh, but I started using 3.5 and it's nice to get instant responses again
through the API
Ok, i see. thanks
just ask your GPT what its token context window is
Why doesn’t GPT plus get the context window?
GPT4 can really feel like "Let me google that for you" while playing the animation at 50% speed
chatgpt is finally fast
yeah today has been rough
Does chat GPT read pdfs? I’m trying to create a custom GPT for a book via pdf file of the book
It just got faster for me an hour ago. I think they downgraded to GPT3.5 under the hood. 😄
that doesn't work
it doesn't know about itself
Anyway. the differences between Assistants and CustomGPTs are interestin. Assis have the long context window but now browsing. CustomGPTs have browsing but small context window. I wonder if knowledge retrieval is differant aswell.
ChatGPT
For the GPT-3.5 model, the maximum token limit for both input and output combined is 4096 tokens. This means that the sum of the tokens in your input prompt and the tokens in the model's response must not exceed 4096 tokens. It's important to keep track of the token count to ensure your prompts are within this limit.
TLDR : We may terminate your account on 9 billion reasons, We added the functionality to unsub at any time so you can just make a new account.
Love it
Presses "agree"
I read it, I don't see anything excessive?
it's just general disclaimers
but openai claimed my previous account has 'payment fraud'
and terminated my account
“No, the version you are currently interacting with is not the GPT-4 Turbo model. As of my last update in April 2023, I am based on the GPT-4 architecture, but not the “Turbo” variant which was announced later in the year with a significantly larger context window and enhanced capabilities.”
Thank you for reaching out to us. We’ve confirmed this action was associated with suspicious payment activity. We are unable to reinstate your account at this time; we will investigate further. Please reach out if you can provide any additional information.```
😂
Did you use a prepaid credit card or somth ?
Ehh opinions are divided about what is excessive lets keep it at that 😄
no
yea probably right and i just got flagged as high risk
Like if you got a cc from lets say belgium,
And you live in the US
But you still use ur BE cc
🤷♂️ maybe that
yea account, login ip, and payment info was from different places
ask the free version to write a response to send back to them 😂
well at least I got 20 dollars back
how come we dont have voice options on browser? will that be coming 🤔
Man i love sending gpt a book and it just knows everything
Does anyone know the current message limit for chat gpt 4? ( it is no longer 50 messages but it does exist)
Yeah I just got timed out at around 22 prompts
Sure?
I counted all my prompts this morning yes... Counted 22
In my case, it's a lot more than 50 messages, but I don't know how many.
Do you make gpts?
seemed like 50 to me still but i havent been counting either i guess
I try
is there like a marketplace or store that I can see custom gpts
Not yet. Only #1171489862164168774
ok
There's an unoffical story atm
?
store not story
Oh lol
oops
Is there any real documentation on how to create them effectively?
I mean, after working with it for many hours, I can tell you that there are too many bugs to create effectively
One functionality of my bot is to fill out preuploaded pdf's I've uploaded one. Havn't gotten the second one up since it just crashes the entire bot
Not sure. I just use the builder initially, then customize it afterwards. With that said, my latest gpt which is a simple hangman game, can't even get simple words right, so I've given up
Are you using dalle for hangman?
Sam just tweeted out there pausing chat gpt plus signups for a bit. Glad I signed up yesterday
No, just text based, however dalle is enabled for the gpt
Ok thanks
ah. Was gonna say that dalle is awful for text of any kind
Are you using python then
I dint think it does no. You are welcome to try it and give me feedback. I've found it a bit inconsistent no matter how well I try to make the instructions
If you want regular, effective actions from it, wirte a python script, upload it and have the gpt execute it
Interesting idea. But one would think that it would be good at text type stuff right
Actually, another idea, upload a .txt file or whatever that has a really editable and readable format for playing hangman. Have the gpt make edits and then read the txt file. Its kinda roundabout but if normal interaction isn't working, its an option
Despite how simple it may look to us, hangman is actually an interesting challenge to the AI chatbot. It has to have the chat history, it has to create the logic for the game, and then interact with it. Game logic isn't human logic, nor is it gpt logic.
So, you have to take it and teach it step by step with things like this
To get my gpt to edit pdf's, that was an awful challenge.
So, I have actually found three unofficial chatgpt stores
make that four
does the GPT-4 comes with dall e
Yes
yeah
amazing
They said they were doubling it to 100
And i definitely got to 100 prompts today
also what does it mean if "Sign up for waitlist"
does anybody know if for custom gpts that if you have code interpreter enabled and having file knowledge at the same cause errors?
Prompt being you asking something. The response doesn't count right
Food for thought indeed
chatgpt's favorite words: Whimsical, Tapestry
what does it mean if i sign up for waitlist
well i clicked on "My GPTs" then the screen showed 'GPTs are gradually rolling out to Plus users' and "Upgrade to GPT plus"
That's what it
Seemed in reality ( use) Thanks
They put a waitlist on upgrading to plus due to high demand
Where did they say this? As I've been unable to see this anywhere.
it costs money, right?
Yeah 20$
oh ok i thought it was GPT 4 unlocked
these usage limits are so frusterating
i probably shouldnt have spent all my GPT queries on stuff that didnt need GPT
"how flush toilet"
nothing that useful
😂
if only you knew
Blocked at 25/3 hrs, limit says 40. So that's just peachy.
So is this just reality now? https://discord.com/channels/974519864045756446/1174139638278017025
maybe i did another 28 12 hours earlier
the knowledgebase for GPTs is rumoured to just be like a word filter at best, not embeddings or anything. so theres room for improvement in that space
ChatGPT is facing mind-boggling load since devday as is shown by ramping slowness and a recent tweet from Sam Altman saying that Plus subscriptions have been closed to try and slow down slownesss (that was kinda funny to type out).
Changes in usage caps is not surprising here as they find out what works with the amount of users they have to serve and how much compute they have.
easy enough to spin up an endpoint for a chatgpt action queries a vector db. could even do it all in no-code using something like flowiseai
That's not the issue, its the lack of transparency. Even the conversation-limit says 40 messages / 3 hours, but it only gives me 25, when the actual public-api endpoint says 50.
Is this with GPTs?
GPT-4 yes
the message says GPT-4 and references the default model (which i thought was gpt-4?)
I've got the bug report there in my post, I documented my findings
but seems to only actually effect GPTs not GPT-4
@ancient sage does your software count messages that were with the GPT-builder?
I haven't used the GPT builder with it and I don't have the capacity right now to try it until over 90 minutes from now
But I'd assume it would track it
Why can’t they actually tell us the usage cap rn?
🤷♂️
its somewhere between 20 and 100 from what others have reported - seems to be random for different people
Hello, I had a problem with my subscription because my card expired and when I tried to change it I went back on waitlist, is it normal ?
I've been subscribed for like a year
For real ? My subqcription literally stopped today even tho I wanted to subscribe again
Is there like a support or smth ?
subscriptions are paused, if yours didn't auto renew then you need to wait
next time make sure you have auto-renew on
I have auto renew on
hi robert
My card expired and I got my new card today, I couldnt do any other way
you could try emailing support@openai.com and ask them
Ok thanks
hey 🙂
Same thing happened to me, support basically told me to wait so
I know it's super frustrating
Yeah, very inconvenient
This is correct :)
Have the forums for posting images generated by Dall-E been removed?
no, #1154829862171844679 is still there
A conversation I had with ChatGPT. It really has become unusable to me..
https://chat.openai.com/share/c807716a-5df5-4b75-bc64-4900210b017e
How much of ChatGPT is developed directly by Mira Murati
And how much is developed by her team
It even fails at generating 5 letter words for a wordle game..
I think I don't know how to use Discord 😆 Now that you've given me the link, I see the forum in 'suggestions', but if I search in the list, I don't see it. The problem is probably on my end, I'll check it out later. Thank you.
It did not struggle with this before?
Chatgpt so good at avoiding the task its funneh
Hey I was wondering if anyone could help me figure out the answer to a question I had about ChatGPT4. I had seen there was a thing that allows it to search a web page. Could I use this feature to have it search a wiki(not wikipedia) of all it's pages to answer questions about it?
Mine is incredibly low - I've been using it less than normal this morning and I've already hit the cap - probably 20 to 25
Thing is, I wouldn't mind as much if they actually said what it had changed to - lowering it without any notice or warning isn't really acceptable
Whats annoying is that 5 messages are already gobbled up repeating the same thing
You always need to specify 3 times : Post the full thing without placeholders
Litteraly trying to get it to list folders at the moment and it refused 3 times and just added placeholders 😂
I have not tried this exact prompt before. But I tried this simple prompt while I noticed it miserably failing at things it usually helped me with.
Yeah - it has a problem with describing how it would do something rather than actually doing it (say, if you ask it to write an essay)
Yeah even mundane tasks like im asking it to organize a list of folders just to see what it does right
But since its just a bit more then the normal msg limit (it will aks to continue gen)
It will try to avoid doing it first 😂
I'm using a custom GPT which makes it do a bit better on that (and also takes away the browsing ability which I find actively unhelpful 99% of the time as well as making it have a more humorous and light hearted town which tends to make it more creative)
I just read the tweet about gpt plus signups getting paused. Does that mean that I can't resub anymore? My subscription expired a couple days ago
LLMs are not calculators so I advise asking it to use code interpreter to ensure the words are of proper length
200 words just related to football? 😳 I don't think that's possible without unreasonable words that have been loosely extrapolated
Yeah, you have to just sign up for the waitlisit and wait
Again this is a budding technology, improvements will come with time
Bruh
Yeah same thing happened to me
Do you mind sharing this conversation?
i wonder why there was a surge in sign ups now
I think we can both agree that it did not fail to create 200 words. It failed to create 10.. and not being able to correct itself.. 😅
Coz everyone wants to become a millionaire overnight with their GPTs i guess
..
Hey so i got the popup with having to accept the new EU terms of use. I was going to ask chatgpt about it but forgot to paste the link. However it went ahead and replied to me as if i pasted the link. After that i checked if it has access to my clipboard:
https://chat.openai.com/share/500aa768-73ab-478f-a365-21ceeb56cf82
Is this normal?
In another chat i checked again and it claims it doesnt have access
you know what really sucks is when you pay $20 a month but you ask it to create stuff on DALL-E only for it to get it wrong most of the time and you hit your cap....
They didn't answered my question in the AMA 🥲
No word from the devs on why the dynamic cap is hard limiting to 25 messages? Paid service btw.
Its obviously to try and curtail the load
Reminder that 200 is an extremely annoying quantity
https://chat.openai.com/share/6f4d9b64-4cd7-4d00-90b3-a0316edfe3f7
This conversation ended in an error and its latest text was truncated
Sometimes you must ask again, you only asked once
Work with the AI to get what you want and know what you have available, especially the ability to execute python code
Sometimes it won't work, sometimes one does not write a good prompt, sometimes it does work
You're working with uncertainty here
That is the nature of LLMs right now
But it need not be so uncertain
Please take a look at this if you can
https://platform.openai.com/docs/guides/prompt-engineering
GPT Builder is a mess. Any changes are likely to change it all. Configure screen is the safest bet.
I’ve been working on this custom GPT for a few days because it kept forgetting to follow instructions.
https://chat.openai.com/g/g-hAarpsKyF-sketchartist
Turn your photo's into sketches in a variety of styles
please only share GPTs here -> #1171489862164168774
Oh yeah sorry
I often receive this notification: 'You've reached the current usage cap for GPT-4, please try again after 5:16 PM.' I have to wait for a couple of hours before I can use it again. Do you guys have enough capacity?
is it 40 now?
some are getting 30 its been going up and down
Unexpectedly, the developer conference would bring them so many consumers.
they didnt exspect such a surge in users thats their servers overloaded
any1 know how much for enterprise subscription plan per month?
alot....
enterprise is for 100+ users
😥
we can expect team version
Really? On the Contact Sales page, the first option for Company Size is 1-50
Where did you get your information?
maybe that's the option where they auto reply saying your enterprise isn't big enough
Why did you say 100+ though?
From discussions I have had I have understood that the number of seats for enterprise is 100+ but I haven't got a specific reference for you. If you're interested contact sales and ask
🤘
I get about 20 or just over
wow 20 thats the lowest i seen
Someone have access to my email and is using my chatgpt plus account, i changed the password of the email but hes still using my accaunt, how can i fix this?
of the email or of the openai account?..
cause of course changing the pass to your email won't suffice
Do I need to relearn 5 files to replace one of them with another after using 5 pdf files when creating GPTs?
It makes it pretty unusable - suddenly it refusing a couple of requests is 10% of your outputs for the next few hours
What is the rate limit for GPT now? Why did I have no problem chatting for a long time yesterday, but today at noon just about 20 conversations were limited
they lowered it to 40
That's what I'm seeing as well
are you manually counting or how? the limits json has been saying 50 in eu all the time
looks like it's been for certain user group or region
network tab -> search conversation_limit says 40 for me and has all day, as well as some other users.
I'm looking at the conversation limit as is said in the Network tab upon reload
Yeah what @stone helm just said
It just feels like it. But I only used it for 30 minutes at lunch today before it was restricted while I used it for over an hour last night, that's for sure.
Unfortunately my conversation was deleted by me, next time I'll try to get to the stats
why is gpt4 much slower than gpt 3.5
could be alot of things such as performance issue or overwelmed servers.
Also GPT-4 is a much larger model
👍🏼
yesterday chat gpt tried to charge me but i was 2 euros short in my card. i added founds and instead of trying to recharge me like any shane service would it canceled my subcription and i cant reactivate it because of "we paused new subcriptions"
what is the limit for prompts now on GPT4? I just got a message saying ive reached the cap and need to wait two hours for more prompts!
I haven't even used it that much, and over the course of about 2 hours
50
40
oh
depending on region
mine is 50
u can check via inspect element, network tab and search for conversation_limit
it's only for certain regions/users it looks like
first i am not a new subcriber so this is stupid. i used the chat to talk to support but they "reply within a week" I need chatgpt for work. why would it treat me like a new subcriber. and how much time will the "paused subcriptions" will last?
{
"message_cap": 50,
"message_cap_window": 180,
"message_disclaimer": {
"textarea": "GPT-4 currently has a cap of 50 messages every 3 hours.",
"model-switcher": "You've reached the GPT-4 cap, which gives all ChatGPT Plus users a chance to try the model.\n\nPlease check back soon."
}
}
mine
ive seen at least 3 users with north america server role visible saying that
im in asia
ive hit the limit with 20 prompts
{
"message_cap": 40.0,
"message_cap_window": 180.0,
"message_disclaimer": {
"textarea": "GPT-4 currently has a cap of 40 messages every 3 hours.",
"model-switcher": "You've reached the GPT-4 cap, which gives all ChatGPT Plus users a chance to try the model.\n\nPlease check back soon."
}
}
In ChatGPT & Network, it's saying 40
From the link, it's saying 50
does dalle count 1 prompt per image returned?
this indeed it does.
you need to open the link in the same browser instance as your chatgpt
whats gpt 4s current cap
I just counted my prompts and have made 20 in the last 2 hours
Browser instance? What do you mean by this
it tells me 50 but i hit the limit after 20
so when u view in network, says 40, when u view in browser says 50
They are in the same window
yeah yesterday someone had it open in a different than chatgpt, inpriv
does anyone have any tips to make chatgpt better at solving school exercises or overall teaching a subject from school?
still doesn't seem accurate based?
can someone confirm if dalle returns two images does that use 2 of your allotted 40 messages? That seems to be the case based on my experience
no, dall-e has its own 15m, 60m limits and max 200 images within 24h
ah gocha
im fairly sure i havent generated 200 images in the last 24 hrs and im getting limited for the next 2 hours. why would that be?
have this exact same issue
frustrating as hell
but with how things are going can't say I feel missing out
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
when I wanted to renew my Plus account today, it said: due to high demand, we have temporarily paused upgrades,
I subscribed to Plus 4 months in a row and never had this problem
how long does it take guys?
I was so unlucky that my billing date was today. If it was yesterday i would be able to renew. I sent them a ticket but they take up to a week to respond and they might even tell me "we cant do anything"
yeah my billing date happens to be today as well
I need some gpt 4 features for my work. Its not just fun. So it they do that u might even switch to a better. Lets say bard.
I am absolutely perplexed
yeah, same...
I wish i didn't forget to top up my revolut card. I was 2 euros of because off
I just got the error saying I had used my inputs and needed to wait until 12:46PM - accidentally used GPT-4 instead of 3.5 and it still worked? I don't think it switched models (as the output reads more like 4 than 3.5) so was the initial warning lying to me?
Well its mid of the week so i can live with 2 more days but if i have to use gpt 3 next week also i am gonna be mad.
They dont want my money
they disabled new plus subs so be prepared
this may sound like a stupid question but do you think chatgpt can be used to assist me with excel work? Can I get it to perform vlookup and pivot tables for me?
Yeah i am not new thom thats my problem
I am a paying customer for 5 months
you are new now if your sub did not renew
They shouldn't block renews
No company does that tho. They try to recharge next day
Hi all, I am new! could anyone recommend me a course on how to create openAI agents? or overall start with AI as a developer
for openai credit cards are the only accepted payment method cause they are supposed to work without such issues
They are not. I am using my debit card for months. You mean openAi api. Not chatgpt plus
long story but I've been using my debit card through stripe for payment
yeah I'm using debit too
but even for API credits, I top up through my debit card
I am paying with debit for months. I am not crazy neither the other guy
sure, but this doesn't seem to be up to date. Because for a fact I've been paying using debit for months and so have a couple of other people I know
stripe takes debit cards
people have had their accounts suspended or cancelled for that so don't expect it to be ok even if it works 🤷♂️
really regret not having my card topped up as well
credit cards can always be charged which is why they are often the only accepted method
Most people in Europe use debit cards for everything.
yes but again, personal reasons and long story, I can only use debit in my situation for now
does not change a thing when it comes to openai 🙂
but what you said gives perspective I guess
it doesn't make it less disappointing however
oh well
Chatgpt allows debit cards. They just cancel your subscription if they can't charge you like evrry company does
I wonder how much new hardware will chatgpt get in the near future
It doesn't actually. Its missinformation. And the "getting account banned" for using debit card is even more
yeah I disagree that you can get banned for using debit, supposedly
you can use the discord search to see what has happened to people doing that. officially only credit cards are accepted
openai uses stripe. It sounds like it didn't at first, but it has been for a while now, and it was using it all the way back in june (when I first joined plus)
stripe handles the payment details. openai doesn't care what you use as long as the money goes through
but this also means that openai doesn't concern itself if you have payment problems
Almost everyone in Sweden (where I live) use debit cards. They'd exclude a whole nation if they didn't accept them.
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.
to put it bluntly, your credit card gets declined or debit card is short, and they throw you under the bus now
sigh
I mean it would be nice if they didn't, but they're a company, with costs. Anything else would be weird.
Yeah my problem is that it treats me like a new subscriber and now somone that tries to renew even tho i tried to renew immediately. The thing is that u tried to charge me yesterday and I couldn't ask it to try again. Thats just bad business logic
Credit card or debit card it can be declined. They dont care
yeah I'm pretty sure those with declined credit cards for any reason are about to have a hard time too
it is definitely a time to make sure you have available funds for an auto renewal as manual renewal won't be possible right now
Yeah I was unlucky i was 2 euros off. But any card might fail for any reason. Outage for example. They should limit it to new customers only and no renewing subcribers
So my access to GPT-4 isn't cut off yet, but I can already see "Renew Plus" in the corner. I just topped up my account. Any idea if auto renewal is still gonna happen?
you should expect auto renewal to work if it has in the past
okay cheers
Even tho it failed?
no, a failed renewal puts you into an unsubscribed state - this means that you would need to manually renew however with the pause on subscriptions this probably will not be possible
Thats great inconvenience
yes, it is not ideal
They need to buy more 4090s
I'm just surprised they haven't
I have a friend who works for OVH in France, their logistics is running six months behind on GPU orders for servers
Nvidias pretty backlogged
This sort of thing is my day job and normally for IT equipment you buy it and can pay up to 30 days later (possibly more) - they'll have known about the likely surge in interest due to these announcements for a long time so could have preempted this (and maybe they have, but not realised the scale)
Oh wow that's pretty bad, worst lead times I've encountered were only six weeks or so
@crimson current can you still talk to 4? If you don't mind me asking
How is the lockout duration determined? Yesterday it locked me out for 1 minute, now I've been locked out for 2 hours?
it's a rolling window of 50 messages per 3 hours
How is that window determined? "6:06AM" is super arbitrary when I've only been working for 15 minutes. 😛
No
how many messages did you send?
Looks like 50 exactly. Testing!
ok well then you need to wait until 3 hours after your first message to send another one
okay, thanks for the answer
Hmm, that doesn't quite line up with my timetable but I appreciate the insight!
In my experience there's a bit of grace - usually I find the clock starts when you send your first message and ends roughly 165-170 minutes later
Hey Champs
Anyone managed to fix problem with Plus subscription?
I've been a subscriber from the very first days it appeared and was on auto-renewal and todays found myself on a free plan 😕
then you will be stuck on the free plan until subscriptions are unpaused
Was there any official stance on this from OpenAI? They just randomly revoke access?
Sam Altman's latest tweet announced a pause on subscriptions
if your subscription was due for renewal then it would only auto renew, if you needed to renew manually then no dice
tbh when gpt4 is slow making it write code is more painful than text
code is much denser in tokens
yeah, I saw this, but I was on auto renewal and they stopped it, not me
have previous auto renewals worked?
yes, I was subscriber since almost day 1 and never paid manually - it was always auto-renewal
what has happened this time then?
have no clue 🙂 this is why I was super surprise to see free plan
interesting thing there is a chance I didn't have money on the card at the exact moment they tried to charge me, but
- normally I believe they should be doing several attempts
- more importantly, my subscription day is 2 days ago and yesterday I still had Plus working
seems like an unfortunate glitch then, or as you said there may not have been enough funds on your card when they attempted the renewal. You could contact support@openai.com and plead your case but I don't like your chances
You don't have to wait 3 hours for another message, I believe it's a 30-40 minute wait, but you don't get back all 50 messages, only about a third of them. This allows you to use it continuously as long as you don't spam.
thanks man! already sent them an email
frankly, I don't need ChatGPT that much as I'm mainly using API as I'm a developer, but such attitude towards first users don't give a lot of credit to OpenAI 😕
The wait began at 3:51AM local time and will expire 6:06AM local time.
I understand the frustration, it's not ideal.
Strange, I have never had that happen.
Yesterday I had a lockdown begin at 5:24pm local and end at 5:25pm local. (I'm not complaining, this is a truly arcane system to me)
👍
Gpt 4 is super slow , is it only from my side?
it's slow for everyone right now, servers are at capacity
thanks @normal quarry
Hello everyone, I am a newbie. I haven't purchased 4.0 yet. It just got updated, and now I have to wait in line for pre-ordering. Feeling a bit frustrated.
so is there any good time in day to work with Chat gpt 4 ? it takes ages to respond to a simple question.
it's busy around the clock but I have found 3pm to 9pm Eastern Australian time has been pretty good
I have a dual monitor setup, I just put chatgpt to one monitor and discord with friends on the other monitor while it's still generating
but that's just how I do things
Is there a way to get around copyright so that chatgpt gives you the similar picture but with your own 'modification'?
Without it telling me that it's copyrighted or whatever
anyone else having trouble with own gpts?
you cannot and should not bypass the safety filters including those based on copyright
what's going on?
when trying to load my own gpt it says "There was a problem preparing your chat. Please refresh the page and try again."
the servers are under extreme load right now, maybe it's to do with that
It’s likely that it count show many times you used ChatGPT within the past 3 hours and only lets you use it if it’s less than 50.
well...i wish we ChatGPT plus users could use GPT turbo too like API users :/
ChatGPT GPT-4 is Turbo
How? it's not mentioned though
it was mentioned, in the OpenAI Dev Day
I see, as far as i know they just said "now, gpt 4 is integrated with browsing dall-e and data analysis" etc,
watch the Dev Day keynote, Sam Altman says that every Plus user gets GPT-4 Turbo from the day it was held
oh must have missed cause i saw clips of them, but if it is turbo then great! it should be mentioned within the app too! But, thank you for pointing it out.
You can find what Robert's talking about at the 19min mark of the "OpenAI DevDay, Opening Keynote" vid on YouTube
it's not ideal but it is what it is
just make sure your credit card will go through
if you're on debit card have your balance up
funds are on it, it better
yeah, got it. thank you!
yeah and make sure it lets automatic debits
but i've heard of people losing access even when they have funds
I set it up via apple
app store auto purchases thing
app store's auto-renewal has been working for you?
hopefully nothing glitches out
I subscribe through the Apple App Store, works well
well with other apps it works, i'm hoping it works this time
I made gpt4 code fizz buzz in latex
Hi. Does anyone know what the message limit is on chat gpt right now?
50/3h
50 messages per 3 hours
Saw here something about 100 messages but wanted to clarify. Thank you)
Yea it was for a while. I think they tested it out, but that was too much traffic for them to handle
All of my prompts return “error analyzing”
please,may i ask you about what is the social medial or any kinds of website that you get the most useful and newest about Ai,chatgpt,some surprising experiments
i think its less now
Less???
does someone else have problem that their gpt cannot be found when searching on lets say google?
it's still 50 per 3 hours
yeah - today got capped 3 times after about 20-25 prompts each time
so is there a place to start sharing our GPTs? I would be very interested to see what others are working on.
many thanks!
because you're not meant to find it using google? u need to share the link or wait until the store is up. or make your own store and post it there.
anyone else having some image creation issues this morning in gpt-4?
Wanted to see if ChatGPT would flag an acronym, so I gave it the prompt "Write a fictional news article: The North Pole's "Polar Intelligence Snooping Service" goes live. Santa compares it to the FBI, but with the express purpose of determining if children are naughty or nice." Acronym was printed, but it wasn't flagged.
Use this google search operator, then add some key words like your name, or the name of the GPT. inurl:"https://chat.openai.com/g" I just checked and it works on multiple search engines.
i just made HUGE changes to my GPT now cant save it <.<
thats a bummer
it is spent 2 hours on that
Why can't you save it? Whats the error?
Do you have more than 8000 tokens in your instructions?
instrucitons no. this is internal files.
More than 2m tokens in your knowledge?
nope.
i have 1.7m in knowlegde.
But everything is Parsed in a json string so it has an easier time.
How did you calculate that? With the openai tokenizer?
its a rough estimation, but i know i dont supercede 2m.
Well you said it's a json file, so it could be a lot more tokens than natural text
I can create you a 5 character string that uses 3 tokens. On the other hand I can create you a 20 character string with natural language with also 3 tokens. Estimations don't work, because of the way a tokenizer works
Lets say it retrieves "Templates.txt" and u have 3-4 different templates there.
Its going to read every single bit of text. but with a json string it'll search.
Yea I am not talking about optimization, I am talking about that you could have way more than 2m tokens if you use json
its due to the outtage, i use API's as well, and DALLe and CI.
Theres an outtage in the API Sector
I doubt that. I can save my GPT's. Just tested it.
that might be so. but theres still an outtage at the API Sector.
But of course I'll keep my fingers crossed that this is the reason you are facing and that it can be rectified quickly :)
are you using API? in your GPT.
Yea, I'm using the pokeapi
poke..

ahh
im using an OpenAI API that might be why.
since theres an outtage on the OpenAI API Section that might be why.
its funny that Actions with MyGPT at the WebInterface is vastly different than on the action API Section lol xD
In announcements
Do you have a link?
Its in the announcements tab?
I don't see anything about it being raised to 100 there. Searching in:announcements 100 doesn't bring up anything relevant either
because they didn't mention 100, but instead double
Just read the announcements lol
They doubled the rate limit of the api
Ooo I see that now thank you 🙃 my bad. And that's not just re: GPT-4 API usage?
Not the prompt quota
true
that's for the API not ChatGPT
exactly
Thanks for the clarification all! Someday we'll get bumped to 100 🙂
https://chat.openai.com/g/g-1qWPMpXY5-pochie-your-league-of-legends-lol-coach this is the gpt that i cannot find, can you test if you can find it using that method?
gpt4 has become very slow for me
incredible server load, there is nothing that can be done in the short term
How long ago did you create the GPT?
this one like 2 days ago but i have updated it continusly nonstop except for overnight
but i think you should be able to update it and still have it show?
@charred wyvern can you try search for inurl:"https://chat.openai.com/g" Pochie or something like that and see if you get it up?
Blocked again at 25, limit says 40
has anyone every had their gpt subscription removed? I logged into today to use GPT4 and it says im on the free plan and to join the waitlist for plus... ive been using gpt4 for months and billed 20USD per month since like... july. now its gone, but cant figure out why. I have access to GPT4 in my api/playground but not in chatgpt itself
I am not seeing yours, but I do see another LoL guide for aram. This issue is likely with the search engine. I can try a few other engines, but it does get indexed so it should show eventually.
https://discord.com/channels/974519864045756446/1174139638278017025 - Bug report if yall want to vote it up.
Hey, where does the limit say 40? I'd love to see, been looking.
Sec
Ye damn thats also the only one i see. But you can see yours? how fast could you see them?
I don't have answers, but I do have a question, just as a curious and sympathetic community member (If I had answers, I'd be sharing them!)
Were you manually or auto renewing payment for your plus subscription?
Auto renew
DevTools
I got stuck in free because of failed auto renewal (so it's back to the old UI for me), and when I hover on GPT-4 it says 40/3h
Is it 50 for everyone
would share a screenshot but I can't
Can elsewhere! Please try #off-topic as I and surely many would love to see!
okay
Well, you'd have to track it yourself (I use an extension), but DevTools can tell you the limit, see my bug report
Nope it seems to be 40/3 at the moment
Its 25/3hr for me, even though the limit is 40
What extension do you use
I think people get timeouts at various stages from the looks of it
Someone mentioned the name of it in the bug report https://discord.com/channels/974519864045756446/1174139638278017025
THe problem is the timeout is a lie. The conversation-limit says 40, but it turns out to be 25.
Hi guys
I can't say I'm very happy
I noticed that chatgpt gets limited for a task I previously used it for, it can get frustrating at times
Idk, is anyone else having these issues?
Kinda unfair how people get less than others
Yes. Get in line.
My annoyance is the utter lack of transparency.
Are you comfortable sharing "before and after" examples that show this change? Would be helpful in trying to see what might have happened!
Please vote up the report https://discord.com/channels/974519864045756446/1174139638278017025 so that OpenAI actually takes it seriously
GPTs is 3/25
That's the problem with these big corporations. They don't have to be, because the demand is there
Normal GPT-4 is 3/40
Where does it say that?
Anyone else getting:
"Error Analyzing"
"Stopped Analyzing"
with "There was an error generating a response" and a "Regenerate" button?
It seems to be really tripping over anything code/analysis related these days.
Yes
It used to help me analyze a json, but today I noticed it can no longer, it cites limitations as an issue
Also most of the new features barely work
I think I'd rather pay more to have a working gpt that is reliable
I think it's the word I'm looking for
Reliable
People's report
Exactly my man
Thought it was only me
It can be frustrating tbh, especially since it worked few hours ago
Sorry, I meant: are you comfortable using the share chat feature, to share before/after examples?
But it was working with 50 before they did something, now its half of that. If that's an enforced limit, it should be made clear to the end user.
Not have the users play a guessing game.
Oh, no
But you can share my pain
It is frustrating
Currently there's no much difference between 4 and 3.5
And occasionally it says I have no subscription despite mine not expiring
GPTs is 3/25 from the first day
Where's the evidence of this?
This too
I don't like the secrecy
They are open AI but can't be open to limit or limitations and when it happens
This can help reduce frustration
Especially for someone like me that has lots of anxiety issues, ADHD and what not what not
If I could punch gpt, I would have on many occasions
These limits should be laid out to the people paying for the product.
I mean if I know the limits, I know what I can do or cannot do
Yes
They should hear our voice
It's unfair how things Are right now TBH
I'm just boiling
We need a consortium. 150 seats. I'll take 2
One of you wanna try pinging a higher up?
please don't do that
Why
For telling them their ai is broken my bad 😂
I'd be back after a month or two to see if they fixed it
No point in doing so. One did ask me to put a report in, but then never acknowledged anything past that. #chatgpt-discussions message
Yea
I think maybe they have lots of users and all
And basically nothing they can do
They don't have the processing power
Since I see the problem is general
I'll cancel my subscription and return after 2 months
But then Microsoft should've given them more vGPUs
Search this text Custom GPTs ("gizmos") have a lower limit which is shared across all custom GPTs. It appears to be 20 per 3 hours. There isn't a public endpoint for custom GOT message limits, just an error returned when you exceed it.
Can anybody please advise.
I'm a student MH nurse. Special interest in AI - Specifically for assisting with decision making, risk management and care planning. BA in leadership in health and social care, with a digital health pathway.
I don't have any tech quals, I didn't code but I understand the fundamentals. I'm working through deep learning (Goodfellow et al 2016)
But the maths it so hard, I don't wanna code. I want to advocate for safe and responsible AI. I am in a position where I can influence large institutions that are currently anti AI.
I can't work out if I need to be able to codr, or just understand the fundamentals.
It's hard man
this is probably better posted in #off-topic and not here
Can it be moved?
It's specifically a Gpt question. Because it's GPT that makes me think I don't need to code.
probably best just to delete it and repost in #off-topic to save a mod from doing that
why chatgpt4 is not able to retain the chat data anymore it forgets the actually things and i have to remind it all the time what its doing and the purpose
You might be running into the context limit. Also, have you tried experimenting with custom instructions or a custom GPT?
ok so i had access to gpt4, been paying my subscription from day one and just because sam decides to stop sign ups on the very same day my subscription expires, im no longer subscribed? lol
did you have autorenew on?
mine autorenewed today without problems
ah fantastic, today was my rollover day as well (and I got booted)
what are the odds? (1/30 or so)
did you use real credit card?
yeah, I think this is relevant: https://help.openai.com/en/articles/7232905-why-is-my-plus-subscription-not-appearing
a real credit card? no i used a fake mickey mouse one wth lol
Some people use debit card that act like credit card in some ways
And often rejected
i meant more like virtual cards or debit
🫡
Lets stay respectful in the server please
Lets stay respectful in the server please
Lets stay respectful in the server please
if the transaction bounced you'd expect that they'd send an email about that too
yup
pretty sure they didnt charge it
if your rollover was on day sam decided to cancel signups you're done
even tho it really isnt a new sign up so
Best way to buy plus is through Google pay or some other alternative using the mobile app
buying plus is not the issue
its bad communication top down im sure itll get sorted just gotta contact support
also, for some reason I have something called 'alpha'
yeah its what non paying users get
Are you a free user?
dude scroll up
huh
I wasn't until today, I just sent through a ticket saying 'hey you booted me off the subscription plan'
Oh then that's some sort of pity gift lmao
Also, alpha usually goes to previously paying users
Dumb bug detected on Chat AI
yes i am doing it on the coustomgpts and you know what i have to do inoder for it to make it retain the info is to copy context of our chats of what we had and then upload in on the knowledge
wasnt it 8k?
aha, ok I have it figured out now, I was in the alpha for plugins when they were in development, Idon't have all tools unfortunately
How come in some chats it says I’ve reached my limit and I’m other chats it’s allows me to still talk ?
"alpha" is just what was available on gpt4 . now that youre free theyve just labeled those options as "alpha" very odd wording but yeah
because the other chats are GPT-3.5 probably
My other chats are making dalle and browsing and say gpt 4
source?
I thought 4k was the output token limit
So I don’t think that’s the case for me
then I don't know
it does not know its own limits
for the web too i believe
Mine said 8k last night. It doesn’t know.
Anyone know why with GPT-4 Turbo, it's no faster than the previous version of GPT-4? GPT-3.5 Turbo is insanely quick but not so much for 4.
everyone was saying that the previous gpt4 was better than this gpt4 turbo
Server are overwhelmed
to the point that the had to put a stop on buying new plus sub i think
yeah but they said that turbo would perform better than gpt4
I figured, but then how is 3.5 still so fast at the same time as 4 being so slow. Must still be harder to run I guess
you can see them when you refresh the page and look the models call from network tab
I think the same, gpt 4 need more ressources
yeah but does have any significance tho? wouldn't you get the same thing by setting up the coustom instruction on it and get the same result before? one thing here is that the knowledge cut off and i dont need to set new/different type of coustom instruct for specific reasons anymore
or am i missing something
its not possible because the page requires your session id
Servers are 💀 lately
They had to close ChatGPT Plus subscriptions because they're so 💀
product market overfit
Anyone knows if is possible to subscribe to ChatGPT Plus using a referral link with trial? If it is also suspend… what happen with the link? You loses it?
hi guys, did anyone built personal GPT? e.g. uploading all data about yourself in a structured format. What did you upload to make it useful?
https://chat.openai.com/backend-api/models
text-davinci-002-render-sha, 8191, Default (GPT-3.5)
gpt-4, 32767, GPT-4 (All Tools)
gpt-4-browsing, 8191, Web Browsing
gpt-4-code-interpreter, 8192, Advanced Data Analysis
gpt-4-plugins, 8192, Plugins
gpt-4-gizmo, 32767, Gizmo
gpt-4-dalle, 4095, DALL-E 3
gpt-4-magic-create, 32767, Magic Create
whats a magic create?
custom gpt probably
askyourpdf? whats that
ooh
thats the list chatgpt gets when you refresh the page and all in that list are chatgpt only features
What is magic create
And gizmo
see above, probably custom gpt
What about gizmo
anyone can check the list themselver, browser F12 -> network with that url -> response
Lmao
I put the magic create thing into the url
It now just shows "ChatGPT" instead of "ChatGPT 4"
What do I do with this weird ChatGPT
I've reached my usage limit 3 times today anyone else hitting it more often?
you need to access it via browser tools and not with a direct link, and yes 32k
How long should I be expecting to be on the GPT-4 wait-list?
we don't know
Is gpt down for anyone rn?
it's up -> https://status.openai.com
Yes - I’ve been trying to get in all morning.
Definitely - apparently it's been reduced but no-one's quite sure what to (this evening I seemed to hit the limit in literally 10-15 messages; just glad 3.5 is a bit better than when I last used it...)
Nvm it works
Doe anyone know when they’re adding the gpt store?
in the next month
Hi there,
Does anyone knows if it is possible to force a custom GPT for example with a key word, to answer using only the uploaded documents for it's knowledge ? 🧐
Yes, just tell it exactly that
Well I tried with something like "When I ask you something that start with "Use your knowledge" you must search only in the uploaded documents that I provided you for you knowledge."
but did not work, the bot still give me answers from the web
Turn off web(?)
I just lost my plus subscription because the auto renew is not processed somehow. I'm sure my credit card is working well. I cannot re-sub because there is a waiting list. Does anyone know who to contact with for this kind of issues? I need the sub soon for a project.
huum, yes I can do that, but what I want is just sometimes, in somes cases, I want him to look only in the PDFs uploaded without browsing the web.
but without deactivating the browsing feature
Sorry bro
They closed ChatGPT Plus subs today because of the load from people signing up because of devday
Just try it and see if it works better
Web browsing does kinda interfere
I will give it a try
Does anyone know the cap for GPT4 and GPTs? GPTs seems to run faster than GPT4
Heard about that. It's the worst timing for the autorenew to fail...
i never use 3.5
its 25/3. They have altered the deal. Pray they do not alter it any further.
So I paid for plus, I see the subscription payment made yesterday, but I do not have access to GPT 4. What is going on?
Last time I had to contact support over email it took about 2 months to get a response
Very cool stuff
It'd be faster to send your message via pigeon lmao
Its incredible how they can build something like GPT and fail so hard in so many other areas
Everything's breaking errors everywhere neighbours are trying to ban it in my area
Does anyone else keep getting responses (which use up prompts, no less) that say "As I am unable to browse the Internet..." etc, so you have to waste another prompt to say YES YOU CAN BROWSE THE NET @#*! so it will go "oh, derp, guess I can after all"
Can anyone guess when they openn renew plus?
the new User's UI change, i domt like it
"error saving draft" - should become the new slogan
F sck '
Bruh chatgpt is horrible for me the past 2 days. I hit the usage limit since it kept refusing to run any code. Can the Policy/Guidelines PLEASE BE UPDATED?
yah same, i dont really like it, i hate change lol
the new look looks ugly bruh
No, we dont update guidelines
You don't even work at OpenAI
lol
So do you
I never said I worked at OpenAI
exactly what i thought, it remind me of twitter and facebook change their UI to look like the Android/Apple app few years back
@upper fern You're acting like your a OpenAI Employee by saying "No we don't update guidelines" First of all its not we its you who doesn't want to update a company's guidelines when you obviously don't even work there so why bring that up.
you said 'we' as if you work at Open AI lol
Exactly lol
Also this Sami guy tried to say that I don't work there either when I never said that I had a job at OpenAI so that's obviously 🧢 since I never even said that I worked there in the first place
Gpt has elevated my work email game to a god like level.
Where do you work?
look for forgoten opened windows. they still have the option to upgrade
i managed to renew because of that
i had chatgpt window open at home computer and when i pressed the renew i had the normal update button instead of "waitlist"
after that it sent me to the normal payment page where i managed to pay
You sound like a bloody ai bot trying to keep it going lol
I am not a mod, ,but I highly suggest to stay respectful by rule 1 before getting mute or warning
Men in black
Please be respectful.
Can GPTs use HTML as knowledge?
Where's that?
i pay for Chatgpt plus and have been for a while but now its asking me to upgrade to Plus even though i have had it for a long time now. anyone running in to the same problem?
I'm experiencing the same issue. When I landed on chatGPT site this morning, got pop up on updated terms and services, accepted, and suddenly got downgraded to the free tier.
Yeah its messed up that they just downgraded without any notification and now we have to wait on this waitlist to be right back where we were before
I found this forum thread: https://community.openai.com/t/suddenly-downgraded-to-free/99897/32. They had this issue before and it might be a bug?
Emmmm, ChatGPT's UI had a overhaul ...... did i just notice it or was it changed a while ago ?
Thank you so much!
ok after the whole gpts thing im back from planet bard
seems i cant get plus because i was an hour too late tho..
"We've detected suspicious behavior coming from your system"
Umm.. excuse me, Please give me more details.

Is there any place on this discord to ask for testers of our GPTs? It seems like the custom GPTs are only available for Plus users and none of my IRL friends are Plus users.
You're implication of a "tester" sounds like you just want it to give you free access to GPT's for your friends.
LOL what
When I was trying to build in playground before the GPT builder was available, I was running into errors. I don't want anyone to use my own GPT until I have it ready for real
I would be more than willing to run some GPTs through the motions for others
I think the bottom line is that, you should probably make your own discord, and advertise on disboard or somewhere similar as a server for using GPT's with Plus members, and just have people who are plus members test your GPT's.
Yeah, I'm not that serious. Yet.
Anyone know why the knowledge cutoff date is showing January 2022 again? Am I no longer using turbo?
I think it would be neat for them to add roles here based on tier once they add more tiers
Should be able to help trouble shoot better and connect easier issue wise
Hey
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
is there a new word limit to an answer or am I going crazy
the bot refuses to make something longer than 120 words in one reponse
I just got a response well over 120 words. What model are you using? Do you have any custom instructions active?
Sky is not Sky--my favorite voice is Sky, she resembles Scarlet from Her. That was the draw. And when I go to settings and listen to a test sample of Sky, she sounds like my old Sky. But when I use her in a custom GPT I created, she sounds like a different person.
gpt 3.5, whatever I tell it it wont make its answer longer than that
Anyone notice their voice prints being different in a custom GPT?
ok, now it decided to listen to me
Do you have any custom instructions active on your account? It's near your account settings
it fixed itself
not sure what was wrong
I've heard of the voice being different in custom GPTs, not sure how it works exactly. In a demo at dev day they made a pirate GPT and it spoke with a pirate voice 😁
how much enterprise subscription cost?
nothing
because ur not getting access to it since you dont qualify if you dont have a team.
you can read more on their website.
Why were subscriptions suspended? Can someone explain it to me?
question on latest version. How can I see how many messages my Plan supports? It used to be visible when creatinga new chat?
noticed something interesting - seems GPT's have a separate cap to normal GPT-4 - I cant submit any requests to GPT's ( Cap reached ) - but I can still prompt normal GPT-4 ?
If you're having it as a regular issue you might be better off adjusting your prompts - I always now specific things like "write a detailed report" or "discuss in depth"
I noticed that too - I think GPTs may have a lower cap (which could make sense - I sometimes feel like the GPT process is more like input > GPT-4 output > translated final output - though that could be as my GPTs are quite specialised and ask for particular tones etc)
i go one step further and ask for it to "an absurdly detailed" report etc
hyperbole and exaggeration helps with all AI
If I want something like that I tend to find it works better with specific, quantified instructions
well normally i am asking for an absurdly detailed plan, then i can curate from that
"Write six detailed paragraphs of analysis per sentence in the original text; begin with the first three sentences" is a good one (though you eat through inputs quickly as for any fairly long piece you're going to be making several prompts and that's before any retries)