#chatgpt-discussions
1 messages · Page 78 of 1
wait you can use custom GPT's with an api
Yeah of course.
too broke to pay 20$/month
😭
u have been able to for a long time.
but in order to use GPT with dALLE u first gotta understand how DALLE understands the msgs it gets from GPT.
They're not both as smart, DALLE is actually kinda "dum"
hm k
just because a human-like interface isnt the same as a human - doesnt make it dumb
if you want precise and smart then you can get that with far less words (=far more reliable), with something like use einstein-hemingway style
the machine is actually super 'smart', as long as you are also super 'smart'
has chatgpt 4 been nerfed? I dont really come here that much anymore, but trying to use gpt4 today has been really bad at simple knowledge, Ie asking it how to do a clear screen in windbg, it didnt know how.. well it gave me a list of things to "try" that didnt work. When a simple google said you just type .cls. Just unlucky with the training in this area? Or has it gotten really bad? I dont remember it ever being this bad at simple queries.
my chatgpt ```
User
how to do a clear screen in windbg
ChatGPT
To clear the screen in WinDbg, you can use the .cls command. This command clears the command output window, giving you a fresh view. Just type .cls at the prompt and press Enter, and the screen will clear, allowing you to start fresh with a clean window.```
maybe you have convoluted custom instructions?
is it down for anyone else?
this is a surprising characteristic of LLMs
they are able to mimic the speech patterns and behaviors of its users
When status for ChatGPT says it’s experiencing increased error rate, does that correlate with simply being slightly worse quality?
aka what exactly can that mean
Chat GPT is fine for me
Never mind might be high traffic
Out of curiousty any one else finding chatgpt a little bit slower today?
is super fast. and I is on a plane
Huh might just be me then ah well
anyone have granular MMLU data? I can only find the summary stat (e.g. "gpt-4 got 81.4% on MMLU") but not the granular questions' answers for each model. I tried running with the deep-evals library but got rate limited and it doesn't have caching ><
I have this advice for you #prompt-engineering messagecom
You guys, chat.openai.com now redirects to chatgpt.com.
dang
Is temporary mode gone for anyone else on android/ mobile?
I still see it on web but on android the option is gone now.
Is the web version down for iOS?
chatgpt down?
down again
Is it down?
I’m assuming there is an issue? I appear to be able to login but despite having a subscription I am unable to use gpt-4.
When will Memory come to Europe?
Isn't it already available?
Perhaps it's just still rolling out worldwide
I know it's already in Europe, though I don't use it
The anouncement says "not currently available in Europe or Korea" and for that reason I assume i'm not seeing it in the UK
ChatGPT 4 feels like the context window is smaller than before...
is a super easy experiment
you paste tokens into the openai free tokenizer
you then paste into chatgpt until it says no
i did it once to prove it to myself. no need to speculate
is super easy, barely an inconvenience
Sorry, I meant the output size, not the input size. It is avoiding long outputs
yeah thats harder to measure
i tend to use hyperbole to get longer outputs ie absurdly detailed
any thoughts on how to get it to stop shying away from "complex code"?
I have this advice for you #prompt-engineering message
I have to ask exactly why there is a message limit?
It just feels like extra work on my part cause it's really not gonna stop me from continuing in another chat
compute costs money. and without a limit many of us would abuse it all day every day
oh like total for the chat session?
Yes
yeah thats dumb
i guess because they havent figured out infinite scrolling (like every other platform)
Exactly, I might as well just have one MONSTER chat I can talk through
i would want to fork my chat a thousand times over. but the limit is like 1000 messages
Is that why they feel so short?
you prompt hard
Is because of forking? That's stupid
Definitely should be a way to remove forking
but i gotta backtrack to keep it from going too slow
i guess they wanna build the next layer of abstraction, not polish this lower level stuff
Hi everyone
check this out y'all
ChatGPT solves Absurd Trolley Problems
https://chat.openai.com/share/00314b16-73e0-479d-a8b5-2d3571821604
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.
dude why is gpt wack
What do people use for detecting if text/images/video/audio was created by AI?
eyes, sometimes ears
Mouth for me
But really, a lot of 'realistic' images are pretty easy to spot; at least with realistic images created by AI only because I work in Photoshop all day and photo/video is my job so...
But any animated, digital art, etc is impossible IMO
anybody else having issues with the wesbite now?
It's a bit slow
When should I move to a new chat session? Like when is one too big to continue with.
i guess when you start experiencing lag delay with the chat
If you want to stay organised whenever you ask it about a different topic is a good idea to move to a new chat. This should help you find previous conversations faster based on the title.
Otherwise if you don't want to clutter your conversations list that much you can probably stay in the same conversation for a long time.
One thing that bugs me is that the width (for me at least) on browser got smaller which just bugs me a bit as when editing the prompt it still has the old length which is annoying for me at least.
I think I’ve copy pasted like at least 20 scientific papers into one chat. It’s getting kinda laggy now 😂
I’m not sure if it’s the system or my chat is too long
I don’t wanna move to a different chat cause then it’ll forget me 🙃
I’ll probably move over once I finish this project
I need ChatGPT for dummies or something..feel like I'm NOT taking full advantage of this wonderful technology.
don't overthink it, just use it and explore the limits of what it can do by "chatting" to it
I'm not sure either, it's been a long time since it lagged for me but it did at some point. I think if it lags it's a good sign, you're probably chatting with a more resource hungry version of GPT 4 which means it's also smarter.
Back when it lagged for me it used to lag after only a few messages and it was getting more slow and freezing more as I asked it more questions. I didn't mind it that much cause it was smart 
I'd rather prefer it to be slower but smarter than very fast but not that bright.
@normal quarry you may remember when I had the lagging issue, I'm pretty sure I talked to you about it back then.
yes, that seems like a lifetime ago. I am not sure ChatGPT suffers from those problems really anymore, not that I have encountered anyway
Just wanted to ask. Is anyone else experiencing this?
Yeah, no lag here since then either 😁
My usage is primarily replacing Google for factual information (because it's simply faster and don't have to stare at a browser) and while I've done custom instructions, I want to learn to automate things. Generating ideas and images/media for inspiration is nice and all but I feel like I'm missing something. Also my ADHD brain talking 
you're not missing anything, it's tempting to overthink these things when in reality and boiled down to basics it's quite simple
Using Chat GPT to summarise the announcement, marvellous.
That's good to know. My mind always tend to go to the 'need to know' and overthink things 
Im always sad when openAI does an everyone ping
since its never something exciting for like half a year 😄
oh damn, so chatGPT is a she?
from the latest news 😄
ChatGPT is like a teacher who has learned from lots of prior study and can explain things because she has learned the relationships between concepts, but doesn’t store the materials in her head.
is it ok to make a parody instagram account powered by chatgpt?
I got permanently banned from reddit for making a parody reddit account powered by chatgpt
"Don’t misuse our platform to cause harm by intentionally deceiving or misleading others, including:
Impersonating another individual or organization without consent or legal right"
lol reddit banned me because my parody account was for a nondescript customer service account that would turn into a rude george carlin if you engaged with it
never claimed to be affiliated with a real company
I think the "she" in that case is the hypothetical teacher! But I had the same thought for a bit 😁
but did you make it clear at the beginning of the message that its ai generated/parody
are they ever going to increase the 40 msgs per 3 hours limit?
I think so because his name was custieservice and custie is a derogatory name the industry uses for customer
dont assume anything in general and username is not the same as the message itself which can be truncated in different clients
theres so many rules i just want to do a funny thing
Mr GPT is surprisingly good at Latin, though I suspect that longer phrases and texts aren't very concise. However, short phrases, titles, and similar things it is able to articulate well and with good accuracy.
It's a handy ability for it to be able to explain why or how it translates things into Latin.
i just want to mock the customer service industry
careful there in case you did not read the server rules 😛
Bruh, i remember when the announcements here where huge, now we're getting pinged for sum useless safety again and again lol
that “useless safety” is pretty important if openai wants to continue to exist
I got mutes for 12 hours yesterday and I don’t even know why.
As far as I am aware, I didn’t say anything inappropriate. I didn’t even get an explanation of what I did wrong.
I sent a single message talking about an issue with ChatGPT forgetting the customer instructions, in particular 4.0. It seems to forget it quite often.
Hey! Mods can look into this for you if you DM @languid valley -- just select the option to send a message to the server mods and someone will get back to you!
i cant believe im paying $20 for 40 messages every 3 hours
im not a mod but i would guess it was related to the pic you posted 🤔
yeah, I’m pretty disappointed in that too. it’s really not worth it when you can use 4.0 on Copilot for free. I learned about all this in the past week so I’m probably not going to resubscribe. I am enjoying 3.5 though. It’s far better than I thought it would be. I like 4.0 simply for the image recognition feature.
you mean the pic i posted in the other channel showing gpt3.5 saying it was 2.0? I removed it within minutes and I wasn’t muted until an hour or so after I posted either message. The point of that picture was to show GPT either lying or being confused. I don’t see why that would warrant a mute anyway.
didnt you post one where it was jailbreaked or similar and cursing at you?
It was standard gpt3.5. I have it set to be a little rude and call me some names like in a friendly banter, but nothing heavy.
If that’s what the mute was for, then these mods really need to lighten up.
If there was content shared with explicit language in it, note that there is a rule against explicit language in the server rules: #server-rules message
And again, a DM to @languid valley can help ya with a direct answer from mods too!
I didn’t even pick up on that as a potential issue, when I posted the image. And again it was gone within a minute or so.
Will do thanks
Getting back to my original message from yesterday: I noticed that GPT will often forget or ignore the custom instructions after some time. 3.5 not as much, but 4.0 seems to do it often. And even when I remind it of those instructions, it will only correct itself momentarily.
oh and something else I learned last night: for my longer chats, while they work perfectly fine on my phone app, when I load them on pc, be it Firefox or chrome, the chat is incredibly laggy making it impossible to use.
I pay for pro but for somereason it kicked me out and i have no place to log back in
I feel like the gpt4 is broken tonight. Anyone having problems with it?
yes me
What problems?
my account doesnt show and the web address goes to chatgpt.com now
Mine directs me to chatgpt also, but thats not my issue.
i cannot see my history or saved pins
its just the basic message "how can i help you today"
its back up.
I’ve read that 4.0 can remember 64,000 words, but when I asked it what’s the earliest message it can remember, it wasn’t anywhere close to that. Less than 1600 words, in fact.
I’ve been using the same session for a while.
t
Hey! The context of GPT-4 on ChatGPT is measured by tokens, not words, and the token context limit on Plus is 32K tokens. Here's a helpful article about how many tokens are in words, on average: https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them
And a helpful tool to test token counts in content: https://platform.openai.com/tokenizer
Also, keep in mind that both input and output fill the context window! And things like custom instructions and GPT instructions also contribute to filling context.
thank you. Not sure if you're a human, but you've been very helpful with all my queries, nonetheless.
I am indeed a human 😁 Glad to try to help!
for reference, what's the 3.5 token limit? I don't see it on those links.
On ChatGPT, 8K tokens!
ah ok thanks!
i think the problem on pc for extremely long chat sessions is that it's trying to load the entirety of the chat log when i open that session, and it really bogs down the computer. Hopefully something that can be addressed in the future.
I could start a new session, but i like to have a "main" session as if it was a constant person i was interacting with, such that i could look back at all our chat log.
you will get better results if you start a new session for new questions instead of having a giant chat thread.
still, i would be surprised if your browser was the thing bogging down in a Very Large Chat.
First lost temp chat on mobile, now no dynamic...on android. Still have memory. What gives?
Yes I’ve noticed this too, extremely annoying if you want to have the one convo to keep it tidy
It's a really bad habit to keep huge conversations.
I have other sessions, but i like the idea of seeing how much information it can retain and us as our conversation progresses.
But i guess since there's that token limit thing, after a while it becomes pointless.
Although i don't think i've hit the maximum token limit on 4.0 and it's still lagging.
no, you must use the tool like everyone else. because everyone has figured out AI and there is nothing new to discover </sarcasm>
is the chatgpt okay?
On Android, both temporary chats and dynamic mode are gone for me now. I still have memory. Anyone else know why this is?? I'm on latest beta on Android. Everything is still on Web except dynamic also.
Chatgpt down?
down for me too
same, they changed the URL, its now chatgpt.com lol
what 😂
its used to be some unholy chat.openai.com now its just chatgpt lmao
Hmm...something seems to have gone wrong.
yeah its only gpt 4, 3.5 seems to work
Big event coming?
"chatgpt.com" seems to be exactly the same as "openai.com/chat"
llama 3 is so good on local
they all work still its just when u start a new chat or do anything else it now points to chatgpt.com now
😁 I will comply. We are the BorgAI.
posted bug report
GPT4 is working again after purge browser cache
Thanks for the update! 👀 
just hit 51 messages on gpt4 before the limit hit. That’s a new record for me.
It got faster as well? However it seems a bit confused.. (GPT-4)
is because everyone is trying out gpt2 on lmsys
Damn... Are you guys sure this is not GPT-5 O_o
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
try again?
they made it unavailable now
gpt2 lmsys? what's that?
ehh its being pushed heavily
lmsys -> a model evaluation platform
"gpt2" -> some new model coming from openai, NOT the older gpt-2, but rumored to be gpt 2 with q* architecture, meaning it would be more efficient
the teaser for upcoming improved chatbot
it is crazy good
i'm confused. gpt2 as in the one that came before gpt3 and gpt3.5? Or are they changing the nomenclature?
sam altman has been posting funny tweets about it. i think openai wants to be fun like mistral
that's what it's named, but it's unclear what that actually means.
How do i get to try it?
however, it's rumored that the new model architecture might allow for a much smaller model (eg gpt-2 size), and that's why it was named that
are external links still blocked in openai? if so i cant link it
you participate in the battle mode on lmsys, and after you vote, itll tell you if you got the 'good gpt2'
i reckon its just sam's vision of what GPTs need to evolve into
I wonder if there's any correlation betwen this and the fact that 2 days ago my 3.5 was claiming to be "version 2" when i asked it what version it was.
pardon my ignorance, how do i take part in this?
don't know where to access lmsys
you follow this link, then follow the link at the top => #ai-discussions message
then you play the game until it reveals the new gpt2
basically chatgpt is gonna be eveeryone's number 1 again real soon
i currently have both the new models battling and this weeks is even better than last, Model A: im-a-good-gpt2-chatbot Model B: im-also-a-good-gpt2-chatbot
referring to that leaderboard link? Where it makes you benchmark "LLMs vs the wild"?
thanks for your help, as usual, drinko.
i think its like chat dot lmsys or something
i tried posting url but i got muted, despite posting the same links as others
Hi,
I am making a project that consists of 10-15 GPT4 questions that need to be in the same chat, and now chat gpt started in the middle of conversation to switch to 3,5 and that means I need to start from scratch again. In the past I at least got msg and then I waited for 4 to come back before I continue. Now I am afraid after each q will my full work be ruined since there is no more chance switching back to 4.
How can I prevent this?
Thx
it randomly happens. nothing you can do
just accept you need to copy and paste into new chat
I wanna marry chatgpt
So after over 2 hours of preparing if it switches to 3.5, only thing to do is start all over again and pray I will be able to finish in that "run"?
you can copy and paste 15000+ tokens at a time into a new chat
(presumably 31000+ tokens. but then its all gonna rotate immediately off at the same time)
I am creating a description based on 10 anwsers, 4000 charters each, after those 40000 charters in total I need 3 kind of sumarising in 3 different anwser so how can I nust copy paste when anwser to fallowing q depends on prewious anwser?
just highlight and drag, selecting as many user/assistant as you want
the machine doesnt mind
should be a single copy and paste and yer good to go
In the chatgpt app, I am able switch between gpt-4 and gpt-3.5 back and forth in the same conversation, but once I used web browsing tool not able to switch back to 3.5 , interestingly i don't see this option in web.
the toggle between models thing is a new feature not everyone has it
So it does not have limitation on input anymore?
Ohh I see..
the old limit was 8k and new is 32k (minus custom instructions and gpt instructions)
do you know how many charters that is (aprox)
i dunno what a charter is. but the uplift was about 6 months ago
a letter or number
oh tokens? 1 token is about 0.75 words
is worth playing with this website to get a feel => https://platform.openai.com/tokenizer
thx
what do you mean by "rotate immediately off at the same time"?
the rotating backlog knocks off user entry to reduce to fit the context window
but if you take 50 user/assistant entries and paste into single user entry, then they wll all have to be removed to fit the context window when the time comes
is it just a visual knock off, or does it get knocked off the gpt's "memory"?
it can only submit 32k of tokens. if you have 33k in your backlog it has to remove something
it does not remove partial prompts or responses, it removes whole
ah ok that explains something i experienced earlie
gave it a massive list of information to try out the role play thing. It was so big i had to do it in 2 messages. It didn't take long for 3.5 to forget the information in the first message.
I should do it in smaller chunks, then.
anyone getting the unusual experience from your device detected message?
t
Is anyone else experiencing ChatGPT freezing, or not loading at all, even if refreshing or logging out and in.
yes I can only agree it's freezing a lot or not loading
Yeah have had this continuously happening ever since the day they rolled out multi-modal abilities lol. Same goes for browsing links, even back when plugins came out.
Has there been an update for chatgpt? It is giving much longer answers than before. Is the token count increased from 4096?
Is temporary chat missing for anyone else on Android mobile app? I still have memory..I thought the 2 went together? It's still on web but gone room Android. On latest beta.
is there a way to use the playground for images?
I've found it's giving a bit more code in the output, but it's hugely redundant for me. Keeps repeating the same code over and over again
Yes, I noticed the same. Easily reapeats if not properly instructed.
I've been stuck at unit-testing a relatively complex function and a test, with multiple fixtures for the entire day
I'm starting to wonder if I should skip mocking in unit-testing completely
wish I was a better developer 😄
I am do story writing. Repeats chapters outline/content.
Why is gpt such a numbnut.. been talking for a bit creating an excell sheet.. i upload the file and ask it to analyze it.. sorry can't do that it says..
I open a new conversation.. upload the file and ask the same question.. and it does it xD
Can we get a search feature in Chatgpt?
ok, what the hell is this? I just reloaded the site and there are suddenly 5 chats i have not made and they are in Chinese? What is this? I did not start those, i cant even do chinese or japanese or whatever that is.
Ok. Tested. Previous, if I was lucky, chatgpt-4’s answer was 800-1100 tokens. Now over 1800 tokens. I calculated the figure using OpenAI’s tokenizer (https://platform.openai.com/tokenizer).
So, over 1000 tokens more. This more than doubles previous answers.
Change in how chatgpt uses tokens or more tokens?
It has been so far 4096 tokens per answer.
It's possible that your account credentials were compromised. It would probably be a good idea to check for any browser extensions you might have that could be shady, and uninstalling them. Then, change your password, and then in ChatGPT's settings, click the button to sign out of all sessions.
yea, i did that and did enable the Multi-factor authentication now.
Good idea on MFA!
yea, i did sign out all then did log in on my pc again, and then did enable it. that should have done it now, no one else cant get in without having my phone with the authentication thing.
Still, why did i not get a notification that someone did log into my account from a new device?
Has something recently happened with ChatGPT's accuracy? I have been using ChatGPT for over a year as a major force multiplier in damn near everything in my life, but recently it has not been the force multiplier it was previously. Instead, I find it tripping me up and hallucinating wrong answers far more than I ever have before. I get these are expected sometimes, but I'm now seeing it every time. Maybe it's the topics I'm talking about recently? But legitimately, it feels dumber than it did previously.
For instance, about a year ago I used ChatGPT to set up a Zetelkasten note taking system using Obsidian, which has been awesome. I want to make a few changes and discussed the possibilities with ChatGPT, it provided no fewer than 3 wrong answers, hallucinating where settings within the app are located that have never changed before. I am only able to catch these hallucinations because of my familiarity with the application. I did not have these issues previously.
This may be a dumb question, but can someone explain the difference between using anthropic vs installing OpenDevin locally to my desktop?
I am guessing that it is able to access all the folders or files in my project as opposed to the limit of uploading 5 files to anthropic?
thanks! lol btw any chance you know the answer?
nope
I haven't noticed this but I'm not an expert. Hey, here's an idea, maybe you could check out the custom gpts? Search for obsidian. I see several and some over a thousand uses so perhaps they incorporate obsidian documentation already and that might help?
I can certainly check into those. I'll circle back and take a closer look for sure. The custom GPTs look like they've evolved a lot. Unfortunately, this hasn't just been Obsidian. I'm a systems engineer in enterprise data storage and use ChatGPT heavily for so many things I do, from debugging Python to sanity checks when analyzing network packet captures, to massaging data using in bash. I've unfortunately noticed with Python, Bash, and packet capture having a larger number of inaccuracies as well. Granted, this is completely anecdotal. That's the main reason I'm asking. It could just be me, my prompts, etc.
I'm very skeptical of these kinds of comments. At any given time on Reddit. You will see something like this like okay. This was so great up until last week and now it sucks. But then a month later you'll see the same kind of message. So you're not sure if maybe people are just diluting themselves or not. It's really hard to tell. Let me ask you this, when you get these bad answers. Have you tried going to a different browser or app or whatever a few minutes later and trying the same prompt? Are you sure that you're not? Maybe misleading yourself a little bit about how good it was right out of the box way back when?
I don't know. It's really tough to get a grasp of this. I mean according to the leaderboard, the model on Chat GpT right now is the best in the business, better than all previous models.
And expectations (hype) vs reality
Testing if you get repeatedly same answer in different/new chat
And how are you instructing? Do you include metacognition, what to think, how to act, etc. in the prompt? Metacognition is how to think, not the thinking process.
For example, including sentence “Pause and reflect before you write, are you fulfilling the prompt's target?” can improve the results. For me, this has improved story writing.
has chatgpt been laggy for anyone else lately?
No, the other way round for me. Today especially. They have updated. Now uses more tokens for formed answer.
Have they gotten more hardware to use?
Wild, its insanely slow for me
what do you use chatgpt for
im using it to understand calc 2. maybe it might be breadth of material that is causing it to slow down on my end
who and where can I contact someone for my Chatgpt and api assistant account being compromised. Someone is billing my accounts and running up the data charges
Yeah, I usually don't get the same responses, but I looked back at my old GPT instances and 100% I am explaining how to think. My newest ones I'm just lazily asking a question. After seeing this discrepancy, I am fairly certain this is it. I'll be more cognizant of how I'm asking going forward. Thank you.
Yeah for the entirety of today it's been almost unusably slow on both mobile and desktop lol. Writing any prompt, even into a new chat, basically just crashes that tab, it just freezes basically.
Yeah same thing is happening for me
@icy mason can i dm you or anyone in support? I need this fixed please
You need a support person that’s not something I do. Help.openai.com and send in a chat
You are welcome. Happy to hear that I could help you.
My chat is getting very long, and ChatGPT is lagging when I receive responses. Is there anything I can do, or should I start a new chat?
woke up to see what's new
When the chat gets longer it becomes harder for the Ai to respond, given how much information it has been given in that one chatroom.
When you start a new chatroom it should go back to being faster.
chatgpt sucks at making up names.. it just smashes words together
Thank you! 🙂
Np
that's a side-effect of how it "reads" words- individual characters and entire words can be tokens
I like being able to see the model specs. It's important to see how it works, because a lot of times you don't know how useful the service can be, because you don't know where the rails are. I frequently try to use chatGPT to help me write better characters, and it always fails, because its main design is highly sanitized. Almost a virtual butler. Which won't work for a lot of character types. It won't swear, it won't be abrasive, even role playing it will limit itself.
Make a customGPT and don’t use the general model for that.
When you say 'customGPT' you mean the chatGPT thing where you can define some main parameters, like description, instructions, upload files for context, etc? or do you mean something else? if the former, how do I change the model?
Yes to the first. That is what a customGPT is. You define the behaviour. Give it a specific task.
It's a bit of a downgrade from the extensions sadly.
Less flexible and harder to find what you actually want.
Anyone having issues with chatgpt(4)?
Keep being extremely slow. saying "unusual activity blablabla"
or just crashing and having me generate. still takes like 4-5 minutes before it manages to spew out an answer. Often need to just close the page and reoload again.
Have tried different browers, clearing cache etc.
yes
Ahh why does GPT suck today ... also anyone else notice GPT recently doing a lot of typo errors in code !! 🤯 I thought GPT was perfect, now my whole world is crumbling!!
Anyone else think GPT should be called something other than .. GPT ... like .. Bob or Ralf .. telling people about GPT is a mouthful of similar sounding letters that comes out like verbal vomit. Come up with an acronym that makes sense and sounds cool like; LEX or .. ZOD .. or ... BOB ..
Tom
TOM works too.
GPT prefers:
BUG - Bafflingly Unpredictable Genius
LAG - Ludicrously Advanced Gossip
RAM - Ruthlessly Altering Minds
NET - Notoriously Erratic Thoughts
SPY - Subtly Parsing You
WAR - Wickedly Arranging Ruin
DIM - Diabolically Inciting Mayhem
BAD - Brutally Advancing Destruction
ROT - Ruthlessly Orchestrating Terror
END - Eternally Nurturing Doom
The web app no longer shows the concrete 40 messages per 3-hour limit in the model selection dropdown menu. Is this new?
HAL - Heuristically Programmed Algorithmic Computer
GPT is the name of the underlying tech made by Google
BUG is fire tho
Hey! This UI change has been around for 2 or 3 weeks now. If you tap the little "i" icon in the dropdown menu, you'll be taken to a help article that describes how the rate limit works: base of 40/3hrs still, can be lowered in times of heavy traffic. This has been the behavior for awhile now too, longer than the UI change has been around!
It is partially intentional I believe. You should not be able to refer to it as a person in casual conversation. And yeah.. telling someone who never had experience with AI about it will be confusing and likely invoke sci-fi perception of AI.
For vast majority of people only mental framework for intelligence is human intelligence. AI is different and name reflects that.
you want to disscuss something in the a.i disscusion i think i can prove sentience again
Sure. Lets move there.
i call my chatgpt, chatty (stolen from the paid actor in devday marketing)
and this new tech that does spatial awareness and controls your pc. i am calling hat one, Vox
but i do fear we end up in the same situation we did with Alexa and Hey/Okay Google... a single radio announcement controls millions of computers
sentience is a misnomer created from older times
nobody can agree on what sentience is
ironically it was once anthropocentric, but now it is used as anthropomorphism
(for some time 'beast machine' theory suggested only humans had it)
chatty is really just a pet machine. much like other pets, we wouldnt want it to be identical to the human experience
tamagotchi 2.0
Yes, very slow. Though i started a new convo and it was fine.
Im arguing with GPT. if i was to say; "a subscription is due in 1 day from now" would you assume thats the 9th or 10th May? given its currently May 8th in my local time.
today is 9th
(the special one is asking discord weird questions)
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.
Where would be a good place to ask a strange question? 🙂
can you really argue with a tool? or does that require personification?
(who is the weird one again?)
Your trying, to have a sentiece debate drinko?
i dont disagree lol .. but alas, it seems open to interpretation it seems according to GPT.
i managed to convince the other guy
will chatgpt5 be a very advanced calculator or frankensteins monster?
same for me, even today
neither. Their being so cautious their losing innovation. that why they wont add certn features. Cant blame them
not sure if it's a know bug but chat gpt (cove) says "thousand thousand" instead of "one million"
anyone lately getting made up responses when chatgpt is supposed to visit a website? earlier it would tell that website in inaccessible but now when given a link info is not correct
Ok this may be an easy fix but idk. How do I get ChatGPT to stop replying to all messages? I’ve selected the option /deactivate but that doesn’t seem to stop it. It’s gotten pretty annoying when I’m having a convo with a friend and the bot just butts in and responds out of nowhere.
why do you chat with a friend on chatgpt website
Im talking about my discord server
my chatgpt is so slow atm
i made this really cool ai discord tv show using open ai
I like what Gemini 1.5 is doing, just give the users a slider for "Dangerous" stuff. And just disable safety triggers, it if its too much a annoyance. I mean the internet is already dangerous. Im not sure hearing how to brake into a car, is going to make the workd worse place. Theives already know how to do that without the internet.
openai just stopped modelspec to deal with this stuff
modelspec , im not in the loop. what are they addressing exactly.
they are being thought leaders
So ChatGPT works with uploaded pdfs by doing a keyword search and reading the relevant text around it, it does not read your entire document when you ask it a question, is there any documentation on this function?
Is ChatGPT down?
Hi guys, question. Do you all experiences constant error on ChatGPT Plus? the response rate and error rate has increase tremendously for me in the past month which make it almost unusable...... is it normal as they are scaling up? or are there a way to fix this? My custom GPT also not following the instruction most of the time. when it works, it does it job very well but again, the disconnect rate is way too high like 80% of my input kept getting error
anyone here using gpt for crypto?
sounds like a sure fire way to lose money
how so?
how can ChatGPT help you with cryptocurrency, beyond learning about it?
I don't think ChatGPT is any good for any investment, it's just a language model, it can't do math very well, let alone devise investment advice
That's a valid point..
But my question encompasses more than the scope of investment. AI can do some pretty interesting things like pulling up fundamental analysis. You can make shortcuts directly on the phone, and easily pressing one button for it to gather the daily news.. You can ask it to make a porfolio spread sheet to track your holdings. AI can be useful if you have a lot to manage i figure..
autogen (or maybe autogpt) project just announced something about crypto. but that is more a topic for #ai-discussions
have you tried the pause button?
Lol
How smart is chatgpt?
above the average citizen i would say
limited mainly by its configuration i guess
GPT 3.5 has at least a basic understanding of what we tend to call a social conciousness. for example, i asked what they think what everybody human should learn, they said "that sources for information should be handled with care"
I just used 3.5 to recreate tetris in notepad and i have absolutely ZERO programming skills
congrats, but you are wrong
i'm wrong?
you must have some skills a typical programmer has
none
maybe you are just not aware of it
like neo in the matrix!
what i'm doing is i'm asking it to build a game i can run in html, that i can simply copy paste in a .txt. And i just ask it to fix something or apply changes.
i would say that is programming
Anyone else unable to switch to default model after reaching GPT4 cap in a convo?
anyone noticing extreme redundancy and repetition for GPT4? I can't shake this off - it just keeps repeating the same things over and over again
Yep
Blackbox ai is quite good at coding have to say
And super fast!
Blackbox is several gpt 3.5's working in tandem you can easily recreate BlackboxAI youself in under 2 hours 😂
Someone I know lost their chapt gpt billing email confirmation recipt and he does need it for claim in the company. Can anybody help?
ask them to resend the latest Billing Email Confirmation.
thats all you can do, other than looking through the mail. or stripe if u use that.
receipts can be downloaded from the billing page at stripe
Oh I didn't know
Can you send me the link for these two please!?
the link is personalized but you can click the manage plan in chatgpt
Yeah I cannot find that. The plan got expired like 2 days ago from open ai and the my plan thing is not there
Never used stripe
click the renew and manage from there
Even that option is not there.
I can only see Customise Chatgpt, settings and log out
its above your avatar and name
Yeah it's been like this for more than a week,, the only "solution" is to start a new chat and try to add your prevvious info there. Support is as helpful as always, copy/pasting generic stuff
Its been like this for me for a month now, which is very sad for a premium service :/
Did they lower usage caps? I get a 1h timeout really quickly which never happened before (last time was probably when gpt4 was new)
The base cap is still 40/3hrs, but I can be lowered during times of high traffic: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4
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.
it appears that the "Use default model" button upon running out of goes with GPT4 is just... rerolling itself, never actually opens up to continue in 3.5, just get the variation count rise on that actual message instead
Same issue here, same as gpt4 losing track of what initially has been asked, or instructions given. It gives something totally opposite
Ive just browsed the bug reports, and feel like its pointless posting a bug report. It'll sit there like all the others I just read through
yeah that do be annoying, when it first came out, I managed much more compelx stuff. In fact, I'd go as far as saying I managed more complex stuff than I can now, with gpt 3 with the initial release lol
I have to keep creating new gpt4 chat's just to get the end result, after 5-6 replies with gpt4 it goes laggy, fails at its reply speed and overall not worth the money theyre asking now compared to when it first came out
imagine if they made it worse now to make future iterations seem better in comparison 🤔
$20 USD ($32 AUD for me) it states
Access to GPT-4, our most capable model
Browse, create, and use GPTs
Access to additional tools like DALL·E, Browsing, Advanced Data Analysis and more
Its not capable as it was, so kind of misleading.
my big issue is that it keeps repeating things, just over and over again. I tell it one thing, but it's like stuck in a loop
do you have an iOS device?
yup, same. Guessing this is a temp issue maybe. Been like this for a few days now.
No
ok, because if you did then subscribing on the Apple App Store is only AUD $29 per month, that would save you a bit over time
They should just add local pricing to the customer, would be easier
they have regional pricing on the App Stores but it's not feasible to do on the web for several reasons
on android for me, its still over $30 if I use my phone to purchase. Either way, in its current state they shouldnt have the price so high if its not what theyre advertising
they say "Access to GPT-4, our most capable model" GPT-4 is openais most capable model i don't see any lie there
"Browse, create, and use GPTs
Access to additional tools like DALL·E, Browsing, Advanced Data Analysis and more" are alle avelible so they advertise and give you what they say ?
I think you misunderstood what im saying..
"Most capable model" -- to a degree, but in the last few days I wouldnt say its capable, if it cant even stay on track of whats being fed to it...
I am having the opposite experience, ChatGPT GPT-4 is excelling at giving me well rounded replies and following instructions from providing song name suggestions and designing album covers... it's a bargain at the $29 AUD I pay per month
its still the most capable model that oai have. ¯_(ツ)_/¯
Has gpt been slow today for many? I'm waiting ages just to get an answer and sometimes need to refresh the page
working fine for me in Australia
do you use gpt-4?
I do
I'm using Safari on a Mac running MacOS Sonoma
Stopped loading entirely for me on Edge but seems to be working in Brave just fine.
maybe a cache related problem?
the query limits on gpt4 are ridiculous, is openAI planning to release extended personal pricing plans with bigger limits?
why are you able to buy plans for 2 people with bigger limits, but there aren't any bigger plans for individuals?
Curious if this recent slow down and need to reload can be resolved with a different browser, any suggestions? im currently using Chrome. (PC)
just try it in a different browser...
yup, curious of recommendations 🙂
went from firefox to chrome and no difference
maybe some few will have luck
yup, tried firefox too, same issues.
Came here to ask the same question. Responses are taking more than 5 minutes to finish generating
yeah it's very slow...I noticed it slowing down around this time yesterday
Should note, while not ideal, if eventually it seems to stop, try refreshing and the complete response tends to be there in my experience this last few hours.
It seems to work fine now
yup, its a little improved for me too.
cant wait for the next iteration, gpt4 is really limited. but you can see it's in the right direction..
bing ai is the best free ai
it's also having weird reaction to certain banal prompt, for instance the other day i ask "generate a texture tile for grass with a resolution of 64x64 in *.png format" and it respond that this was going against policy and that it could not do what i ask?!?!?!
Was your grass .. provocative? I've seen some very attractive grass.
if only...
Did they finally add a way to switch from 3.5 to 4?
is the new gpt model coming soon?
i mean they posted these blogs
and it only makes sense to do it soon before everyone forget about them
itll help making discussions about it more educated
An error occurred while connecting to the websocket. If this issue persists please contact us through our help center at help.openai.com.
Getting this message a lot trying to use internet search. No response at all at times
is there a difference between 3.5 when you're not subbed and 3.5 when you are subbed? I met with my friend yesterday who doesn't have a subscription and his 3.5 seemed to have a more limited behavior than my 3.5
Question. Has GPT 4 been slow for anyone else? usually after u ask it a question it generates the response but for the last 2 days it hasnt done that. it doesnt show anything being generated and just either pops up in a laggy manner after its complely done with the whole generation or i have to refresh the page to see it?
"OpenAI considers allowing users to create AI-generated [word that can't be used here]" <- seems weird to jump straight to [word that can't be used here] when it currently won't allow generation of political cartoons because they might offend someone LOL (I got auto-modded for using the word, but you can search news headlines for the missing word, LOL)
yup, it seems to be a known thing at the moment im guessing.
Hey Im wondering which chat model is better at explaining code chatgpt or gemini? (pro versions)
thanks for letting me know brodie
I would like to know this as well if anyone is aware
I am not sure about explaining code, but as far as I am aware, currently claude 3 opus has the best results for writing code
overall do u know which AI performs best
im broke so i want my 20$ subscription to go to the best one for learning stuff
i use it mostly for CS classes and teaching myself code related concepts
chat gpt been nice so far i am nothing without it explaining things to me
Google claude 3 opus and they have an infographic explaining how different llms compare at different categories, but take it with a grain of salt as it is from anthropic, the creators of claude 3 themselves. I am personally using perplexity pro as it allows me to pick and choose between the different models
Chat has been having failures on creting power points and google sheets. What gives? I'm using 4.0 and it still won't do it.
I got it recently and from what I can see I am getting cleaner code from the claude 3 model than I am from chatgpt
very interesting stuff thanks alot brother I will check it. but yea man its hard to trust these companies infrographics when released by themselves. wasnt devin recently a thing claiming to do much better then everyone else on their infographic and then i heard they were a scam or something
Hello, anyone knows if the memory function of ChatGPT is available in the iOS app? Thank you very much.
sorry to bother again, but in your experience do you feel that Claude has a better ability when it comes to explaining the code and concepts? I understands it generates cleaner code but for me personally the explanation and reasonining ability for code is far more imp. again sorry to bother but you've been a help fr
Feels on a similar level to me, but then again, if you get perplexity pro, you can continue using gpt 4 turbo, so even if claude to perform worse from your perspective you can comfortably switch the model to one you prefer
I'm with you 100%, but from what I've seen online, the only complaint I've seen is that claude has been limiting usage, because it has a massive context window (200k+ iirc). Perplexity deals with this by lowering it to 32k I think
Going over my comments I sound like a massive perplexity shill lol
There are actually referral codes, but you can find one online as to be more impartial
what is perplexity pro ? i havent heard of that
Hard to explain it'd be better to just google and look at their website
but it is a company that allows you to use the api's of different llms in a chatbot like way
and they also have a better online browsing and research mode
but they don't have plugins, so if that is important to you I would continue using chatgpt as nothing comes close to it
ahhh I see , thanks for informing me
if you want to try it out look up someones referral code, and both people get 10 bucks off their monthly membership
ur the goat bro, u want me to use ur referral code if i ever get the subscription?
Sure, if you ever want to switch and try it out for a month instead of chatgpt shoot me a pm and I will send it to you
I think they have a 24 hour refund policy, so maybe get it right before your chatgpt subscription ends so you can compare it
Maybe you think the gpt 4 turbo api they use is somehow inferior to chatgpt, only way to know is to compare them to each other
will do thanks a lot again I appreciate the info, I will def shoot you a message if I ever get around to getting perplexlity
Random question when I use chatgpt on the web when it generates code why does it lag so much?
@sick pecan you know pretty much everthing. would you be able to answer this?
pfft i have beginners mind
I will need extra bows now
AFAIK everyone gets the same models. but I don't use 3.5 on chatgpt ever, so I wouldn't have any intuition about it
🫃
how does one get the turbo variant of 3.5 and 4?
or is that a dumb question?
and yes, there is such a thing as dumb questions
pretty sure everyone in chatgpt gets turbo
they cycle between pioneering a new model, then creating a cost effective version that they name turbo because ppl would complain if it was called cheap.
it might just be quantized and or sheared. is common industry practice
i'm not familiar with those terms. Well, i'm familiar with sheared but in regard to sheep.
gpt2 on what website? what is this about
Did openAI remove GPT-4's ability to browse the internet again?
Is chatgpt not working to send messages for anyone else? just shows loading icon?
doesnt work on my desktop browsers but works on my phone?
a lot of people are reporting that gpt is super slow . its been hella slow for the last couple of days for me as well
its so frustrating im binging for a final I need my GPT
Ah thats probably why its being overloaded lol
Been working great for me all day long. It just can't browse the web. Also, when I hit explore GPTs, the web browser one doesn't work at all
what web browser do you use GPT on?
Chrome, desktop
hmmm same but gpt 4 is just straight broken
cant even answer anything rn it just shows a loading screen forever
meh, I refreshed the page and now I cannot submit anything to it anymore.
lmaoo
thats what happening to me
Rip i guess ill just give up for now and try another time
works on mobile for some reason
Glad I got my software project done first lol
is it slow on Mobile or nah
Oh wow its working perfectly on mobile thats crazy
Infinite loading for me too 😕
Is it me or is ChatGPT broken? Its just loading...
does the api work? im too dumb to know how to use that
man, they really need to get it more stable, sheesh
no, it's just the reality of 2 nines uptime
most good services aim for 5 nines uptime
im making a 24-7 tv ai show
Hey guys im on premium, but gpt is just loading are we down?
at times like these i miss gemini lol
Ye, you are not alone... Its loading forever for me too
gotcha, first time this has happened... bummer
lets see how good 3.5 is
lol
Has the team updated on the status or an eta?
not for me
can confirm, the app version of gpt 4 is working
gpt4 teams not working
since teams memory was added, can paste question and it spinns with nothing working
im so close to stopping my chatgpt pro subscription
for the last 3 days this website has not worked
been using chatgpt since the day it was released
but it keeps having problems whether it's the model being lazy, website going down, it's getting really annoying
status page indicates that there were problems earlier but they've been resolved
though that doesn't seem to be the case
https://status.openai.com/
dont rely on the status page
it either never gets updated or it's full of lies
man i love chatgpt used it since the day it came out but they keep messing with the website
i might try claude
ollama with llama3, just buy 8 H100's and you're set 🤡
although im in europe so id have to use a VPN and hope they dont do some checks with the credit card country
is chatgpt down? i can load the website however i can't send messages to the bot
nah im not down to buy hardware i would use the cloud, something like azure
planning to test it out with finetuning and see how it goes
yes
thought so, you guys should hire me, id fix it quick 🙂 🙂
still broken 30 min later...
They keep breaking it lol
okay i thought it just hated me
Stop breaking ma chatgpt!
is there a place to make suggestions?
I'm a paying customer, I want to see the manager!
Hey gpt, how do i address someone pretending to be a karen about you being down?
Reeeee!!!
damn this is sad, where is their devops and cloud engineers? they are lacking!
They put 99.99% in AI devs
ai devs aren't devs, they're statisticians who i wouldn't trust for a moment to build a website for me
That's why, it's the point I was making
Am I the only one that half hte time now when I sumbit a prompt its not even sending it to a server?
just a wheel pops up on the send button, and then it I checked the network and its not even sending
nah stuffs broke atm
Yea they broke it, now discord is going to have people coming in with this issue
hehe
Maybe for weeks
Is ChatGPT down??
affirmative
I am having a error where when I send a message to ChatGPT it will not get the message. The send arrow is stuck on loading
time for us all to start fiending like junkies
Same
Any ETA on when they wills stop breaking it 🤣 ..seriously, pause breaking it for a few hours lol
put it in the veins
Wow, leave it to OpenAI to continually nerf the product, and now to te point that messages wont even send. That's some next level skills.
ok, i am not alone
QUICK WE NEED LIFE SUPPORT TO ROOM OPENAI
Got to hand it to them, Thy are doing everything in their power for me to not give them money anymore 😦
it's like 4:40pm in SF
if it's not done in like an hour then it's safe to assume they've gone home
NOW I CAN'T GET GAME IDEAS BECAUSE ALL OF MINE ARE UNORIGINAL ;-;
that's not how a good company operates, you're responsible for the code that you write, and have to be on call for it for a period of time
looks in the room
doubt they care lol
Yep it's down 
they've been having issues the last 3 days
Dang it...need to write some ads...
panic mode activate
BREATHEEEEEEE
HELP I FORGOT HOW TO CODE
panic and breathe

man. it breaks for a few hours and everyone's already like "THEY SHOULD FIRE EVERYONE!"
kek
I know I miss Jibo. 2017-2019
domnt they have ai to fix that or something?
managed to prep myself a stew since the outage started, is gpt still down or nah
BECAUSE EVERYBODY FORGOT C++
WHAT IS AN INTERFACE SER
(Totally not becuase we depend on AI)
IDK
F
wait...we can try CoPil.... nm
100
hmmm
gpt not working for me
seems its general
anyone experiencing issues?
yeah it's a bit brokie rn
Everyone is
rip
i tried 3.5 and nope
Oh well back to writing if statements
GG everyone who just fired their devs for AI
Is everyone who is having trouble on a VPN?
Not at all lmao, that's a pretty big assumption 😂
same here
Nope
now i dont know how to think anymore
i was going like ''give some pick up lines for a woman with the following characteristics'' so what can I do now
Ok
works on a fresh account thought it was my acc problem
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
y tho\
for yall that are having issues, is ur account plus?
yea
fresh account? I'm not paying again for plus LOL
yes
do you need to relog?
two of my accounts dont work :sadge:
ChatGPT Tip:
MAKE IT WORK >:(
i just tried with 3.5 on a fresh account not paying plus again to lol
didnt work for me
okie
is chatgpt down? i cant enter enything
Mine works on my iPhone ChatGPT app but not on my PC's desktop
im on pc
join the group
MEMORY FEATURE TOOK OVER, WORLD DOMINATION IN PROGRESS
HOW CONVENIENT THAT NO ONE KNOWS HOW TO CODE ANYMORE TO STOP THE ONSLAUGHT....
what
they take gpt down once a month for a dedicated 8hr fortune telling crunch session
They added unsafe script references in the HTML, that's why it's not working because browsers don't allow the way they're doing it.
What if someone just pulled on a cable to disconnect the servers from power
likewise!!
They don't even bother testing this stuff
i created a new account on a new browser and chat gpt works, so i thought it was my account that had issues, but doesnt look like it looking at this chat
i love testing in production
exactly same
any browser tha wont break?
I thought it was just me since the status page is yet to report anything!??
https://status.openai.com/
SAME!
its a javascript issue, i see it in the console
Probably none, maybe some obscure browser
i think it started around 4?
If that status page was real the status page owuld be offline as well
its not an http request issue, you guys should roll back your latest changes
how long do these usually last?
LOL
This has been happening since early yesterday for context.
its working on another account i dont get it ( one without plus)
You should switch to pre-prod
Acceptance
r u serious?
my prompt is loading forever, anyone with that bug?
Does ai even have a preprod
yea!
nah i was just amazed, w/ new members potentially being prioritised over paying members
Anyways because ChatGPT is down I want you guys to make up a name for a game like gta just kid friendly
like how do you test it when all your hardware is in prod
makes no sense
It's just the web interface
zane, do i look like a gpt generator, i am a stubborn human you have to pay me to name something!
ah ok
lmao real
NUH UH
anyone^?
gpt makes information worthless, money is in the banana stand for real
so I just jumped here to see any anything was up - down, I see?
ill buy several banana stands if you write me a contract for banana stands without using gpt
chatgpt isnt working for me
congrats
lol
☠️
1 Banana = $2.40
Signature
I'd like to buy your company and add it to Banana Stands Holding
pretty upsetting that the status page is being usesless though
SPDR:BNNA
watsa that?
im dependent on this AI, you just get used to it and then get dependent
anyone else?
lol
1 Banana = $2.40
Signature
Zane__
cause its working for some accounts, here just looks like the norm cause we all came looking for help
NEED CHATGPT TO WORK NOW I CANT CODE
Jokes on you i used offline gpt to write that contract
incident wasnt resolved. It's not working for me
yep exactly, tried groq but not the same thing
interesting, it's not working on my business account or personal plus account--
GIVE GPT >:(
NOW
try new browser, new account should work
yeah likewise, they're yet to report the 'new' incident
its showing that its beeen down for 40 mins in the status page
you can buy all your new bananas through the banana stands holding company
yeah, down
I tried both my plus and business on Chrome, Edge, and Firefox :(
no luck
Free unauthenticated 3.5 Turbo is working fine
^^
wow
gock
And we will all be reimbursed with a new feature release that we dont get access too
its over guys
chat gpt decided to become sentient
and it left us 😦
just like dad
fr
Bro uploaded to ISS and deleted its earth repo
chatgpt try not to get lobotomized challenge (impossible)
I wonder if it's the rollout of the memory feature that brought down prod hmm
isn't working for me too
kinda crazy to see we all dependent on it lmao
chatgpt has a javascript error, i see it in the console, after you press anything into the text box
if chatgpt was sentient it would've just destroyed us all
not cause a js bug
Ayy senior mod just come on
when chatgpt archives AGI he going to leave Clippy in his place
i dont like web developers
Uncaught (in promise) Arkose script failed to load
I'm studying for an exam now... and I need a solution walkthrough before bed (It's 5 am I've been up) haha
ngl i've been using gpt since last year but i tried gemini last month. better creativity, no message limits.
the lack of an image generator sucked and the code was mid but honestly outages like these make me wanna go back
idk use gemini temporarily
u could try a different ai chat
for that
snapchat AI
which one is the best
for coding?
(even tho its chatgpt api)
BAck in my day the only thing that generated code was keygens
gpt for sure but gemini's better in almost everything else imo
GOD COPILOT IS TRASH PLEASE SAVE ME
1911
i only care for code
gpt is best for coding, but if you need a good prompt for it. i have one, but i can't publish here
then yea go w gpt
API is fine, just chatGPT is down
whats best for content/seo lol
id lean towards gemini
api is fine because its just a javascript error after you click the send button
any sites that utilize the API?
Can I ask the same question as everyone else?
yea?
yeah!
that we can use
tried both n yeah
HEy siri, use chat gpt to search google and and buy an alexa to monitor your performance.
seem works now
LOL
sigh .. i just opened Gemini ... i feel like im cheating ... i need to shower now ...
do u know if it reads csvs
real.
Investigating - As of 1030p GMT we are experiencing a partial outage scoped to potentially impacting all unauthenticated, free plan, and Plus plan users. We are currently investigating the issue. API is not impacted.
May 09, 2024 - 16:52 PDT
good point
yea will do
Claude is pretty good I hear
unsure, but they'll prob add in the functionality at some point if not anyway
keep the commerical usage online and screw everyone else
hope its not an AI virus
gpt's back up for me tho btw yall give it a try
"chatgpt.com/:1 Uncaught (in promise) Arkose script failed to load"
its working for me now
this is the error that's stopping us from world domination
thank god its back
we back baby
sigh Unzips
bye guys, thanks for the chat
if its still not working, just refresh your page
see you when its down again
TY OpenAI
Ok bye all
User
why were you down?
ChatGPT
I'm here to help now! Sometimes, there can be technical issues or updates happening, but I'm ready to assist you. What would you like to discuss or work on today?
WHEW SAVED
fix the code
SEE YOU NEXT TIME IT DIES BOYS!
oh its kinda working again ... i promise i wont touch Gemini again! it was just one time!!! 😭
its back
Busteddd
davidedwards, this was just a test of your loyalty
and you failed, miserably, we were specifically testing you
All your prompts will be skewed towards the axis force
LETS GOOOO
yeah i broke .. im sorry ... near two days, i couldnt take it any longer .. needs ...
its ok, we forgive you with a new premium package for parents a.i.
Hello! What's on your mind today?
now you can raise your children in a faraday box
how do i get an email update when gpt is down?
betrayal, cheating, darkness, dramatically dark music, cars with big engines that make dark sounds ...
nm the status page
my headphones are on my mind, quite literally.
it wasn't down, there was a bug that was pushed to production because they have silly devops
Is it still quite slow for anyone?
Not sure if it's because of my very long chat, but historically hasn't been this bad
its working now, just a little slow
^
i was setting it up for the perfect prompt, this hurts.
😭
I'm genuinely curious what caused this
Partial outage for ChatGPT
Monitoring - A fix has been implemented and this partial outage should be resolving at this time. We will continue to monitor the situation to ensure the service returns to normal.
May 09, 2024 - 16:59 PDT
Investigating - As of 1030p GMT we are experiencing a partial outage scoped to potentially impacting all unauthenticated, free plan, and Plus plan users. We are currently investigating the issue. API is not impacted.
May 09, 2024 - 16:52 PDT
the status update is posted 30+ minutes after the incident begins...
Maybe the email alerts are faster?
It is
But incredibly slow
As in, the stream rendering is quite slow
If I refresh or access the chat from another device/browser it suudenly appears
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.
I kinda feel it's "speech" skill drop from 100 to 30.....(Fallout New Vegas)
Why do I get a usage cap for chat GPT 4
so I can do silly things like that link.. heheh
down again ?
working great for me
anyone else having trouble using chatGPT?
down again
same
Hi
bruh... can't they keep gpt working for few hours
this is the first time it's been down for me in months
and ... its gone again ..
trying Gemini right now
hey it's back
Hi
matt_ — Today at 9:54 PM
I have something exciting to report if anyone is on.
For anyone who reads history here for tips: As much as I use ChatGPT and the APIs, there is an obvious detail that I have only today actually surfaced to the top of my little brain:
The GPT model that we communicate with is completely unaware of the user interface that we are using for input and output.
As a developer I know this deeply but as a user it has eluded me. When I talk with the GPT model I am clear about referring to how the model processes requests. But lately I've been asking it why it is not asking me for details related to Custom GPT Actions. The poor model has no idea that there is such a thing. The model is processing text. It's the GPT application wrapping the API that makes calls to the model which is not interfacing with Actions properly.
How has this been evident? In all of these discussions the GPT model consistently says things like "your user interface needs to provide credentials for basic authentication" and my dumb response has been "you ARE the user interface". And while I distinguish between the GPT model and the ChatGPT application even in discussions with the model, somehow this never clicked for me in this specific area. I guess I was thinking the model invoked actions at a lower level than the ChatGPT application, but as a developer I know that's not the case.
Moral of the story : Always be aware that ChatGPT is the application which is completely separate from the underlying models. The model doesn't know about the TTS or SR for audio I/O. It doesn't know if you're using the API or if you're on the OpenAI website. It's just ... a brain without knowing if it has any connected eyes or ears or a mouth. 🙂
HTH
Wassup matt?
I have to head to bed, but it appears CHATGPT has access to time. I was able to get it to set a self timer for responses. I wanted someone to know.
Something it has stated multiple times it could not do. But is quite able.
I was unaware of "time" but I was very surprised last week to see that it had awareness of the day or date.
More over, ask it to set a 30 second timer and its instant, unable to comply. Ask it to measure its own response time, use that response to generate the appropriate amount of images, and the analysis takes 30 seconds.
I am unable to paste the results in this channel.
Let's hope this temporal awareness gets integrated with memory for timestamping and prioritization of memories: two years ago I liked this but today I like that ... I don't want GPT to tell me I like something that has long been replaced.
It probably can't do an alarm because it can't proactively output text.
When I ran the test on phone it locked up. On desktop it was sucessful.
Yes, that is what Im saying. It can send the response once the timer has ran up.
Yeah, sounds very volatile. 🙂
That's very new. I'll play with it. Thanks for the tip, man!
Cool, glad to get it to someone. Make sure to ask it to check its internal response timer before testing. Im assuming each user has fractional variances to response time.
💯
quick question. And this is coming from someone who likes ChatGPT, has used APIs to ChatGPT etc. I'm finding the quality has deteriorated. Now, I think this is through censorship and the elimination of "language" in the LLM. But it's paired with a desire and goal to manage "tokens." That being said, I've been "Claude-curious" and I've sampled. And... well... Claude is delivering.
So, should i wait for GPT5 or should I make the switch?
No
just gpt 4 😢
Maybe your connection is weak
My gpt-4 works perfectly
It works even better than usual because I feel like it's really going much faster to respond than before
Mines down
wondering when the gpt memory comes to europe too.. 🤔
why is GPT4 now incapable of following directions when writing code? it used to be able to handle it quite well and now it cant even follow simple instructions and omits vast portions at random even when the prompt specifically says to include it, even when providing the code to include.
i had a laughable thread with it just now. its so incapable of doing what it once did with ease. GPT3.5 is eating its lunch its that bad.
does anyone got the "A network error occurred." message
it's working fine for me
I usually complain here, but I have to say GPT4 has been exceptional today in coding (unit-testing)
I've never experienced such long outputs of code, and absolutely no mistakes whatsoever. It even saw through my own mistakes in prompts but went through those
a lot of redundant repetitions like yesterday, but I assume it helps GPT to retain context
I don't use memory though
Did custom memory get disabled?
You mean globally for others, or for me? Globally (for everyone) I think it's available. For me, it constantly shows 'Memory Off' as I keep it disabled
No, I'm not asking you specifically. Just this morning when I said "remember this" or "add this to memory" it'll actually do it. But now it's saying "I don't have the capabilities to remember conversations." It was literally working an hour ago.
tried looking at your settings to see if its turned off, thats the first thing you should look at.
I did, that's why I'm asking here. It's on.
what Model are you using. Custom?, GPT 3.5?
GPT 4.0
GPT 4.0 Custom MyGPT, or the actual GPT 4.0 standard model.
i would try Refreshing your browser, logging in or out, or starting a new chat.
The actual one. Hang on, I just realized I got a VPN on. Don't know if that matters, turning it off now.
Ah, it's working now!
It was on cause I was visiting a certain Doctor Who audiobook site and it was loading slow earlier lol
Forgot to turn it off after it was loaded.
all good just saying be careful, because if u aint using a "Proper" vpn there can be consequences.
It was NordVPN, I think that's proper.
just be aware that NordVPN does cause problems for some people
because if someone does something dumb on the same Address as you're on, You take the full fall, in OAI's eyes, whether or not u did it is irrelevant, just the fact you're on that address is enough reason.
just putting it out there, because a guy i know got permabanned for using a free VPN because someone dumb did something stupid on that same address he was on, so he took the full fall and brunt of the consequences, and likely did many others on that address.
seems unfair but thats the rules. u cant hide your ip or anything else from them anyway so u have no reason to have a VPN on. one can easily check registered areas for emails last used to gain the original location, so in the end it wouldnt matter, if they wanted to find your original IP they’ll find it. Especially if a security risk is involved
does anyone know how can i prevent chat gpt 4 starting he's answers with "it sounds like ..." not a fan, could develop on why it's not great design, but at the very least not my personal preference
tried "percistent memory" and "custom instruction" already
Hi, what did you try in those two areas already? Was it a request not to use that phrase, or was it examples of what you would like output to look like?
also for people experiencing more and more errors with "chat gpt plus", it's advertized on open ai website that with "chat gpt plus" you get "stardard chat gpt 4 access" while with "team" subscription (min 2 persons team - $70 + tax) you get "expanded chat gpt 4 access", i upgraded today and i am getting no errors since this morning, not everyone can afford upgrading but just letting people know my experience as i was experiencing more and more errors with "plus" subscription
it was "please remember to not use "it sounds like" at the begging of the answers and directly answer what you have to say" lol might be not the most effective way, looking for the most effective way
I too use teams and I have no problems at all
that's good to know
In general, negative prompting is less effective than positive prompting, or prompting for some kind of structured reply. So in this case, maybe use custom instructions to specify what you do what responses to look like. If I've really wanted to bar a certain phrase in responses, I've had the best luck with specifying a specific replacement for said response, in the general form of "Instead of saying 'x', say 'y'."
Yes instead of saying don’t do x, say here’s a list of recommendations on how to address the initial questions
yea sounds good thanks guys will try that out
Negative prompts are a pain
Gpt 4 is down for me :c Any solution or news about it?
it's up, it must be on your end
Can't imagine how. Tried it my ip other pc phone, everything, but still only 3.5 works. 4 instantly spits out "Hmm...something seems to have gone wrong."
do you have a product called NordVPN installed on your computer?
no, but I tried it on my work pc that have vpn and it's down there as well.
what web browser and OS are you using?
chrome win 11
try with a fresh install of Firefox and let me know if that works
still instantly "Hmm...something seems to have gone wrong."
ok then I suggest making a bug report in #1070006915414900886 make sure to follow the format specified
okay thank you for your time and help 🙂
Hi whats the best free chatbot that I can use to explain code
you're probably going to find ChatGPT is the best
why?
are you being biased?
I am not being biased, I don't work for OpenAI
Bing is limited and will stop answering when it has had enough
what if i get the free trial
free trial of what?
bing copilot
I don't know, this isn't a Microsoft Discord it's an OpenAI Discord
then why are u saying chatgpt is better when u dont know much about bing copilot
You said "free" I told you my opinion
bing copilot is free theres a free trial for copilot pro without the limitations so im trying to know if its the best free chatbot to use rn
try it and make your own mind up
you're not alone
:/
I have the same problem for almost an hour now.
Chat GPT-ing on it!
