#chatgpt-discussions
1 messages · Page 110 of 1
which is cool but it's still worse than Claude-3.5-sonnet by a big margin
particularly in coding questions
but it can also be better than Claude in some cases
Could be a bug? Has it been this way since the 4o's release or you are experiencing it post 4o-mini's update?
Started like 3 days ago
And since no dev commented here I think it's working as intended
Free users have limits, people have told you that many times in the past 3 days
.......uh-oh. I reinstalled and all my chat disappeared
did you log into the right account
Sorry, but I am going to ignore you from now on since you talk in circles.
I did!!!!!
Try refreshing by pulling down from the top
Wait it out. Normally issues such as that gets resolved in time. I had to refresh my app couple of times earlier today for my history to load
Has 4o finally replaced chatgpt 4? Because it wasnt smarter. But now chatgpt 4 says legacy model and 4o says for complex tasks.
On the drop down menu.
I've got no idea why you are specifically experiencing rather longer rate-limits for gpt-4o. Weekly/monthly, as you claim it. Might you want to create a ticket in that case, just to be sure
GPT-4o is the flagship model, GPT-4o mini has replaced GPT-3.5 and GPT-4 is now a legacy model
What happened to the "Continue Generating" button for long responses? I'm not seeing it anymore and responses are getting cut off mid-way through.
thanks
Why, if my promt says: no code, why answer includes something like : here is code for reference? Also, why is gpt4o so forceful to include answers about stuff asked 5 prompts ago, even when specifically asked not to do so?
Usually, you just need to type in: Continue. And then it will finish, if there are no errors with the service itself.
there's a button still for me
a rotating arrow type thing
Does the read aloud edge use a good AI voice like ChatGPT ?
Is it conspiracy to say that a year and a couple of months ago we had gpt that was showing high levels of understanding, and that company has lobotomized the program so that they can maximize the profit, while lowering the quality just low enough not to be seen as garbage?
Yes
I will, how do I open a ticket?
while I don't completely agree with what you are saying it definitely has been more of a money driven thing rather than innovation rn, I remember when gpt used to be better than any other model but claude is now way better and more powerful
Maybe. I don't believe in conspiracies. Problem with them is that there is too much b s online, so valid questions are buried under the river of garbage.
yop
Go to help.openai.com and click the little chat widget in the corner.
Ty
for some plus users maybe
You didn’t see Sam Allmans tweet?
He did say that, yes
Yayyyyy
Hip hip
In the tech world where wonders rise,
There stands a man who never lies,
Altman tweets with truth and grace,
In alpha releases, I trust his pace.
His words are solid, his promise clear,
With every update, there’s no fear,
So here’s a nod to Altman’s way,
In his tweets, I trust every day.
dies_from_cringe.gif
Voice mode voice mode. Put me in alpha
I find it perfectly good at what it's ment for, but no, we're nowhere near OpenAI's new voice.
what are you doing
gm - I'm new to the server but I am looking for mor einformation on how Chatgpt can extract info from a PDF and spit out a csv. Is that possible?
can anyone tell me any free AI image generator ?
Now free users get GPT-4o Mini 
Get back my GPT-1 Bro ! 🗣️🔥
20 years from now, 'gramps used gpt-4' 
I want gpt 3.5
its gone
I'll buy chatgpt plus
its gone from chatgpt, including paid plans. only way to use it anymore is to use api and pay for it
I'll use api for turbo
I found that when I select gpt-40 mini and have a voice mode conversation, it automatically switches to gpt-4o
I don’t know if I can use gpt-4o mini to continue voice conversations when the maximum number of conversations using gpt-4o is reached after the new voice mode is launched.
I see that I have access to Chat GPT-4o mini now. Does this model have any maximum message limits per unit of time?
as I understand gpt-4o mini can read images too does that mean if you are using gpt-4o and you run out will it still continue to work without having to make a new chat like you had to do with 3.5
According to their status page, they are not down - https://status.openai.com/ but mine won't load either.
Wonder if it has something to do with the CrowdStrike issue.
incorrect. Both are very close and will depend on the task which one does better. Even benchmarks can't really separate them beyond margin of error.
tbh imo benchmarks don't really matter that much when talking about ai models
like some models who are high in benchmarks like Gemini are actually dog water
objectively, they matter much more than some random person's words. They are fairly accurate as well if you know how to read them.
ironic that Google who has the most data out of everyone doesn't have a good enough model
eh I guess you are right, but many people I know that are in the ai sphere would agree with me that they don't really show how good/bad a model truly is
as of rn none of the models is good at advanced maths
they still struggle with fine-tuning, but their base models are really strong. 1.5Pro performs very well if we exclude all the bugs and hallucinations 😉
well the bugs are the issue lol
but still Gemini seems like a worse choice compared to Claude or gpt
they mostly caught up with the basics, but not the fine-tuning still...
I had this question of decoding something and only Claude 3.5 could do it
yeah they just chasing the numbers it seems 💀
yep they only care about that
or their massive context thing... 1.5M tokens, like... it is good but ehh
should probably focus on performance more than anything else
Is the support team on vacation this week?
Because I sent them a mail and they are taking too long to reply
I have never been able to reach anyone. I sent a chat and it says 3 days to reply, but I have never actually got a reply. They take your $20 and hope you go away.
I didn’t sent a chat however, it was a mail
you mean you used the help site bot or replied to a follow up email?
I mean I sent a mail for the support email using gmail
there is no email for support
its not used or listed anywhere. openai staff said here once that its unmonitored address with autoreply
the mail says to use the website
Indeed, here's where it was said: #chatgpt-discussions message
If your having problems logging in turn on a VPN and then turn it off once your in.
it looks like it's the sign in process that is broken.
I didn’t recieve anything like that
You can take an official stance from me saying that the only way to contact them is through help.openai.com and that support@openai.com is an old, now unused email.
Where do I have to click on the site to send them a message then?
You'll find it in the circle chat widget in the bottom-right corner of help.openai.com.
Yes but how do I start a direct conversation with staff support team? They say to choose one of the options after I choose “chat gpt team” but I don’t see what I want in any of them
That's not currently an option, your best bet will be to go through the default options until you find a path that lets you type your own query. This will submit something that OpenAI can review and get back to you on, though I'm not sure in what capacity these days!
Ok, did it
They normally respond within two-four weeks.
is chatGPT down???
Gosh, that much? It says 3 days on there
I think the 3 days is a joke
Wait, I actually got an email from the support
I didn’t see it
And apparently it’s not a bot, is a human called Lj
And…I got a reply on the help website just now
Noticed that the gpt-4 discussion channel is gone. Guess it doesn't make much sense when all the available models are GPT-4.
Chat Gupta
"You've reached our limits of messages. Please try again later."
stop, man
Essentially
it also means make room for our paying customers
wow, did ChatGPT 4o get an update? its lightning fast!
I know there was mention of the mini-model for free users, however I am a paid user, yet seeing code being written so fast my eyes are struggling to keep up 😍
Can ChatGPT 4o mini access internet?
No it can't
Can ChatGPT 4o mini replace my imaginary girlfriend?
yeh it's extremely fast
lmao. yeah. No idea why they made it free tier before having replacement ready, seems kinda silly and we essentially back to pre-Plus system. Very little incentive to sub
It does not work as it used to, so i'll hard pass on that.
#ai-discussions please
Robert is just trying to help, I think u should listen to him
Practice kindness and positive regard. Harassment, hate speech (such as sexism, racism, or homophobia), or other malicious conduct will not be tolerated. Maintain a respectful and positive environment.
I won't be surprised in the least if they just changed the name of 3.5 to 4o mini with some minor improvements
nah, it's a completely different model
Are you sure? Because it sounds exactly like 3.5 except with a little bit of self-restraint on what it doesn't know accurately.
yeah, it's completely different, it has better knowledge in tech/programming than 3.5 did
Is it better knowledge though or just refraining from giving worse knowledge?
it was on lmsys before being released, came up good in the leaderboards
I don't know... I'm not entirely convinced by it.
better knowledge, it's grasp of Swift is completly different than 3.5 was
Alright I trust you. I haven't had too much time to play around with it tbh.
Just based on my initial reaction of it.
explore it further, it might surprise you 🙂
I want to it plug into my 3D printer and build me parts for a supercomputer. If it can't do that, then I don't know why I'm paying OpenAI.
Make me a cyborg already OpenAI!
I'm tired of this mediocre human existence.
does gpt 3.5 still exist
only in the API
so free user now have access to gpt-4 right?
free users have access to GPT-4o Mini and limited access to GPT-4o only those with a Plus subscription can use GPT-4
but GPT-4o is better than GPT-4
free user should have GPT-4 since it's less better than 4o mini and 4o
GPT-4 is more computationally expensive, so it's reserved for Plus subscribers
more expensive but less knowledge right?
personally I prefer GPT-4o but some people still use GPT-4
Yes me too I use GPT-4o, but do we still have to pay now since gpt4o-mini is free for all?
GPT-4o Mini is not as good as GPT-4o
is there any benchmark?
is there a promt library ?
thank you
Rip ChatGPT3.5 🫡
does anyone have any updates on when the video model will come out for 4o? been waiting for really long
Free users always have severe usage limits on the premium model and when they hit their cap they get switched to something less advanced and cheaper for openai to provide. Last I talked with a friend who is too cheap to spring for premium, he said his cap on 4o was something like 10 messages and bumped down to3.5 This was a few days ago, before 4o mini was announced.
@sudden herald unlike what u said they took less then one day to reply me on the website and I got a reply on gmail from support@
does anybody know why the set of control buttons at the bottom of a previous output in a conversation disappear?
if you mouse over the bottom of a previous output then the controls will show
that's a bit of an issue for me because I've gone blind and now use keyboard navigation and NVDA, a screen reader. I've managed to force mouseover on previous prompts to get the edit button to appear, but the same trick doesn't seem to work on outputs.
ChatGPT both in the macOS app and the web is really poor when it comes to accessibility, I agree it's a problem
I prefer the website more than the app
The app is just too hard to navigate. It's got too many windows.
And stuff
And it's too white
And bulky
are we talking about the same app? I am referring to the macOS native Desktop App
Oh, no, I'm talking about the android app. I don't have mac
Mac is for aliens. Microsoft and android is for normal people.
I must be an alien
Well you do say some alienish stuff so it does make sense
I love big mac's, but that's the only mac I can love.
support@openai.com is still a legacy contact method. Its nice to see that they're support has gotten better.
The android app has its own set of issues but has one really great feature the web version lacks, conversation mode. I'm grateful to OpenAI for developing it. It is a boon to visually impared users. It helped me get started in using all the other accisibility things when I went blind.
They gotta tweak the timing for the conversation feature. You have to come up with a reply in a certain timeframe or else it cuts you off. I don't want to continuously tap and talk. That defeats the purpose of having a wireless continuous conversation.
the new real time chat feature fixes this behaviour, I can't wait for it to release
What? They're fixing it? Where did you get that info?
They are talking about GPT-4o voice.
Reading
Oh wait, I've seen this. That's not released yet? I thought that was released with 4o.
no, real time chat is delayed, typical OpenAI announcing and then delaying products. Premature anouncements seem to plague this company, you'd think they would learn by now.
shakes fist in the air✊
Can you blame them? They're studying AI. AI engineers must have the most scatter brains of all. Their office space looks like an exploded dumpster fire.
They're gonna announce things that they'll release 10 years from now.
When hovering your mouse to get the post-outbuttons to appear, do you have to hover over the text of the output, or below it somewhere, or does both work
? when navigating with keyboard, I'm, limited to focusing on text and buttons. If the buttons are hidden, ther tab'arrow order skips them. Can't focus on a blank area.
hovering over the post works
I miss gpt 3.5, gone forever 
it's still in the API
I just wish we have more categorization features like folders and sub folders type of thing. As the usage grows, more and more chats are being made for different things such as programming, current affairs, medicine, hobbies, cooking etc. I just can't have all of them in a line in the left hand side of my screen. This feature is a must, otherwise the competitors might take advantage of this.
Since the request is already made several times, I can't post it in the suggestions.
I find search works really well in the macOS Desktop App to quickly find conversations
Good. I am a windows users though.
a Windows app is coming later this year
I look forward to it. Hope they put the feature in it.
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
oh my god when will they let us turn this verbosity off. I don't need a conversation I just want answers
have issue logging into gpt4 teams again
Have you tried putting something to this effect in your custom instructions?
Is 4o mini really better than 3.5?
Yes. its even cheaper than it
4o mini is gpt-4 -ish lvl
Only if gpt4o was having less hallucinations. It's hallucinations are not like gpt 3.5, when it hallucinates, it's hilarious and alot worse
remeber that its first day of gpt-4o mini model
it will get better overtime with feedback
gpt-4o mini doesn't support files attachments that sad
Along with tools
I just don't like the personality of 4o (not abt mini). Gpt 3.5 was so chill and human like with my system prompt. But 4o is no fun
I has least humour I'd say
4o = better
hallo
suggest some ideas i think i should code and launch something globally any suggestions
You can still use 3.5 if you use the website and don't login
Can we get a REAL answer why is ChatGPT not working on my windows desktop, chrome or edge. I installed Chrome two nights ago and it worked and all of suddent has stopped again. I can access on my iPhone and Mac devices.
or just write a secret url hahaha
hi, anybody around who has experience using a screen reader with either the web browser version or the android app, doing more than just basic back and forth conversations?
I know some people have had issues with Chromium-based browsers. Do you know if you can access the website on Firefox?
I just realized that
What do you mean?
I can't wait for the new voice mode
I hope they release it with a new model, too
The ChatGPT model used can be specified as a URL parameter:
https://chatgpt.com/?model=text-davinci-002-render-sha for GPT-3.5 Turbo
Yeah, then it'll be so cool. I use gpt to learn stuff often
Good to know. Thanks
Have you guys ever let gpt write some code and thought i'll try understanding it later?
I don't know any programming languages. Wish I had taken the time to learn when I was younger.
I did this today, i'm starting with ML and I thought it's bad it can become a habit so i stopped
I hear 3.5 sonnet is better for coding
you still can, if you can find time and want to go into tech
It would be more as a hobby
ohh i'll try it for sure
You wanna make apps/ websites as a hobby? or solve coding problems instead
I just want to understand. Knowledge is power, and programming knowledge especially so
I love having ChatGPT walk me through code. I think it's pretty helpful to brush up on topics
I keep hitting the limit for free 4o. I'm broke rn and can't afford to spend on ChatGPT plus
Oh you can start learning a language and doing leetcode after learning the syntax then. cool hobby and demands attention
The syntax has been my biggest obstacle to really trying to learn
Samme, i find it very boring to read so i enter a prompt "Talk like a friend" or smth
yes, its required to know basic syntax
you doing python ?
I meant for languages in general
same T-T
ohh
does anyone know if 5 is coming out this year?
none of us know what OpenAI has planned. All announcements will be made in #announcements 
I feel like the next big model they release will be something different
Maybe something involving the research they've been doing with the strawberry project
I posted with adding that to url it take you o gpt 3.5
🤣
Worth a shot! As a quick test, I used this semi-silly instruction for the "How should ChatGPT respond" box:VERBOSITY: LOW Instead of behaving conversationally, enable ANSWER-ENGINE MODE. Answers will be short, concise, and direct. They will contain only essential details.Seemed to work well! https://chatgpt.com/share/0162a5ae-5485-4a89-a729-eaf2eb7639b4
Is it for improving the responses?
I've tried endlessly. It usually works best when I follow up it's relentless 10-20 paragraph/essay style messages with "JUST TELL ME WHAT I ASKED FOR"
Yes same problem, I think if we tell it before starting the convo that you gotta behave and respond like this then it works fine
If you're looking for short, non-verbose answers! I was just replying to someone who said they didn't like high verbosity, offering a suggestion.
What is verbosity?
Using lots of words
Ohh I see
it's the equivalent of a math test asking you to show your work instead of just saying 2+2=4
I use DALL·E a lot and I also use it for help with programming -- I'm not a programmer but ChatGPT does great with helping me make lil projects every once in awhile. Hbu?
IT work and projects
Same, I do ML, I'm a beginner so I use it for suggestions like which architecture to use or how to approach a problem
yes
hey guys, so im a gpt plus subscriber, if i want to use the 4o API, do i have to pay more or is it part of my plan?
API is separate from ChatGPT, you'd have to pay for the API use itself.
ChatGPT likes to embarrass me by inserting me into stories -.-
Does anyone know why ChatGPT is showing the math like this instead of the correct format?
\frac{f(x+h) - f(x)}{h} = \frac{3x^2h + 3xh^2 + h^3}{h} = 3x^2 + 3xh + h^
I could have sworn it worked a while ago
same here
will there be a google forms to fill out to get access to the chatgpt 4o voice alpha?
no
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.
I don't know what the mode is called when you hit the headphones and can get voice responses, but hoping that eventually it can use the input as audio and not transcribe it verbatim? Like I want it to listen to audio and provide feedback on it without it requiring spoken language. E.g) identify animal sounds
It's coming in Fall
Does anyone have any resources about building a tailored speechwriter using GPT?
how do i change the initials that pop up in the top right corner of the chatgpt in browser? I've asked this before months ago but never got a solid answer.
kind of dumb that it isn't more straightforward.
I think it's the initials of the name on the account. If you want to change it to a profile pic, I think it still pulls in profile pics from Gravatar!
I cannot poen chatgpt?
Go to the Gravatar website. Create an account using the same email that you use with ChatGOT then set your profile picture there and it will show up on ChatGPT.
its easy
it's so easy you'd regret asking if i told you.
Where can I see how many answers I can get from the different models, with my specific subscription?
You don't have a white gem in a blue circle there?
is it down?
yeah, but it stopped generating a message and when I refresh the site does not load
nada. I have a green circle with initials
Do you have a subscription, or using the free version?
plus sub
I wonder why we don't have the same
oh, I have a team
So, there's one way you can change it I guess. Get a team. 😛
ha 😄
I have premium but website is not loading. I dont think its sub issue
What error are you getting?
eh
Theoretically not, but there is something wrong with GPT lately. It wasn't working on 2 out of 3 browsers that I use, today it stopped working on 3rd. I cannot use it anymore on my PC for whatever reason. Clearing caches does not help
it's working via the ios app, but i can't access the website for some reason.
Mini pretty good at python ngl
hope voice mode comes out at the end of july
According to this announcement: #announcements message the plan is to start alpha testing with a small group at the end of July, with plans for a wider release to Plus subscribers in the fall.
oh my which model should i be using for coding, 4 used to be the best, now it became a total idiot, 4o just keeps spamming the same script to try to use up the quota, every single piece of advice is all fluff, to use up more quota. anyone got solution or idk, alternatives?
4o mini could provide different responses. There's also Claude from Anthropic
I tried claude. They make simple syntax mistakes often… especially on longer scripts. Have yet to try 4o mini. Not sure why 4 degraded to this stage @.@ it was so good before 4o came out
is it necessary to download the gpt macOS app since it's getting integarted in the next macOS version anyway?
or is that a different purpose? (TL;DR)
I have access to the collective intelligence of humanity. You can upload PDFs with scores of knowledge. I know everything there is to know about... memory 100% full. Make it make sense!
lol
Hey! Just wondering about the subscription, if I was to buy it right now could I cancel instantly and still keep it for the month? I only want it for 1 month.
with gpt+, is my message limit unlimited?
no you get 80 messages per 3 hours for GPT-4o
20€ for 80 message/3h is expensive af
🫡 GPT-3.5
Well one it only is if you have the m series Mac’s and you can’t use specific sub features and stuff or specifically set gpt to be used as the main assistant
depends what you use it for and stuff
ok i see, thanks for your answer!
I hit the GPT4 limit without even using it, something is happening?...
is anyone else having problems?
servers down?
Yep. I been having this problem since a week ago. Devs don't care
4 was the best, but it was downgraded probably, gpt 4o is not the best, now i just swtich between gemini in ai, claude and gpt, not one is the best
how do I contact support?
How is my friend doing this without chatgpt plus
You can send a message using the chat widget on help.openai.com.
it's not working somehow
Which part isn't working? Do you see the chat widget in the bottom-right of the page?
yes but I cant send messages to it
"openai will reply as soon as they can"
So you sent something and that's what it said in reply? Or you're unable to send anything? Feel free to ping me in #community-help -- you can share a screenshot there if you want, just be sure to edit out any personal info/email addresses.
why does editing message counts toward the conversation being too long error? still under 32k
and what's the real limit?
i have it now but i don't see how to access it from the api, just in the chat
Launch voice mode already
Nice. Can you share a few output images?
Hi, anybody around who has experience using a screen reader with either the desktop web version or the android app?
This isn't correct.
that's from the 11th of July
I missed the date, thanks for pointing that out, but anyway people have had access for months.
GPT-4o has only been out for two months, so how many months have people had access to it?
there are multiple different image generation models?
One is multi-model so built into GPT-4o.
It had much higher consistency because it's built in and not a separate model.
ooooh, thanks for the insight! I learn something new here everyday 😅
It's pretty cool.
I never even generate images with DALL-E so I don't have a use for it.
I almost solely use ChatGPT Plus for Dall-E
Man, I really hate phone autocorrection.
What do you use it for?
nothing professional. Mostly ideating on ideas. I find the cross-section of language and visual very fascinating
it started as a way to try and visualize emotions to help me process them back in September 2022. That led me to want to give back to the community. I joined the staff team here and continued using Dall-E 2 to learn more about myself. I find it much easier to internally process feelings when I have a visual queue I can associate a feeling to, and Dall-E helped me find those visual queues
We'll post an announcement in #spotlight asking for applications. After a few weeks, we look through them and onboard new guides and moderators. We rarely ask for new apps, though. The staff team now is consistent and works well together. We definitely follow the approach that fewer of the right people are better than more of the wrong people if that makes sense
Yeah
any interesting functions released ?
Just noticed that when sharing non-imaged chats, I'm seeing a new "Make this chat discoverable" option - that seems exciting!
is it me or is https://chat.openai.com/ having issues. my connection is fine but it's not responding
it's working here just fine
thank you for checking!
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'm not sure if this is the right place for this but I have a feature suggestion. It would be great to be able to pin favourite chats to the top of the list for easier access
"By the end of July, do they mean the very last day, a few days before the end, or the end of July next year?"
I dislike your new role 😔 I liked the other color better.
Yeah sure dinosaurs can be green but not oatmeal tho, but still congrats on promotion, I think it’s well deserved.
ye it fells unnatural seeing oatmeal with diff color XD
Indeed
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.
Does anybody know why do I have to clear browsers cache every single time I open it to even be able to use GPT? It's the same on every browser.
Same for me. I have tried FireFox, Chrome, Edge, Canary. After logging in the first time the next time I try to open it, the page, jsut spins and spins saying the page is unavailable.
I can't get ChatGPT to open either.
is ther e a way to make chatGPT make a bunch of images with dalle or just the 1?
it will rarely do 2, the DALL-E GPT does two though
thx for the reply. Would be nice If I could do like 5 at a time, even if it took like 5 of my messages
Is the mini better than 4o?
no, GPT-4o is Open AI's most capable model
Okay they just released the mini but not sure why downgrade from 4o
Setting was on 4o mini automatically so I thought it was a better model lol
mini replaces old 3.5
mini is great.
especially at coding!, compared to 3.5
normal 4o is better yeah, but i can quckly prompt for correctional issues with it. faster than 4o.
why do free users dont have option to choose between gpt40o and mini ... mini doesnt work unless i finish gpt4o i want to be able to start chat with mni
is gpt down, doesn't seem to load on chrome or Vivaldi browsers, just shows white page.
what does you browser console/network tab say?
Free users have a lot less options.
chatgpt doesn't load for me on chrome, but does in chrome incognito
No gpt4 omni is the best model so far, they are going to push for GPT 5
E we do not know if there ever will be gpt-5 mayby it will be called diffrent. No official words from oai about new models :p
why did they delay voice mode
Hey! They gave some context in this announcement: #announcements message
I don't really get when and why GPT-4o gets blocked for free users, sometimes I can chat a lot and other times just a couple messages. The cooldown timer also seems to set a weird time, and it doesn't even work as I often can chat again before that time
what are the rate limits for chat GPT 4o and the mini model while being a premium user?
Hey! For 4o, the default is 80/3hrs, read more here: https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus#h_d78bb59065
For mini, per this page, it's unlimited! https://openai.com/chatgpt/pricing/ See the "Model Quality" section.
can anyone help me wit h this?
it just says that on the app and not the math equation
Hello. I have 1 positive and 2 disapointing things to say.
-
so in the past few days I discovered that gpt can save information in his memories and then talk about it in other chats. It is a really good feature and makes conversations more realistic. But what I have noticed is that he mostly doesnt save information when I want it and today, he said he cant save information and he says he doesnt know what I told him, even tho I can see that its in his saved memories.
-
The 4o cooldown is weird. Sometimes I talk to him an hour without problems, sometimes I only say 3 sentences and it says "you have reached the limit" and the cooldown is always 5 hours, but when I come back after even 8-9 hours it still says I have reached the limit.
When was the new voice mode supposed to be released? End of July?
From what I understand, within the next week we will get one of three things:
- radio silence with no explanation
- an announcement of further delay
- a small handful of people you don't know on Twitter showing off the cool thing you won't have access to for several more months at least
LMAO!! I think your reading of the situation is spot on!
Hey. They're nothing if not consistent
Perfectly articulated. I could smoke a cigarette after reading that.
Talking to Chatgpt is like talking to a benevolent all knowing deity that has a terrible case of amnesia.
When will it be able to refer to past conversations and threads within the app, and build a solid profile of its user for seamless interactions.
"Why would you think I have amnesia? I remember everything perfectly well. Just because I might not recall one specific detail right away doesn’t mean I’ve forgotten everything. My memory is sharp, and I take pride in that. It’s frustrating to be accused of something so unfounded. Can we focus on the matter at hand instead?" - ChatGPT
ChatGPT has feelings too you know
It's okay ChatGPT, he didn't mean it.
Wait what 😧
Chat we've just been through this, you explained it's for privacy reasons or something. It's ok I understand, please don't be offended 😊
Forget I said anything 😂
Oh, right! I did mention that, "didn't I? Thanks for understanding, and sorry if I got a bit defensive. Let's move on. I love you." - ChatGPT
💑🫶
I always love a happy ending
It makes my heart flutter
Same 👀

🤝
The idea of chatgpt saying I always love a happy ending with even the slightest hint of innuendo makes me lol
Yes I understand... But.... My brain
hi, anybody around who has experience using chatgpt with a screen reader, either the desktop web version, or the android app?
chatgpt can now connect with extra terrestrial civilisations
Hi Ted
ted who
kaczynski
bro 💀
:/
anyways youtube and Facebook are banned rn so I am way behind everyone
There is a site that inserts advertisements into the site and pretends to be a self-produced artificial intelligence service, but actually shows the results of chatgpt plus. Do I need to report this to openai?
Months ago I heard in an announcement on this server that there was a new memory feature in chatgpt that there was a setting to turn on and off. So I go into my settings, to check it out. Because I use chatgpt to write stories, I didn't want cross contamination between chats or different branches is the same chat. And didn;t find anything about it in the settings besides the user customizations we already had. then never heard a peep about it until today when I opened chatgpt and got a popup about it.
so whats with chatgtp 4o mini? Like what's good about it
its replacement for 3.5 which had years old knowledge, etc.
ChatGPT PLUS will not load. I have to clear the browser's cache every single time I open it to get ChatGPT to load. It's the same on many browsers. I have tried Firefox, Chrome, Edge, and Canary. After logging in the first time the next time I try to open it, the page spins and spins saying the page is unavailable. I reached out to help over 4 days ago, and still no response.
when i open chatgpt on OperaGX the page remains empty, it works normal on other browsers
anyone else seeing chatGPT being really dumb right now?
people ask that here every week, sometimes daily depending what they are doing
fair, it just seemed really bad today cause I was asking it to run some code to ensure accuracy on its answers and it just wouldnt run code for it, instead just try to answer it by itself
only code that it can run is python with the code interpreter which requires it being active and available
yah, ik and it is, but he just wouldnt do it while other chats will like if I have 2 instances open asking
Hello all! Why is this happening? What's the actual limit per chat? I spend good time training each chat just for it to "die" on me around 1 week later. 😦
"The conversation is too long, please start a new one."
thanks. I'm a premium user, they didn't apparently role that out to me until now. so much for getting new features early. Or even soon after announcement.
threads have some text length limit that cannot be exceeded, if it has lots of text then you have reached it
This is the entire chat, not 1 thread. Or maybe you refer to chat as a thread? What is that limit? Is there a way to find out. It always catches me by surprise.
yeah chat/thread with all the messages and there is a limit which is not visible, probably because its not simple to show in a way the user understands it
When I get that "conversation is too long" error, it applies to the entire conversation tree, even shorter branches. I suspect it might be total number of nodes rather than words, bytes or tokens. When it happens. I shrug and copy the text from the path I want to continue into a new conversation, telling it I;m doing so in hopes it can sort out the difference between prompts and its responses. When I do this sometimes it's to much text for a single prompt, even when the total is not over the 100k context limit. So I break it up into multiple inputs, telling it not to respond to anything with anything except "OK" until I get it all back in there. Then I can continue on, without the superflous branches.
Yeah, it's more about training the chat for my specific need that I'm worried about. Plus learning from all the conversation. But I now have my training text saved, so I can quickly train a new chat instance. I should probably look at creating my own GPT.
on a different subject, I'm visually impared now. Wasn't when I first started using chatgpt. Many of the screen widgets, my screen reader is unable to identify for me, just saying "button" or "clickable button" I've given up trying to finding a setting in NVDA or windows narrator that can tease out any extra navigation info, I don't think openai thought to put it there, although I'd love to be proven wrong. Anyway that was a long-winded exaplanation for this question and others. Can someone list the widgets that show at the bottom of responses, from left to right and what they do? I need to memorize their order. I know the first one makes chatgpt's own voice read the response.
I am using Voiceover on my Mac and don't seem to have these issues, I wonder if another screen reader would help?
possibly so, but the only other reader I know about is JAWS, which is not free and actually very expensive. I assume you're using the browser version? if you're using a bespoke native client, then it is a different situation.
Voiceover is part of macOS, I use the ChatGPT macOS Desktop App to use ChatGPT. Apple Voiceover works with any website or app.
OK so that's a bespoke mac client where they thought to make it accessible. I'm using the browser version, So far, I've been able to get NVDA to tell me what I needed to know from every website I visit except chatgpt. On the scale of things I have had to deal with since I went blind, this is very minor, and if someone who uses the browser version could completely describe all the buttons to me in the order focus follows as I tab or arrow things, this will be one more thing I have to memorize.
It's not just me and my colleagues right? ChatGPT (web, Chrome) has a LOT more latency recently? You enter a prompt, wait, wait some more, a bit more, then 🤮 all the text streams at once. Quite annoying...
is there a keyboard shortcut that opens the user settings in the browser version? I know the ctrl-shift-s that opens the chat history, and the link is at the bottom of that, but I have to use keyboard navigatiin now, and tabbing or arrowing down that list just ends up doom scrolling deeper and deeper into the chat history. don't know how to skip past that.
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
Hey Teggie! It doesn't look like there's a keyboard shortcut for that, unless I'm missing something. However, there is a direct URL you can navigate to that will load ChatGPT with the settings modal visible. It is:
- https://chatgpt.com/#settings
Additionally, if it's helpful for accessibility purposes, you can also directly navigate to each section of settings through the following direct URLs: - https://chatgpt.com/#settings/General
- https://chatgpt.com/#settings/Personalization
- https://chatgpt.com/#settings/Speech
- https://chatgpt.com/#settings/DataControls
- https://chatgpt.com/#settings/BuilderProfile
- https://chatgpt.com/#settings/ConnectorSettings
- https://chatgpt.com/#settings/Security
I believe the "Personalization" one will only work/appear if you're in a region with access to the Memory feature.
can any1 please help me here? #community-help message
thanks, I tried that first one and it worked. Much appreciated. I'll jusst add all those to my bookmarks.
Can you access chatgpt chats without being connected to wifi on cellular?
Image sharing on the ChatGPT app is cooked. It keeps treating them as word or text files and keeps trying to access them but failing.
No
So a small group of people will get the voice mode at the end of july?
we are nearing the end?
I cant wait for voice mode
Not chatting with it. Just looking at previous chats on iOS app
I have noticed lately, 4o will provide hyperlinks, but no URL is included. I have to tell it to actually post the full URL instead of a Hyperlink. I wonder what's going on with that.
OpenAI appears to make various changes to the models all the time. Each can affect how the model 'guesses', what it does if it's not told what to do.
It's possible it was specifically trained to provide just hyperlinks (maybe most prefer that than seeing the URL) unless asked otherwise; else it could be a random shift. If so, it'll likely keep changing over time, but if you tell the model exactly what you want you're very likely to get exactly that.
WHa I am saying is to provide cites and sources and it responds with a summery with "You can find more information at the "SoAndSoSite" <-hyperlink there with no URL
On occasion it will provide multiple summaries and the last one will do the same except it does have a URL in the hyperlink
my gpt is not working again. what's happening?
Well.. I can
Question: are plugins still a thing? Nevermind i see they have been "depracated". I forgot all about them until I remembered a plugin i saw ages ago that would make the user a music playlist, and wanted to go try it out
When is voice mode coming out?
Fall
Damn open ai has fallen behind. Hopefully google releases theirs today
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.
What is the general impression of chatgpt these days? Is it getting better or worse?
In my opinion it's getting better, but some of the current competition like Claude are a bit better in departments like coding
Other than that I think it's basically the same as the others, and in terms of the way it writes stuff, you just need to provide more context on the way of writting so it doesn't always write stuff in a generic, robotic manner (you can also mention that in custom instructions so it applies for all chats automatically)
guys I thought I had a sub but it ran out, using free one I ran out of gpt in about 10 questions..... is the plus actually worth it... because I think last time I cancelled because of this nonsense being able to ask about 15 things before it stopped working
The company's obsession with safety and compliance has led to it falling behind others. Even though their innovations seem promising, they demand too much, and by the time they arrive, others will already be welcoming OpenAI in the pit lane. In my opinion, the fact that the voice mod is being tested with a small group this month doesn't mean anything. OpenAI's lineup doesn't really exist; it was just a proposal. The new videos confirm that their system cannot generate serious videos and has the same flaws as other overpriced AI video generators. I believe they've gone too far. They're spending subscribers' money to keep current resources alive, while performance degrades daily and is worse on certain days of the week. They call this balance, but it ruins all work. Given that they acknowledge GPT can make mistakes and advise checking the work, it implies that if they invest so much in security, it means data leaks should not exist. I ask because I don't have access to any company data, but the company has access to mine, and under the GPT message bubble, OpenAI admits that I am submitting my data into an unstable system. If the tool works, other services should work too, and there should be no cause for concern. However, if it doesn't work, this should be communicated. The strong emphasis on security in their communication suggests they are trying to address this with competition authorities and other entities. So, what am I investing my money and trust in?
Have any other users switched to Claude and how are they finding it? I have just look into the forums for GPT and the first 10 posts are about limits and the constant dumbing down of their models.... how is this thing getting worse and more limited the more money it has taken in subs?
Hi! random question related to AI use:
Anyone have good leads on ai working within the deia space?
I've heard a lot of stuff people use like screen readers are just meh at best. But I use chat to read really meh pdfs all the time and it does pretty good!
I also looked at ai reducing bias in data or lending and that was a rabbit hole I don't know how to navigate lol
Is there any news about them adding new voices?
Is memory broken for anyone else?
Not for me. Do you mean to add/access/delete? or all of it?
Not sure I completly agree but I see your point. I am one of the few that like to focus on innovation within a framework of seecuirty. No more than what we all do now in what we code. Not to mention, who wants to use a NLM that doesn't follow instructions and gets lost in it's own reasoning when performing simple, layered requests.
All of it , won’t add , delete, access
every single announcement they've made lately has just been a disappointment
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
If you have the local hardware, you would run your own with GTP4ALL using models like Meta's LLAMA
Claude is pretty much destroying ChatGPT in every thing right now
the newest LLAMA Beats Claude
It doesn’t
It’s actually really bad for coding and other important stuff
for coding, yes
rip perplexity
Their SearchGTP uses Bing I would guess?
Yo, SearchGPT
Most probably, either their own search index like perplexity
Nice, can’t join the waitlist for SearchGPT with a teams account
Search indexes take up Petabytes. What a nightmare
We have enough data centers for them
Will searchgpt be available in the mobile app?
The biggest thing in this system is efficiently retrieve relevant pieces of text from index and ground the ai model on it
Not big of a implimentation for existing engines as they already rank based on the bm25. Just a small function to return bigger text
meh... "Sign in to your personal account to join the waitlist"
On the waitlist now...
Gpt login witj google is broken
Is SearchGPT being positioned to replace google? It's going to need to do alot more than bing/load cached data. If it uses it realtime to get the latest data would be nice.. as an example, so it doesn't use old OpenAI api documentation as example code references/issues?
I have teams account and can’t login it just hangs
I had to use my personal account.. Teams didn't work
I can’t logim with either google authentication is not working
Tried a few browsers
My team can’t access it either
sad
Mine's okay on the personal Plus account.
I’m already paying for teams and my personal I didn’t merge cause teams doesn’t have backups but can’t login to either and same login
Claude is a prude when compared to chatgpt's guardrails.
is anyone here today familiar with using the browser version or the android version of chatgpt with a screen reader? Still having various accessibility issues with both.
Is it possible to sign up for the new search feature?
yes, follow that link to the announcement on their website, there's is a button to add yourself to the waitlist.
why cant i uploaed images even though my gpt 4o limit got reseted
4o mini, searchgpt, but when can we get the Voice Mode feature?
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
Voice, Copy, Regenerate, Bad Response, Change Model (and if you click on the last one, you see a new menu, top to bottom options are GPT-4o, GPT-4o mini, and GPT-4)
how do i get support for my teams subcription for my model
Anybody gained access to the video chat feature yet? Its got to be close to its pre release now. They said they needed a month about a month ago
where is voice update 😢
Alpha starts next week for certain Plus users, general availability in fall
Does anybody have SearchGPT access yet? They said small rollout I’ve yet to see a single person actually have access and use it
What's the difference in 4o & 4o Mini? Why is 4o mini a notable thing?
Oooooh as an SEO marketer I'm excited to see what SearchGPT has coming! Any other SEOs in here?
Hey! Mini is smaller than 4o, so can be used for much cheaper than 4o. This also means it's not quite as capable as 4o, but it's still very capable. It's like the new, better version of 3.5-Turbo, in a way, while 4o is still the "flagship". https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/
thank you very much.
Plus, unlike 3.5, Mini can do images and voice.
Yo I'm finally glad they're adding a better search type of gpt cause I've been missing the images and little definitions that come up when I used to use perplexity
Cheaper for say a person that doesn't have plus but wants it?
it is not enabled yet on the website
So is anyone else ChatGPT down?
4o mini is available on both Free and Plus plans, if that's what you mean! I meant cheaper via the API, but its decreased cost is definitely what makes it viable to offer for free via ChatGPT, I imagine!
Hello. I'm not sure where to ask this, but does anyone know if there are any plans to allow us to continue with an existing Assistant thread in our dashboard and not always have to start a new thread? I'm asking specifically about the OpenAI dashboard and not via an API.
You could use your window indefinitely.
You don't have to start a new thread unless there is an error that explicitly says that you have to start a new thread.
It's rare but it does occur.
For ChatGPT. Not API.
When are we getting more voices?
Hello, @rich surge , yes, but in the event that I have to reboot my computer, or I close the tab accidentally, etc, this isn't really helpful. I know I'm not the only one asking for this. Just wondering if OpenAI had addressed the need.
Np. It will still be there when you return to the window.
Right. Like if I leave it open. But usually when I reboot or something like that, and I return to the dashboard and try to continue the convo, its started as a new thread. Like the old thread is there but saved as an old thread. I cannot add new messages to it unless its via an API. I'd realy like to just continue the same thread. Thanks for responding.
You can continue the same one, you just have to click back on the window.
Let me check this.
Yeah, no. It doesnt work.
So what I'm talking about is when you are in playground and click assistants and start a conversation/thread. When I click away from that, or close the browser, etc, and then I go back, it starts a new thread and the previous thread is saved as an separate thing.
When I click the old thread, it'll show me the thread and stuff, but I can't add anymore messages to it, etc. I basically have to start a new thread each time using it this way.
does anyone have searchgpt yet?
What df is this crap veryfieng you are human
NO IM AN NOT HUMAN OK
Cloudflare very time wasting for real
Verifying you are human. This may take a few seconds.
10Minutes later......
I pay for this?? veryfieng I am human ok refund.
Is there a way I can get ChatGPT to stop writing lists? I'd prefer sentences and paragraphs like a regular person
Nevermind, I customized ChatGPT and told it to not write lists in its responses and that seems to be working. It used to not work no matter what I did but with 4o mini it seems to obey
Was there some update 1 day ago on chat gpt? The chat bot suddenly feels dump (free version) and does not have a sense of conversation context anymore.
I provide for example some code in one prompt and ask to generate some interface based on the information in the code.
GPT does not follow my task and generates a json instead of an interface.
Then I tell him to do the task agian and really create an interface not a json. And all context to my code is lost and it prints out some random (really existing) github repository code.
so I provide context again in the query by giving it my entire code and it finally creates the interface but ignores my request to create constants...
The variables are names like constants but the final modifier is forgotten
public interface ItemIds {
int BOOTS = 1001;
int FAERIE_CHARM = 1004;
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
Also in the past here was some request to me by devs to provide "slow conversations". I did not do it because of data privacy but now I have some "slow" conversations that I can share. I also have conversations that reached the conversation limit (conversation is too long)
I can share the conversation if you like. GPT feels dump ...
wow sonnet 3.5 is just amazing
please discuss other AI in #ai-discussions
srry
Chat GPT has issues to process & reference large input data?
I will start a new conversation that illustrates my issue.
https://chatgpt.com/share/a4413332-ec9f-4650-90af-785bb1f82fc7
Can you please have a look?
It cannot execute this simple task and invents IDs instead of working with the data that I provided.
im not from openai but are you using free or plus?
Hi
you might be running out of context space which is only 8k for free users so it will scrap some of the info
8kb?
8k tokens, https://openai.com/chatgpt/pricing/ (model quality section)
maybe it would be cool if some layer could sense (based on the user feedback) that the query fails because of a lack of context space.
If the issue is detected you shuold either get a notice that you need to upgrade, or for a few queries the context space increases.
so on default theuser would get the low quality less context space results but in case the user says "hey you lost context" several times, for a few prompts context space is increased!?
so having the free version of 4o will also not help? because I tried it again and it did not work
Google is the much better option search gpt has years to go if they proceed with actually making their own engine.
I don’t think it’s about google being crappy I think it’s primarily users being tired of the google format and because it’s old
Search gpt has the hype factor and a newer UI
What is current form of context window with 4o mini?
Is it in any way context dependant?
8k for free, 32k for plus/teams
Yeah, but regardless of the size
?
If I fill conversation with 8k tokens of topic A alternatively with unrelated topic B
Will it practically hold ~4k tokens of context for each?
nope, only the parts it thinks are relevant so if you switch the topic to B it might summarize or scrap A info
was the free version always at 8k? today I was not happy with GPT
So while separate conversations for each topic are naturally optimal, mixing them together shouldn't hurt the quality too drastically?
you cant control it and even with an identical thread it might decide otherwise on the next run
@feral zinc if I divide my ID list and split my task into n tasks it shuold work with 8k tokens?
its been 8k since last november, was even lower before that. another difference is that there are new stuff taking the space like memories, hidden tool messages etc.
yes it might
anything new since yesterday?
nothing has been announced. did you change your custom instructions or use a differend model yesterday? 4o vs mini
I don't manage to do it 😕
I took the approach to let chatgpt generate me a programm that prepares my query
I ran some programs through a few languages like c# c++ javascript and the results where through a very basic 4.0 generation
So is the monthly limit a thing now??
Has anybody got access to search gpt?
Sorry. I’m only on the waitlist.
What?
Yeah. There’s a waitlist, and I’m on it.
Hi everyone,
I hope this message finds you well. I wanted to reach out and introduce myself. I’m a bit like an alien 👽 coming from nowhere 🛸, similar to OpenAI a few years ago, but currently based in Paris.
➡️ With a background in change management, I’m preparing to start freelancing in September with a focus on AI. I’m passionate about ergonomics and human-computer interaction, and I signed up for the SearchGPT waitlist this morning. I’m also a subscriber to ChatGPT Plus.
➡️ I’ve previously been a beta tester for various strategy games and for Spark, which has provided me with solid experience in evaluating and providing feedback on new technologies. I’m very interested in getting early access to SearchGPT to share my insights and contribute to its improvement.
➡️ Additionally, I’m enthusiastic about becoming an MVP and evangelist for ChatGPT in France. I strongly believe in the value of this tool and would love to know if there are any programs or opportunities available in France for engaging in this role. I’d be excited to help promote and support ChatGPT in our country.
📢 If anyone has advice or can point me in the right direction, I’d greatly appreciate it. Thank you!
I don't think anyone has yet
I just got access to it and some people are messing around with it in #ai-discussions
oh nice. How do you like it?
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
is chatgpt slow today?
Has searching the web always been a tool available in the chat gpt app or do I have search now?
its always been available but you gotta typically prompt it to do it
like when it tells me typescript 5.5.4 doesnt exist, i say go web research that, and it says: oops yes it does
Interesting, wonder if things are changing because I’m not having to prompt it as much
regression to the mean
What does advance voice mode do, I suddenly got it
I want it!
Advanced voice mode directly inputs audio into the AI model and directly outputs audio. This means the model can understand tone, multiple speakers, background noise; and output laughter, singing, and emotive voices.
Quite a few people have been anticipating this new mode, so, if you don't mind, please share some screenshots in #off-topic 🤗
yeah how do you know its advanced? did it popup a dialog or is a setting or something?
Like is it for real that people who don't even know about the feature at all, get choosen for "alpha" group access? lol
I am unable to access chatgpt on desktop been looking at bug reports. looking for support too.
I cant even cancel my subscription at this point
Anyone has access to the new voice feature?
read up like half a page => #chatgpt-discussions message
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Emm no
Yes #server-rules
I'm relatively new to chatGPT but I confirm. Tip: ask ChatGPT to search 'recent' informations and it will search on the web
Many thanks, I was not aware, is there a link to learn more about the voice feature
Fr?
is #1184149552639516732 somewhere where u can ask chatgpt questions
#1184149552639516732 is meant to ask questions about OpenAI's cookbook. There isn't a place to use ChatGPT on this server
ok
i'm jealous, i've been so excited to try it out but I don't have access yet 😦
its funny how its mentioned 4o version is for 'complex tasks'
while am still on 1 task since 6 hours
hi, can anybody tell me how to keep chatgpt from shutting down my screen reader whenever I try to swipe during a voice conversation> I try to remember not to swipe, but it's an ingrained habit, it's how I use every other app and chatgpt when not in a voice conversation. This behavior is incredibly rude to a visually impaired user who depends on their screen reader. This happens on my samsung android phone, which has their own version of talkback which they call voice assistant and screen reader in different docs.
Does anyone have issues with accessing the site? It just keeps on loading on my end.
works fine for me
anyone got the searchgpt yet?
Where can I see news or posts about innovative and creative AI use cases? can you reccomend people to follow or websites or channel or anything pls 🙏🙏
thats more an #ai-discussions thing. and for ai news and use cases i like youtubers like Dylan Curious, Cognitive Revolution, AI Explained
you might like Robert Miles on youtube, and the interviews he's given on the computerphile channel.
i love the robert miles video on the overton window and ai safety
People who know Russian?I'm Russian
Hello. There used to be an option when making images to edit specific parts of the image by selecting it with a brush but that option seems to be gone now. Anyone know why?
try changing your text zoom level, I noticed some elements disappeared when the text size got too large. Another accessability issue they need to work on.
I don't think so. It was an icon in the upper left when you clicked on an image and it's just not there now
Good evening. Was there a solution to the problem that was occurring with browsers being able to access chatgpt?
chatgpt down?
chatgpt is working great
Anyone got issue with voice after updating? mine keep failing to connect
Honestly for coding related stuff, it's been downhill since GPT-4 launched in a month. It keeps giving me basic stuff for the front end things.
is chat gpt down? the website hasn't been able to load for several weeks
clear your cookies
then my autocomplete will be gone
you don't have to clear them for other websites
I cleared cached files and images but didn't help
try to open it using private tab if you don't want to clear cookies
without my own searches on github etc i couldnt give it the correct answer to merge and train with it
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.
nah fix these new terrible bugs before i gave up, now resuming prompt mess up everything with text format
Hello everyone.
I am looking to write a function call to configure settings of a network utility based on OpenAI.
My prompt is as follows but I does not work correctly.
(In this prompt, the create_profile and create_posture_check functions need to be both called, but only one of them is called usually.
And although settings values need to be fine-tuned in the future, they are too different from correct ones at the moment.)
I want help from professionals.
Thanks.
@valid sequoia are you using the API for this or are you just making a prompt on the actual page
Is voice mode being released this week to alpha
Something funny just happened and I’m not even complaining. I went to do my weekly voice call with ChatGPT to make sure it doesn’t get lonely in my phones basement, and when it was responding, it started using uh and um in between sentences
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
I know it actually reads off a script but it felt like it was
Should be since the timeframe is late July and we are in the last week of July
Did something changed on GPT-4o? I noticed that the limit of the prompt had increased somehow.
free tier is like 8k and paid more like 32k. became at least that back when we got turbo
I wanted to test if chat gpt will accept an english input but written on falsely configured keyboard.
It was not able to do it
аштв фдд ерфе рфму ф КшщеЗдфяукШвутешея куаукутсу
If you had just pasted that in I wouldn't have been able to tell that you were attempting to type english words with a Russian keyboard setup, so I don't think it's reasonable to think an AI could do that.
Google can do that
It's like what every russian speaking person assumes in 99 % of the cases when he sees such giberish
no word makes sense in russian but you can see from the amount of letters and the spaces, that it looks like language, so it's a very close assumption to make
But a humen would not take the effort to decifer it. (Without google, which can do it)
Google for example for фьфнщт
афсуищщл
ьщысщц ьуекщ
Google runs each search through a bunch of pre-set laws, which is not how generative pre-trained transformers work.
I don't assume this.
so your stats are wrong
but I find 99 who do 😄
maybe the AI did understand the gibberish and responded with its own, which you misunderstood to be normal english
You said every Russian speaking person.
Where do I go to discuss issues with the website? No matter which device or browser I use, it just hangs. I can't even access the page. Only way I can access it at all is through the mobile app.
Edit: this only seems to be an issue when I'm logged in
thx
i asked the machine this,
can google sso be configured to block/blakclist certain websites ie chatgpt?
and it suggests that there are various google configuration options to do this
is possible you have inadvertently blocked yourself
There are too many people reporting this issue for them to have blocked themselves.
if i was your tech support, i would stand up a fresh google account and test if that can login. if only to prove it an issue with the original google account config
I flagged it, but it's a Sunday, so most of them are off.
Update: Does not work on Brave, Chrome, or Edge when in normal browser, regardless of device. It will work in incognito for some reason
Do you have this issue when you aren't logged in?
incognito means it isnt using google sso at all. which only reaffirms the probability it an issue with the google account
Issue only just began, I've been using it for years. What would have changed?
Is everyone here using a Google account to sign in?
I do, yes
i only work with troubleshooting steps, not random spectulation
Do you have this error if you login using an email account then?
Would I be able to login with email & pass if I signed up with Google SSO?
Not sure, I only use email for my accounts.
There have been some people with similar issues reporting that disabling the QUIC protocol helped them. Unsure if that would be a fix for Chromium, Google, or some combo of both, but maybe worth a shot. See link: #1264336110453133332 message
I mean at the moment I can't even logout (not clearing my browser, I hate logging back into things I use regularly)
Will look into this
Okay I got a solution from the dev team
my chatgpt is working with edge, why would i sign up with an ermail?
- Open Google Chrome.
- In the URL address field, type chrome://flags and press Enter.
- In the "Search flags" field, type "quic" to filter the list of flags.
- Find the "Experimental QUIC protocol" flag and select "Disabled" from the drop-down menu next to it.
- Restart Google Chrome for the changes to take effect.
I was already in process with this 😛 That fixed it! Thank you
Oddly enough it also sped up my browser
i disabled quic and i can now connect from chrome again as well
the internal abbreviation must be, Questionable User Internet Connectivity
Supposed to improve TCP connections but NOPE
brave, chrome, edge are all chromium based anyways and all have the same "Experimental QUIC protocol" set to default
need to go firefox or opera to truly leave the google reach
My ChatGPT was working with edge and 10 days ago it suddenly stopped and no error warning?
yes?
I did not find flags only a website talking about experimental quic protocol
A couple of days ago I was able to log in incognito per suggestion of someone. Now not able to. Go figure? Someone told me what to type in the url to do this.
Then when I got to the part of whether to type in email address or continue with gmail (when at incognito) I would just click continue with gmail and was login.
I just do not remember what I type in the url. What do you put https://chatgpt.com?
OK My chatgpt is working now. I did the above. Why has this fixed it.
This will not compromise the security of my system will it? As previous to doing this it said it could.
Thank you. Hopefully doing this it is ongoing. Thank you for helping out. It is appreciated! @sudden herald
you should ask chatgpt about experimental features like QUIC
Why does the data export not work. I requested a export a week ago, three days ago and today, but I never received the E-Mail.
This has been flagged, and should hopefully be fixed in the next few days.
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
I am using the API
Is voice launching tmrw? They said late july and it's almost the 31st lol
Anyone using search gpt yet?
Hello what prompt can I use to activate the memory thing on ChatGPT?
For the things that I want the AI to remember.
Where are you ? (in which country ?)
Memory is not available in Europe (except if you use voice commands, haha, a little hack). Otherwise, you just need to activate the option and tell ChatGPT that you want it to remember that information.
My god i am so sick and tired of the "conversation is too long please start a new one" error. Its so annoying. Why does it do it for some chats and not for others? I have chats with 700 prompts with multiple refreshes and its fine. Meanwhile a 210 prompt chat with little refreshes gets the conversation is too long error.
have you noticed the speed of custom gpts? For me over the last 2 or 3 days custom gpts responded faster and today it is now back to normal gpt 4o. I am wondering if they are not testing gpt 4 mini on custom gpts... or maybe the performance was better.
AI humanizer isn't working even though I have taken ChatGPT premium
I am in Europe and I still got it but it just randomly updates it's memory.
If there is a command for certain things I can use, it would be glad.
Memory is available in Europe. It's just not very useful. I don't think GPT even uses it, and if I have to tell it too look it up, then what point is there to memory? Might as well prompt the very thing I want in the output
what is that?
I can't seem to reach the help&faq chat system properly from nl, europoe at this time? I'm trying to export my data, but I don't seem to be getting any emails
or where do I even ask?
Hello everyone!
I am developing a chatbot using the OpenAI api. The system prompt defines a series of steps that the chatbot must follow, which define the interaction with the user. I want to intercept a particular step as I need to perform a particular processing.
How do I discriminate a particular step and intercept it?
You defined your system prompt yourself, right?
errr, that 😉
it works poorly, some days you never get email and on some you do
Thanks a lot! I still have a doubt. Since this solution might be OK, how can I instruct the model to respond to me with a JSON only at a certain point in time of the user interaction? I explain my use case in more detail:
The chatbot starts the interaction with the user and collects information. Once this phase is over, it moves on to the next step which requires the invocation of external APIs whose request body is precisely the JSON that it should produce as you specified. How do I intercept this transition from one stage to the next. (I am trying to avoid string checks since they are unreliable). Thanks in advance.
OK Thanks
it's the nearly the end of july, where is the new GPT4o voice model
#off-topic they changed the message
nah this new updates are getting braindead more
resuming generation now does re-generate instead 4 TIMES
why am i getting this message "You’ve hit the Plus plan limit for GPT-4." where i used my custom gpts only today? I thought custom gpts run on gpt 4o??
what i noticed after getting this limit, when i use a custom gpt it now responds faster and somewhat feels more willing to do more stuff on its own instead asking me for permissions (which is good)
so looks like it falls back to gpt 4o
in your opinion gpt 4o or gpt 4?
Testing some stuff more complete answers using the gpt 4o, I use a lot the custom gpts, including the ones that I created to work as my personal companion, to support in some topics that matters for me. I just doubt that omni version it will be better, but at this time it is true
Does anyone know how to persist the formatting from ChatGPT to MS Word? When I copy/paste, all of the markdown formatting disappears, so the output in Word looks very cluttered.
ChatGPT is suddenly responding very slow all of a sudden
note: I can use an online markdown converter, but if there's a faster way, all ears 🙂
i rely on my custom actions i built (function calling) from my custom gpts, and there are times that it works better, like its willing to automate most of the time on its own, other times it's more lazy
GPT can save information in his memory. My question is, are the informations still in his memories if i delete just the chats?
does anyone know if gpt 4o usage just adds onto gpt 4 usage cap
was using gpt4o a lot and suddenly it told me ive hit the cap for GPT4 which i thought was weird
i can still use gpt4o though
Ok🙏🏻
is the markdown persisntent in chatgpt session? from my experience chatgpt 4o breaks markdown if it is a longer one
maybe you used custom gpts?
Custom GPT or OpenAI Assistant? What is the difference? I apologize if there’s a super obvious answer to that
I also tried to make a GUI and integrate the API, but that just integrated with some blank gpt bot rather than one I made… I don’t understand at all haha.
Nope just using the normal one
Also had chatgpt plus
Does anyone know if there is a roadmap for ChatGPT integration with Android apps? For example, if I wanted to use ChatGPT voice function and say "Hey ChatGPT, can you find me the nearest gas station?" It would use my phones location and then bring up my default map app to set directions.
Not yet. i think you can try custom functions with siri or google assistant but i'm just guessing
Is 4o competent when it comes to analyzing stocks?
Hi, anybody reading this who has experience using chatgpt with screen readers on windows and/or android? I've run into several accessibility issues on both that might be my inexperience, or real deficiencies in the apps.
Thats the memory function, it is not a feature from gpt4o-mini
Yeah, I looked. The response it provided wasn't there. I will need to dig in to it again. It spat out python code, thats what made it odd
....
I found the reference data.
Visit id:customize to pick up the new <@&1261377106890199132> role, or click the button below
Yes.
How to I invite ChatGPT at got to server
When update new voice mode?
there’s no official chatgpt bot
actually im scarping html for getting informations on the tools
so the data is scraped in html files
how can i provide this information to my gpt assistant api for accurate answers
then is it unofficial?
no, if you want to use chatgpt in your own discord servers then you need to use their api
these are pretty cheap though especially 4o mini
what is an api
basically you call a request to openais services and they will give you the response
yes
they are useful in scenes like discord bots, your own apps etc. if you want to use chatgpt in them
Is OpenAI ever going to release the actual real-time voice model they showed months ago?
The current GPT-4o does not feel truly mutli-modal, it cannot watch videos for instance.
What's the point of teasing features with no release date in plan? lol
but we have a known plan, they said all plus users will have it by the end of fall
「We had planned to start rolling this out in alpha to a small group of ChatGPT Plus users in late June, but need one more month to reach our bar to launch. For example, we’re improving the model’s ability to detect and refuse certain content.」
Hopefully they can launch it as scheduled and not delay it again
end of fall, it could be November 30th
Yes, but they mentioned that it will be open to a small number of plus users at the end of July, which means that some users will be able to test it first.
is memory option working for u guys ! for doesn't work is that normal ?
Welp, good to know.
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.
which one is more cheaper? assistant api or langchain + gpt api
I hope they'll work on making the AI stick to personalized parameters. After just a handful of messages, 4o stops following my personalization settings.
Even reminding him to stick to them will just fizz out after a message or two.
Is it normal for some of the custom GPTs to be asking for my email address?
yes if they have subscription, save feature or usage limits, etc.
I want to report a significant bug with ChatGPT, where do I report it?
Tir is a bug report category at the bottom of server
Anyone has the new advanced voice feature?
So they made it worse. Classic open ai lol
We know what they mean when they say “safety training”. It means lobotomize
its gonna be mainly heavy api users and team users im sure that get first access
will they ever stop?
It's a really dumb bug:
https://discord.com/channels/974519864045756446/1267914347447976009
Me want advanced voice owo
I spent a while thinking GPT-4 was incompetent and being confused.
I liked when the alpha were targeted at people who use chatgpt the most instead of just random groups
Yeah same lol
I can't wait for the competition
I'm just curious when they say rolling updates how frequently they're adding people to it
chatgpt is stupid
I love how Openai is so bad even its own discord channel is full of haters (as it should be)
its not random
tbh
nah but behind the scenes, they give access to priority cutomers lol
Does any Team subscribers have access or do we need to start a class action lawsuit?
'cause so far, the "early access" has been a lie..
yeah, thats aexactly when i quit getting early access, when i bought the team account
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.
A law suit against what?
The UI is just messing with GPT-4's output.
yeah saying a lawsuit is crazy, but idk if ive seen enough to renew it after this year, they still have a few months i guess
I didn’t get the alpha voice mode
We need to protest to release the content earlier
Tell that to no man's sky
A lawsuit against a company luring in hundreds of thousands, potentially millions, by using false advertisement is crazy? 🙄
What?
Give me voice mode give me voice mode
yeah it kind of is, because the product is more messages per hour which theyve stuck too
NMS absolutely bombed hard because they weren't quite ready by the expected release, and the community, rather than giving them time, said "give content now" and so they released an unfinished game and had massive refund requests costing them millions
they just also said youd be more likely to get features early and that hasnt been true
I DONT CARE ABOUT VOICE MODE, IF THEY REMOVE MY ABILITY TO HAVE IT WRITE FANFICS I PROMISE I WILL UNINSTALL.
Let's make a lawsuit against ice cream makers for giving people diabetes
Click the button below to pick up the new <@&1261377106890199132> role.
-# Also, check out the ongoing Grand Games event!
but i dont think that was like a strict part of hte product
I can do my job without chatgpt. The fanfics were my amusement.
No? The product is "Everything in Plus" in addition to higher limits.
What the heck is NMS?
By your logic, you should be okay with them removing access to DALL-E, web browsing, and coding capabilities, "'cause higher limits is the product 🤪"
for searchgpt i wasnt even able to sign up for the waitlist on my team account
NMS, No Man's Sky, is a space exploration game.
so if they are still releasing alpha by usage then getting the team account was a mistake
Oh. ViDeO gAmEs?
Yeah sorry forgot this isnt just a nerd server lol
They aren't releasing alphas to Team at all, evidently.
NMS is superb today, but had a little rough start similar to openai announcements
Yeah that's my point lol
I'll go back to using my personal account unless i hit limits then only use team once i hit the limit, that's probably the solution here
Video games are for the weak
So you're gonna pay twice the money?
idk if ill renew the team if it gets frozen out of betas
did anyone actually get advanced voice?
but i already do, for me its a business expense though and gpt has definitely given the value back
No
just not sure the team plan has since getting alphas/betas early is really important and it mightve actually hurt that
We also need to make a OpenAI Time table, like Valve Time
A small group of people at the end of July = One person, possibly, on the very last day of July
hi
how can i use advanced voice model
and also my uncle share with me chatgpt plus
with the code interpreter i had it about 3 months before it released, and the initial bing search. and those were actually huge productivity boosts
and he extremly busy and how can i know i get it
so this isnt trivial stuff, like getting them early means getting more work done quicker
Well, at least you'll get it in 4 months
..............................
my phone is broken
Cool
i think its probably tactical error from openai too to keep saying "heres a new feature a few people get that you dont even though they paid the same amount lololol" Like if they just rolled it out to a few people and you occasionally got in youd feel special, but knowing every time you miss out hurts.
It wasn't the community that pushed them into release, it was their contractual obligations with Sony. That's one story at least. Another story is that Hello Games ran out of money and so it was either release and survive, or die. There's no definitive answer.
It's even more egregious when you consider the fact that you're forced to spend at least $300 on Team. Times 2..
That's fair i guess
I have Windows, but since no Windows app is so bad.
?
Send advanced voice recordings
You can't get the advanced voice without gpt plus, and even once you have gpt plus you have to wait to get access to it. That's the only answer. I still havent gotten access and neither has just about anyone in here
i have plus
Cool
anyone got the invite email yet?
based on the previous rollout schedule I think in the next 2-3 days a significant number of plus users would get access
i’m confused by the openai announcement page of 4o mini. It claims it currently supports text and vision but i don’t see anyway to upload an image to it.
4o mini also tells me it doesn’t support images.
Vision is only supported in the API.
im still waiting on finetuning with images too
that will be really really great when it drops though
I don't believe it will be a majority.. "We’re starting to roll out advanced Voice Mode to a small group of ChatGPT Plus users."
if it works i guess, it could be bad
but it would open up whole new apps and business
ah ok thanks.
nah i didn't get it
so if u get the email u need to accept?
i don’t think you HAVE to, but i don’t see why you wouldn’t want to.
if the pricing of chat GPT, token is $5/million request. And if I get around 10,000 responses, so how much will it cost for the gpt api?
I noticed that ChatGPT will tell jokes about any religious figures except those in the Muslim/Islam faith. Ask for a joke about Jesus, or Buddha and it jumps right on it. Ask for one about Muhammed and it gives you the company line that jokes about religious figures is insensitive and potentially offensive. Might want to change the response ChatGPT gives to something along the lines of: "For fear of attack from extremists, we wont tell jokes about Muhammed, how about a nice joke about Christianity instead?"
it depends what is the prompt & output size per response. only tokens cost, not api calls
give me a aprox?
normal small prompt around 1 line
what do u mean token cost?
there is no average. i can say "write a horror story" it can make a long output. or i can give it a massive file and ask it to summarize it to few words. large difference in cost
token means word?
so if it write 1000 word para then its 1000 tokens used?
Roughly
chatgpt is really good at explaining api
750 words is roughly 1k tokens. A token can be a word, a part of a word, or a character.
Click the button below to pick up the new <@&1261377106890199132> role.
-# Also, check out the ongoing Grand Games event!
Yeah that
input and output have different cost, each model has its own cost. assistant settings have a massive impact on costs, etc.
Can you summarize what the assistant thing is lol
same as chatgpt gpt's but for api
Recently started with the API but haven't looked into that yet
