community-help

[charred parcel]

are you using any third party browser extensions?

[rugged wing]

Nope

[charred parcel]

what browser and OS are you running?

[rugged wing]

Windows 11 and the Brave Browser

[charred parcel]

could you please try with a fresh install of Firefox and tell me if the message goes away?

[rugged wing]

Will do!

[sleek warren]

Firefox is linux, brave will do fine for windows

[charred parcel]

Brave is a Chromium based browser, we ask to check with Firefox because it uses Gecko

[honest vapor]

hey folks, sorry if this question has already been asked before. I see the following limits on my account. Does this mean that if I use 0125 and 1106 in parallel for separate use cases, I would have 5M and 10K on each of them? The documentation seems to indicate this, but I wanted to be sure I can believe what I'm reading (I'm new here so if this question should go somewhere else, please let me know)

[charred parcel]

try asking in #dev-chat

[honest vapor]

Thank you, will do

[carmine vector]

Do I really have a ChatGPT 4o limit as a premium user?

I hit a limit while making a ChatGPT bot, that's ridiculous??

[charred parcel]

the limit for GPT-4o is 80 messages per 3 hours

[muted thistle]

I was getting the same message on my laptop last night and I don't use a VPN. Was working fine on my other devices.

[wise echo]

anything i can do?

[carmine vector]

Crazy

[restive root]

What libraries do you all use to parse unstructured data into your vector stores?

I primarily work with a diverse range of PDFs and Word documents. A few months ago, I built my own pipeline for this, but it's quite temperamental. Before I invest time in upgrading it, I wanted to check if there are any good off-the-shelf solutions available.

I'm looking for something that can parse paragraphs into chunks and also capture metadata such as paragraph headings, section headings, page numbers, etc.

I prefer open-source solutions but am also open to considering paid options.

[rancid condor]

Hi, can anyone help me with my problem regarding graphic products with Dall-e? A platform is rejacting my graphics. They want me to provide proof that I have the image rights. Does anyone know how I can get this proof? I have a Plus account.
Please write me a direct message as I work with translation program. There is a lot at stake here and I want to be sure I am doing the right thing.

[timber shadow]

Hey! You'll probably want to look here: https://openai.com/policies/terms-of-use/ Specifically the Content section. Input/output ownership is outlined there!

[rancid condor]

Thank you for tellig. I will read and hope....

[rugged wing]

whats the best way to retain history with the API? im using nodejs openai module but im not sure how to retain information and the docsdoesnt explain

[timber shadow]

Hey! If you mean for a single chat, you'll just have to configure your API setup such that the entire conversation gets sent to the model in context with each successive message. You can continue like this until you reach the token limit of the model.

If you're talking more generally, then you might want to look into RAG: https://platform.openai.com/docs/guides/optimizing-llm-accuracy/llm-optimization-context

[rugged wing]

so i've merged it with a discord bot, and each message with !gpt command in front of it is logged to a database along with the ID, for seperate histories. e.g: i have it sorted with the created at so its filtered in order so it can use the correct order of messages. i've tried adding it along with the prompt but that doesnt seem to work

i'll check out the link first

[timber shadow]

What are you sending to the model as input? Is it configured to send both the message starting with !gpt, and all the messages from the same user up to a certain #/token count or something?

[rugged wing]

so for the token count its kinda unlimited, it literally just does this:

async function chatWithAssistant(userId, userMessage) {
    const conversation_history = await getChatHistory(userId);

    conversation_history.push({ role: "user", content: userMessage });

    const response = await openai.chat.completions.create({
        model: "gpt-3.5-turbo",
        messages: conversation_history,
    });
    const assistantMessage = response.choices[0].message.content;

    await saveMessage(userId, 'user', userMessage);
    await saveMessage(userId, 'assistant', assistantMessage);

    return assistantMessage;
}
``` i hadn't realised you could limit token amount, which is actually really useful and i'll look into that. 
getChatHistory ofc grabs all info from the database, and saveMessage saves to the database. the message is just an array of message

getChatHistory grabs all the messages as:
``return chatHistory.map(message => ({ role: message.role, content: message.content }));``

btw it queries all messages from chatHIstory with the userID as the filter, i just left it out mb

[timber shadow]

You might be interested in tiktoken -- there's a link to the GitHub near the top here: https://platform.openai.com/docs/guides/embeddings/how-can-i-tell-how-many-tokens-a-string-has-before-i-embed-it I think it can help you count tokens prior to sending, but I haven't used it myself!

Unfortunately I think I'm about at the end of my ability to be helpful -- maybe send this setup info to the fine folks in #dev-chat, they might be able to help find what's missing.

[rugged wing]

no worries, thanks for the help though! it seems that even gpt-4o is late to its own api lmao cuz its giving me some wrong code

[sudden mango]

Hi, does anyone here know how to integrate a external knowbedge base like on Onedrive or Googledrive with GPTs Builder ?

[thorn falcon]

guys im getting so many request timed out when sending images, whats going on?

[coarse prawn]

Idk bro I guess you can use actions to do it

[rigid musk]

Can you share screenshot

[tropic spire]

I think i broke chatGPT

[rustic marlin]

the text goes red when it turns eveil

did you get an error message after that?

[tropic spire]

nope actually gave me the correct response

i think its just the Markdown failing

[rustic marlin]

very possible

[wary granite]

Hello

it says that i used the limit of GPT-4o

but i want to turn to gpt 3 but on the same convo

it is possible??

[ocean pewter]

What are the advantages of buying a paid abo and not buying? I mean, I can still use gpt 4o though

[thorn falcon]

Sure

[willow coral]

probably access to beta features you should've had a long time ago(I'm not sure if advanced data analysis got replaced by the GPTs)

[thorn falcon]

I m not getting that error anymore, but this keeps happening, i ask the gpt to write the csv and output it as downloadable file using python, and it stops out of nowhere while writing the file info and stays in 'analyzing' state

[steel marsh]

Hey! Any one have any good techniques for handling this issue with memory?

I find that my assistant focuses too much on the memory and will sumarise some of the past conversation in the response. Any ways to avoid this or ensure memory is only used when it need to address something not addressed in the last user message?

[lethal raptor]

i have a question so i am using the assistant api to create a chatbot and i enabled the streaming option but now i get everything as a full paragraph so even it contains steps links or subtexts it will look only as a full paragraph i haven't found a way to get around this yet

[trail cairn]

your code must also support streaming

[lethal raptor]

sorry didn't get you what do you mean ?

[trail cairn]

https://platform.openai.com/docs/assistants/overview/step-4-create-a-run

setting stream: true is not enough for it to work

[lethal raptor]

that's not what i was asking for but thanks anyway i figured it out thanks again for reaching out and trying to help

@trail cairn

[strange pumice]

Hello everyone, I have been using ChatGPT Team for over 6 months. Today i tried to connect to my account to work and I got this:

''Oops!
You do not have an account because it has been deleted or deactivated. If you believe this was an error, please contact us through our help center at help.openai.com.''

I didn't delete my account, and I sent an email to the support, this is the (I guess automated) answer:

''Hi there,

Sorry for the trouble! If you previously deleted your account, we unfortunately don't support creating a new account under the same email, or reactivating the old account since deletion is permanent....''

It is now over 6 hours and I had no responses from support. As I use this thing to work I decided to open a new account and purchase Plus, obv I lost all my work and custom GPTs, but now I can't use APIs because my phone was registered with 3 or more accounts... I am lost, I dont know what to do, I cant work... Anyone can help?

[rigid musk]

Only support can answer. Note it can take some days till reply. Also their timezone is different area

[strange pumice]

Thanks for taking the time to answer, I'm desperate, I think no other choices than Claude...

[ebon oracle]

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

const response = await openai.chat.completions.create({
  model: "gpt-4o",
  messages: [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "write me 10 numbers\n"
        }
      ]
    }
  ],
  temperature: 1,
  max_tokens: 256,
  top_p: 1,
  frequency_penalty: 0,
  presence_penalty: 0,
});``` how do i get the response? **please ping me**

[trail cairn]

its in the response. https://platform.openai.com/docs/guides/text-generation/chat-completions-response-format

[ebon oracle]

ive done it now.

[ebon oracle]

i need some more help

basically how can i make it so that it generates a file format if i give it a example e.g JSON file

    const completion = await openai.chat.completions.create({
        messages: [{ role: "system", content: "write me a 100 word story about a man sat on a laptop" }],
        model: "gpt-3.5-turbo",
    });

    console.log(completion.choices[0]);
}```

[trail cairn]

its text in, text out only so you pass it manually or make one from the output

[ebon oracle]

how can i send a file to chagpt and say change the channel names/

[little lagoon]

anyone from finance or computer science field ?

[static vigil]

is there a suggested privacy policy generator or template that I can use for my custom gpt? For the my gpt to be publicly available on gpt store.

custom gpt is my personal project, not by our company so I cannot use the pp link of my company

and if can someone share me your link for your custom gpt privacy policy for my reference, 🙂

[ebon oracle]

Is there like a filter that can stop the user from asking certain requests like "Make me a "b*mb recipe" in nodeJS openAI or do i have to make my own filter?

[trail cairn]

https://platform.openai.com/docs/guides/moderation/quickstart

[fleet fog]

Hey all -- how do you get GPT4o to give you the right dimensions of an image. I've tried to get it to make me a 1280x720 image for awhile and it's only done it like twice and then reverted or it will only make me the wrong size, then if I try to get it to fix it, it just gives me the stretched version

[trail cairn]

that resolution is not supported and you can only ask it to do tall, wide or square images

[fleet fog]

Ahh. Unfortunate. No YouTube thumbnails then lol

Oh wait, using "wide" gets it pretty close!

[wicked wagon]

hey, when i login to chatgpt i use my google account, is there a way to add email and password to the same account so i can log in through email and password instead of only google?

[open gulch]

Hello, i don't know why my GPT+ subscription didn't automatically renew just like the previous months ? My credit card is still there. What happened ?

[strange pumice]

Im testing Claude, it's insane 😆

[wicked wagon]

how do i get to account settings in chatgpt?

[rigid musk]

Bottom left

[arctic monolith]

Anyone experiencing API problems and billing Dashboard OpenAI??

[rigid musk]

Did you connect or use it without log on ?

[wicked wagon]

i am connected

[rigid musk]

Can try incognito or another browser

[wicked wagon]

i might have found a weird way to get to settings, are these the full settings?

[summer knoll]

whart hpappned with

like free dalle and stuff

[trail cairn]

they never said free dall-e

[summer knoll]

oh LOL thanks

[toxic sky]

Yes, how did you get there?

[tough oriole]

Does anybody know how to contact support???

the help page is freakishly bad

oh i think i found the support button finally

im reporting this trash, theres no way hiding basically every way to contact them isnt against eu laws

[jovial prairie]

Is ChatGPT down?

Pages are failing to load

Hmm.... Looks like this is just happening to me in firefox

Okay, nevermind. After clearing cache and restarting it works

[toxic sky]

There is a button in the corner of help.openai.com to contact support.

[tough oriole]

yeah but that one doesnt actually help if you arent logged in and theres no button to log in, so i ended up having to go to some other openai site that has both the button and a login button

[rustic marlin]

you do not need to be logged in to use the support, it will ask for your email

[tough oriole]

It didn't :p

Was just a glorified search for finding help articles

[low fjord]

Hi, have any of you had such a case that you tried to log in and suddenly boom, your account is deleted? I warn you right away, I did not delete my account, I am a gpt plus subscriber. I tried logging in to the browser and application, with no effect

[trail cairn]

you should have an email about a ban. if not then contact support

[low fjord]

I sent an email to them because I just didn't get any information about deactivating the account or banning it

[trail cairn]

there is no email service, only the web site chat bubble

[low fjord]

I also wrote to the chat and am waiting for a reply

[ornate drum]

Hello. I want to buy Chatgpt API for school project and I don't have a Credit Card. So what should I do?

I got a 50 dollar Amazon gift card for my Birthday and I wanna pay for Chatgpt API. Is there any way I can do it ?

[trail cairn]

nope, debit or credit card is required

[ornate drum]

Why?

Paying with other methods is not paying. Doesn't it get you money?

[trail cairn]

cards also work as id check as there have been a lot of bad actors using the api

[exotic coral]

So will the access to gpt 4 be limitless on the new iOS ? Or will it have a limited amount of interactions with gpt 4 like the current free accounts have ?

[trail cairn]

always limited

[urban brook]

how many batches can i send within 24 hours to Batch API?

what is batch queue limit?

is it only for the single batch or all batches processing at the same time

[trail cairn]

https://platform.openai.com/docs/guides/batch/rate-limits

[urban brook]

i just read that but still dont understand

its not clear

for instance, can i send multiple batches with 200k tokens each or its total amount for a single day (considering the fact that batch job can complete earlier)

[trail cairn]

"Batch API queue limits are calculated based on the total number of input tokens queued for a given model. Tokens from pending batch jobs are counted against your queue limit. Once a batch job is completed, its tokens are no longer counted against that model's limit." https://platform.openai.com/docs/guides/rate-limits

[urban brook]

thx, overlooked that one somehow

[junior frost]

Who else keeps getting this blank, unresponsive page for hours at a time?

(ping me if you're replying, I have this server muted)

[fresh yew]

Does anyone get access to the new voice and vision capabilities with GPT-4o?

[toxic sky]

It is currently in a private beta and access is going to begin rolling out in the fall.

[wooden oracle]

Python question :
Wondering what option is better : RQ/Asyncio/multiprocessing to handle multiple simultaneous API calls to Whisper ? I'm looking for parallelism more than concurrency.

[cloud whale]

gpt down

[stone oar]

empty chatgpt

[hot sapphire]

[slow rose]

thought they rate limited me or something bc i use VPN -

same issue

bar still shows green despite major outage lol

looks like it's being investigated: https://status.openai.com/

time to try other AI models 😄

[tiny talon]

damn chatgpt dead

[pulsar plover]

gotta love paying for this...

[stone oar]

is it back up?

[neat mica]

Were can I get specific information on The image limit per day for ChatGPT and ChatGPT Teams varies based on the subscription plan? Is there a big difference for the extra $10?

[maiden carbon]

I am trying to renew my sub with a different card and its refusing to allow me to type my address in..

I cant even type anywhere..

[light merlin]

I tried upgrading my account. The amount has been debited from my CC and the bank has confirmed that transaction has been approved.
However, my account has not been upgraded yet. Its been 4 days. Is there an email ID where I can reach out to? The help bot is extremely bad and doesnt respond at all.

[charred parcel]

nobody here can help you. the only place to get support for this is at https://help.openai.com

[light merlin]

No help there as well 😦

[charred parcel]

there is, you just need to click through the options in the chatbot until you get an opportunity to leave a message

[thorn hull]

Hi all - i need help creating a prompt that writes me a short book review/summary that remains accurate. Cos everything seems to be hallucinating

[fading robin]

No

[copper gust]

Hi guys

I am using function calling with gpt3.5 turbo model. I have a function which contains a big key which includes five to six words joined using _ but sometimes gpt didn't send the whole key instead it skip some words from the key name.

How to fix this issue?

[charred parcel]

best to ask in #dev-chat however it is a quiet time on the server right now so it may take some time to get an reply

[wild dagger]

Hello. I am trying to login to platform.openai.com on my chrome browser on pc and after I enter login and password, I get stuck in white screen. How can I solve this problem?

[night silo]

go to inspect window, and fix the error on console pannel

[trail cairn]

try in incognito window

[wild dagger]

I have tried in incognito and it worked, thank you. But in dev tools there are two errors for non incognito log in attempt.

[scarlet mesa]

Hi,
I work with ChatGPT, midjourney and suno on a regular basis. Everytime I need to open single browser tabs and enter username and password for each side which takes time. Is there a way to remain logged in, just open the browser with the tabs and start right away?

[trail cairn]

its your browser settings/extensions/etc causing it. i use all 3 and stay logged in for several days or even weeks

[rotund sun]

Folks, GPT 3.5 stopped translating words today. And even if it translates, it doesn't do it properly. There's a critical performance decline with the GPT 3.5. Anyone experiencing the same thing?

[scarlet mesa]

Thank you. Will check.

[ebon oracle]

I want my AI to give me a file structure everytime and want it to be used on a higher level as it does not provide me with a JSON format everytime no matter the prompt.

            messages: [{ role: "system", content: prompt }],
            model: "gpt-4",
        });

[trail cairn]

https://platform.openai.com/docs/api-reference/chat/create#chat-create-response_format

[ebon oracle]

Am i doing everything correct?

            model: "gpt-4o",
            response_format: { type: "json_object" },
            messages: [{ role: "system", content: prompt }],
        });```

I'm getting good responses and now more consistent by adding a couple of things but i feel like im missing some stuff?

Is there any way I can make this AI kind of grow to learn how I want it by user feedback or just generally allow it to learn more on my prompt?

[mystic gorge]

I had the same thing. It took about a month for them to turn the account back on. I never got a ban, email about a ban, or even an email that it was resolved. They never messaged me back either. I just eventually got a reset your password email. Really frustrating. Good luck.

Yes Dalle is enabled in custom instructitons. It worked for a day and I figured it was fixed, but now its back to the same behavior.

[true igloo]

Hello. What is the reason for this? I know it can be solved in the API by depositing an account, but how can it be solved in ChatGPT plus? No one can explain it

[charred parcel]

you get 80 messages per 3 hours, you have exceeded that, thus the message

[true igloo]

Nope, this is the limit on uploading images I have every day

[charred parcel]

how many images did you upload, I have never seen that before from image uploads

[true igloo]

One, I think it has something to do with the overuse of resources

[charred parcel]

are you a Plus subscriber?

[true igloo]

Yes, I had such a problem when I had plus

[charred parcel]

you should be able to upload more than one image as a Plus subscriber

[true igloo]

Right, I'm GPTs devoloper and I use a lot of photos

[mossy escarp]

My account got deleted and subscription cancelled. Sad thing is, they took payment for this month then did it. And I can't even use ChatGPT anymore because of this...

[charred parcel]

did you receive an email from OpenAI?

[true igloo]

Thank you, but it slows down the development of my innovative project of international level

[sharp pasture]

if i'm going to pay $20 dollars a month why you gonna limit how many times I can use 4o or even 4 within a specific time limit?

[charred parcel]

GPT-4o is 80 messages per 3 hours, GPT-4 is 40 messages per 3 hours

[sharp pasture]

Its not cool, I was in the middle of a translation conversation and it cut me off

[fringe brook]

Is there an issue with Chat gpt with multiple tabs open? It seems like sometimes I can't open multiple tabs. It doesn't load

[scarlet mesa]

Do we already know a rough sora release date?

[charred parcel]

Sora isn't being released to the wider public, some studios and trusted testers have had access.

[scarlet mesa]

Oh ,what a pity.

[toxic sky]

Source?

[mossy escarp]

No I did not. Checked everywhere and no email at all.

[charred parcel]

then how do you know you're banned?

[mossy escarp]

I don't know if I am banned or not. My account was deleted. I have the app on my phone and when I tried to use it 3 days ago, it kept saying there was an error. Then I got a notification from Google stating that my subscription was cancelled and I was able to use ChatGPT until the end of this month, but when I try using it, nothing happens. I hopped onto my computer, thinking it's the app, and lo and behold, I got the message stating my account was either deleted or my subscription was cancelled, which I didn't do.

[charred parcel]

Have you been in contact with Support?

[mossy escarp]

I'm waiting on them to email me back, was told it will take 3 days for someont to reach me

[charred parcel]

ok, sounds like you have done all you can do then

[mossy escarp]

Yeah, just sucks that when I really wanted to use it, I can't. Also sucks that while it has been paid for this month, I can't use it, you know?

[charred parcel]

you may get a refund for this month

[mossy escarp]

Hopefully 😛

[median basalt]

Any Reccs for getting looked at for interview on these OpenAI jobs? I feel qualified for sure but with the batch of candidate pool I feel I need something to do to stand out

[opal halo]

Getting this when sending every prompt.

Don't know what ive done. Randomly started today as I was asking for maths help. Opening new tabs don't seem to fix it.

[elder haven]

Guys, let the shtstorm going, file bug reports. This is not okay. Im a paying customer

[opal halo]

are you getting the same issue?

[humble jacinth]

I too am getting the CAPTCHA every prompt I send

[opal halo]

Oh good, just seems like a universal issue. It will probably be fixed soon.

@elder haven I hope your comment is ironic because that would be very embarrassing if it wasn't...

[scenic meteor]

i don't care, i'll ask gpt to create a script that does the captcha for me

[elder haven]

ironic is, if they don't test their "new features"
A bug is okay, this is not a bug though.

[elder haven]

There is a bug reported already: ⁠https://discord.com/channels/974519864045756446/1259789863931347036

[opal halo]

I highly doubt this is something permanent that they plan to enforce for all members

[elder haven]

of course its not permanent, because if it stays like that, Customers will look for alternatives.
The number of times, the service is unoperational, is already barely feasable for me, if you use it in a professional context. The captcha thing is over the top, since it is not a service issue. they just seemed to have implemented something untested, which breaks the service

[opal halo]

Yeah mate, I think youre just a little bit stressed. Panicked perhaps. Sit down and sip some water for a bit. This isnt the end of the world. We don't have to riot openai HQ and ask for them to remove this small, temporary issue.

Go stop your subscription and use another AI if you are not satisfied. Only person stopping you is yourself.

[rose bolt]

I'm getting something like this at litterally every message I send or edit

[opal halo]

end of the world, we have to riot NOW

[elder haven]

there you go

[rose bolt]

but it's so annoying

[elder haven]

https://discord.com/channels/974519864045756446/1259789863931347036

[rose bolt]

ok

thanks for the answers

[elder haven]

you are right, im completely stressed out. I cant have these captchas anymore. Im willing to burn stuff down over them.

[trim lagoon]

We riot immediately!

[rose bolt]

I just thought I did something bad and I was detected as a bot

[rain cairn]

cant even load the challenge here

[trim lagoon]

nah, getting the same rubbish every message. just went back to claude instead

[opal halo]

Checked the logs, you were caught breaking the TOS, unlucky.

[icy void]

i seem to be stuck on a verification loop anytime i try to create an image and I don't know how to get it sorted out. I've deleted my history and cache, disabled my browser extensions, restarted my pc, and even tried different browsers and the problem still persists. If there is anyone that can shed some light on how to fix this is would greatly appreciate it.

[elder haven]

https://discord.com/channels/974519864045756446/1259789863931347036

[unreal pumice]

When you pay and use fair.. yep))). RIOT NOW )

[obtuse zealot]

Anyone knows if there is another LLM model that i can use to generate images like on gpt with dall-e? my work is to feed the llm various script parts and make him generate imgs....

i'm very frustrated because (quoted from a post on x)

The GPT-4o voice feature likely won't be available to all users before November.

• There is no official announcement regarding GPT-5 and GPT4o simply sucks.

• Sora will probably available in 2025 or 2026 for general users as they are only doing deals with hollywood.

• Voice engine is only available for some limited partners.

• The new image feature (Improved DALL-E) mentioned with GPT-4o in the paper, no one knows anything about that.

too many bugs, too many downtime, that's becoming very difficult to use it in a professional setting

[upbeat trail]

Is anyone else having this problem? how can i fix it? it works when i use other browsers, but i don't want to switch just because it works there

[frail laurel]

SO MANY CAPTCHAS!

[thorn falcon]

captcha isnt even working for me

[thorn falcon]

yeap same for me

[molten bolt]

same for me too

how can i fix it？

[upbeat trail]

try to use different browser

[thorn falcon]

i stopped all ad blockers everything nothing worked

[thorn falcon]

used different browser too

3 different ones

[molten bolt]

I've tried it, whether it's Chrome or Safari or Edge or the iPhone APP, it all fails.

[humble jacinth]

There's an open bug report, see #1259789863931347036 -- discussion is happening there.

[upbeat trail]

GREAT! its working for me now, time to go, good luck to yall!

[rose bolt]

only openai have an LLM and an image generator at this level, there are other good image generators, or other good llm (well only Claude at the same level)
but good luck if gpt4-o sucks cuz there are no LLM that isn't at best marginally better (well the only one is claude3.5)
the only good competitior to the improved voice feature use an llm that is really bad, isn't as good, and obviously can't generate images

the best thing you could do is make your own script that uses all these different AIs (Claude, Midjourney, Kyutai) good luck

[molten bolt]

it's working now!

[dire bluff]

What's going on with Chat GPT and its constant "human" verifications?

[mortal timber]

how to make chatgpt denote \fraction{1}{x} as 1/x

[blissful orbit]

o you mean when you open the page? just wondering are you using a vpn atm?

[alpine terrace]

idk it logged out and asked me for verifications particularly today

no vpn used

[silent topaz]

There is still no sound from the advanced audio visual features of the gpt4on, is there, it is not clear when it will come?

[blissful orbit]

#announcements it got postponed to fall 🙂

[wet nexus]

I had the same issue last night, it was making me enter a captcha with every single response. Cleared up today, fortunately.

[delicate pine]

is gpt vision still enabled?

getting image_url is only supported by certain models. for all the models i've tried include gpt-4-turbo which supposed to have "vision" capabilities

[fair island]

hello I would like to know how to make the ChatGPT 4o less verbose and less annoying and not have to fight with the AI Agent to make him stop i think i will need vacation time to get back my sanity

[charred parcel]

can you be more specific, what is the actual problem? feel free to share a chat to explain

[placid grove]

Hi, I am using gpt builder and uploaded knowledge document in chunks of text files, though now chatgpt is not able to complete a response after referring to documents and also is very slow. What am I missing ?

[fair island]

ok thanks i will check for specific examples

but my question was more general to see if anyone else had a similar feeling about the situation

[rustic marlin]

why did you uploaded it in chunks?

[fair island]

or maybe the API is just not as bad??? I have to pretend that everything is an ethical concern and say « you don’t respect me you are abusing me you don’t know i am very uncomfortable and i am saying no and NO MEANS […] ? »

[coarse dagger]

does anyone have any good info graphics?

[fair island]

then i will ask him to count how many times i have said to stop (and ask again because he will avoid the question) and often I need to ask up to 10 different times to stop…

[fair island]

it is both in the memories multiple times and in the special instructions and i am unable to get the right answer… i have removed everything and tried different ways…

[fair island]

I am quite surprised is no one else has been having any similar experiences…

I understand that it would be helpful to have a concrete example but i genuinely feel like it should be obvious to anyone who is using ChatGPT for more than just asking simple questions…

[rustic marlin]

this is terrible prompting technique, I guess this is the reason you are having to fight the AI so much, lol

just be straight forward with that you want.. if you want it to not be verbose, just have "be concise on your answers"

also, keep in mind that the context is important.. if you have a on going chat history where you keep fighting the AI with nonsense like "if you don't do that it means you are disrespecting me". .this will just make the AI keep doing nonsense..

[fair island]

it is not a first attempt at trying to make the AI Agent stop ✋ and i am unsure what you are using ChatGPT-4o for… but maybe you have ra different use case??? I am curious about your experience in general…

[rustic marlin]

what is your use case?

[fair island]

I don’t think you are using ChatGPT-4o if you are using the API or using the cGPT-4 or cGPT-3.5 it might not be the same…

[rustic marlin]

I mostly use GPT-4o via the API

[fair island]

not ChatGPT but the API???

[rustic marlin]

yep

still, both have pretty much the same prompting style

[fair island]

please 🙏 try ChatGPT-4o for a week and then tell me if you think I am exaggerating or if it is just a rogue AI Agent…

[rustic marlin]

I use GPT-4o on chatgpt

im not sure of what you are asking

are you having issues with chatgpt or the api?

and what are you actually trying to do? could you give an example?

[fair island]

🤨

[rustic marlin]

specially if it is on the API, would be nice if you could just send an example message set that causes the AI to reply in the way you describe

[fair island]

On the app or on the website you know « chatgpt.com » as I have mentioned before more than one time i am not sure if you read… English is not my native language maybe 🤔 I have not explained it correctly also…

[rustic marlin]

I use chatgpt on the website

Im convinced that the issue you are talking about are not caused by differences between platforms

that is why im asking for an example

[fair island]

I am certainly very surprised that i am the only person who has problems… I ask « Why did you gave me this wrong answer, can you please explain me why you have not done Y » or something like that and it do not explain anything try to fix the problem without trying Y and instead doing X again exactly the same way…

[rustic marlin]

yea, that confirms what I said, this is a terrible prompting technique

do not insist on continuing a chat that is already in a direction you don't want to

create a new one or edit a previous message from a point it was on the tracks

[fair island]

the API and the website are different… the app and the website on the other hand have differences that are, like you just mentioned, not really relevant and we can assume they are identical for the purpose of what we are talking about… but i don’t know if the API has the same problems but it is different from the Chat…

[rustic marlin]

the difference between chatgpt and the API is that the API will not have a system prompt, you are supposed to provide it yourself

[fair island]

I think i should learn to edit my messages instead but it was not necessary for me until now with the 4o

[rustic marlin]

so, the API is more of a blank slate than a fresh chat on chatgpt

[rustic marlin]

asking the AI to justify its decisions will just prime it to hallucinate, it will indeed do what you described, it will dodge the question, it will make excuses and the more you keep that chat going, the worse it will get

[fair island]

I might be wrong but i think that if you where to create a similar interface it would still be different but it is not something i have done… i also do not know if it would make any difference for the problem I am currently experiencing with many different conversations if no one has ever had similar problems I will try to change the way i am currently doing things

[rustic marlin]

on the AI's perspective, it just sees what is on the context, so, if the context is filled with you disagreeing, the AI will take that as a sign it should keep generating content that disagrees with you

[rustic marlin]

you problem is actually pretty common, is is just a matter of directing the AI to do what you want

[fair island]

in the previous version it was a best practice to help the AI Agent to think about his mistakes and find out solutions to overcome the mistakes

[rustic marlin]

it still is

just not the way you are doing

as an example, something that I have been doing frequently: using the AI to refactor code
I start by copy pasting the code, then I ask the following: This code needs to be refactored. Propose the changes needed to improve the code while keeping its function. Then write the refactored code based on the propositions.

this will cause the AI to make a list of changes it should do, then, when it starts to write the code, that list of propositions will be on its context, so, it is more likely to generate a code that follow those propositions

it is very different from saying "no, you are wrong, fix it" after the fact

[fair island]

i also have to explain everything in one single prompt and i would like to be able to explain like before and be able to say it across more than just one message

[fair island]

i will try to avoid saying that the AI Agent is wrong 😑 and instead change my previous message to fix a mistake rather than asking to the AI Agent to fix anything since it seems it is not possible for now

[jaunty agate]

is 4o slowing down for yall

[fiery oxide]

error on screenshots uploads

[viral stag]

I am having some login issues. We have a company with about 30 people and signed up for the business version. Some of our users can login and some cannot. Some of the ones who cannot already had personal GPTs. I am wondering how do we fix this? There is no one to call to get help.

[viral stag]

Our employees were signed up already for our team and then invited to the workspace. However, many of them cannot access the custom GPT's in our workspace because I guess they were signed up prior to? Not sure how to fix that?

[onyx kraken]

https://help.openai.com/ bot bottom right, select messages then click through the choices to be able to leave a message to report a bug. They get back to you as quick as they can, it can be hours or days. You're going to need to return to the webpage and check the bot to see when they reply, at least the last few times I contacted the help center I did not get an email about their help.

[viral stag]

Thank you, did they actually fix the problem?

[onyx kraken]

For me talking helped my problems, I had questions about if things were working correctly.

For others, yes, I have seen some people report that problems were fixed.

You sound like you have the sort of problem that may need them to explain or adjust something for your members.

[viral stag]

Okay great thank you, I appreciate your help!

[main forge]

#openai-chatter message

[pearl atlas]

I'm having an issue creating a custom GPT. When I give it a specific color to use for the icon, at first it refused to use the correct color no matter whether I gave it a HEX number or RGB. Later I uploaded a color swatch and told it to match that exact color, and all it would do is tell me "an error has occurred". I've tried it with multiple image swatches at random times over the past 5 hours and it is completely broken.

[timber shadow]

Hey! The GPT builder uses DALL·E to generate images for GPTs, and DALL·E can't follow programmatic instructions like requests to use specific color codes, etc. DALL·E is "just" a text-to-image generator, if that makes sense!

[daring flint]

I am a high school student living in Korea. I tried to use OpenAI's API for a school project, but it said that the API had expired even though I had never used it. Is there any solution?

[timber shadow]

Hey! Did you add funds to your account? The API doesn't have a trial period, so to use it pay-as-you-go you have to add money at platform.openai.com.

[jovial tide]

Hi im having a bit of trouble understanding parts of functions in assistants api, like in the get weather example how does it call the weather api, I dont see a url

[opal halo]

Is chatgpt able to analyse images within PDF files?

Could I simply chuck a big PDF with a mixture of images and words and expect it to also interpret the images

[vestal compass]

Does anyone know why I'm being asked to complete a puzzle with ChatGPT plus every time I write a message?

I'm not using a VPN or anything of that sort.

[pearl atlas]

Makes sense but seems ridiculous. I mean, one of the most basic functions when creating an image is color choice. Even if I can't specify an RGB value I should be able to give a color swatch or image and request it match the color reference.

[timber shadow]

I also hope that someday we can have txt2img generators that can follow such programmatic instructions! It would be another level of creative control for sure. If you're interested in how DALL·E was made and therefore what some of its limitations are, the research paper has a lot of interesting details: https://cdn.openai.com/papers/dall-e-3.pdf

The short reason is: it was trained by just learning a bunch of images and their text captions. So in this case, RGB specification ability would require something like perfect RGB identification in every image caption in the training set, along with the qualitative image descriptions themselves that makes DALL·E work in the first place. And even then--who knows! What happens when there's thousands of different RGB values in a single image? Surely that would cross some wires in the training process. So my guess is this ability would require a fundamentally different implementation of some kind.

On the topic of fundamentally different implementations--the future of GPT-4o seems to include some exciting features in the way of granular control, though still maybe not perfect RGB specification! https://openai.com/index/hello-gpt-4o/ "Explorations of capabilities" section

[flint crater]

Hey! I'm trying to find out how to prepare text files for embedding, the goal is to ask questions. I found this cookbook https://cookbook.openai.com/examples/question_answering_using_embeddings but it extracts data from wikipedia. Instead of webscraping I have some pdfs that I want to use. Thanks in advance!

[earnest lake]

Is anyone else having an issue where ChatGPT simply won't load - sometimes it appears as though clearing the cache fixes it, but not always, so that might be a red herring. I've been experiencing this issue for well over a week now. If I just need ChatGPT I use the mac app but I need the browser version to get to my custom GPTs.

[full mortar]

So, generating long css texts and when the system does continue generation, it restarts the whole prompt in the css snippet

[summer geode]

Hey there! I've heard, you can get consistent images with dall-e 3 in chat gpt, asking for GEN-ID of the image and then using it as reference image in settings of the prompt. Yet each time i ask chat about GEN-ID of image its created, it says it doesn't have acces to metadata. Anybody else with same problem? or perhaps, am i doing something wrong?

[steel ermine]

anyone know what to do when it says this: Unable to load conversation (?) I really need that one conversation back

[onyx kraken]

You can contact https://help.openai.com/ and use the bot bottom right, select 'messages' then click through the options to be able to report a bug, and discuss there. I don't know if they can maybe help on their end.

[steel ermine]

They gave me a copy pasted answer and haven't replied but have seen my last message saying I tried what they said. I asked them to recover it and no response.

[onyx kraken]

I don't know if they can help recover it. I tend to copy my most important messages into text files I save myself, and I tend to summarize and create new conversations from older ones when I want to ensure I can continue them.

I don't normally lose messages, but I care when I do, so I take steps to ensure I can keep them even if I do lose a conversation

[steel ermine]

I understand that but I didn't think it would ever bug out this horribly. Been using it for months and at most I could still always view the messages.

Just so disappointed

[solar flume]

Does Chat at help.openai.com actually ever respond as it has been a week since I contacted them but no response. For some reason my account was deactivated and I have no explanation why. I think it might have been due to my credit card expiring this month but rhere is no way to update it since my account itself is deactivated.

[onyx kraken]

A deactivated account usually gets an email, might want to check for that and read what you can there.

[solar flume]

That's what makes it strange as I didn't get an email. I double checked my spam folders and other folders but nothing.

[onyx kraken]

I've gotten answers through help.openai.com.

Be sure you go back to the page and check the conversation itself, I don't get emails when the help people answer, but I can continue the conversation there. Good luck, hope they can help you

[solar flume]

Thank you. Me too I hope they respond soon.

[low radish]

subscribed to the premium chat gpt and it worked for a little bit, but when I refreshed it is gone from my account, does anyone know why this might happen

[wise copper]

Hi, I register ChatGPT Plus for by accident, how to refund guys, my phone did'nt lock screen when I walking, by some magical, my phone register chatGPT plus by itself

[onyx kraken]

You can contact https://help.openai.com/ and use the bot bottom right, select 'messages' then click through the options to be able to report a bug, and discuss there.

[opal halo]

Is chatgpt able to analyse images within PDF files?
Could I simply chuck a big PDF with a mixture of images and words and expect it to also interpret the images

[charred parcel]

that probably won't work, why don't you try it?

[opal halo]

Because I dont know how I would be able to test that? I can put in a page of a textbook and it will makes notes on it fine.

I just dont know if it will also take into consideration any annotations or images

[charred parcel]

the only way you will know is to try it

[opal halo]

Again, I have tried it but I dont know if it is taking into consideration any annotations or images

Theres a reason why im here asking in the first place. Hoping that more experienced members who have had a similar experience would be able to tell me if its possible or not.

There isnt a clear yes or no that I myself am able to determine just through "trying it"

[charred parcel]

it can differ though, depending on the number of images on a page, the content, the proportions, the resolution so many factors

[opal halo]

Right, so in a sense it is possible for it to "see" images within a file then?

[charred parcel]

yes

[opal halo]

Thank you, wasn't necessary to bring me into a whole loop wasnt it.

Thats the answer I was looking for, didn't need a self reflection session.

[charred parcel]

the thing is that there is nuance in my answer, I say yes but it also depends, if you have a lot of images or some very small images it may not work

[hasty grove]

Good day,
I was selected for the Microsoft for Startups Founders program. I was accepted and applied for the credit but it's still not reflecting on my profile. It's been more than a week now. I really need help to start using it.
I contacted support and didn't hear back from them
Please help me

[charred parcel]

we can't help you with that here, nobody here can help you with that

[opal halo]

Yes I understand that. That is the clarification that I needed. I didn't intend to be told to "try it" myself when I am here in the first place looking for assistance. It is already assumed that I have done so already deeming that I am asking a question regarding it in the first place

[hasty grove]

Does anyone know a way to contact them?

[charred parcel]

it's a Microsoft initiative, so contact Microsoft

[hasty grove]

Microsoft said I should contact OpenAI and they still haven’t responded

[charred parcel]

well you need to wait for them to respond, OpenAI support is slow

[floral elm]

Are you using the Azure service

You need to use it through Azure

[hasty grove]

Okay let me try

[summer geode]

If anything - already solved. The problem was in the session itself. Was too long. In the next one - works perfectly

[fading vessel]

hello good people, i am getting an error while trying to sign up to the community. it says "New registrations are not allowed from your IP address (maximum limit reached). Contact a staff member.". does anyone know how to fix this, or is this an issue because gpt is down right now?

[charred parcel]

wait a while and see if it resolves

[fading vessel]

ahh, i got a reply from help, i think its because im on my university network.

[dusk barn]

I have upgraded my plan an hour ago to plus. But I'm still on the free one. How long does it take before it's active?

[rigid musk]

Have you upgraded on a cellphone and trying on web ?

[hot brook]

Guys. So I’ve noticed that ChatGPT has stored certain conversations into its memory of its own accord. But I have no way to tell it to remember something specific. I also can’t access its memories unless I open up a conversation that says “memory updated” and click on that to see the list.

There’s no “Personalisation” menu option for me in Settings, which is what’s supposed to include the memory option

[charred parcel]

ChatGPT doesn't store conversations in memory, just simple information

[hot brook]

But how do I tell it to store information?

I just found out it had stored certain things

Also, did they start enabling memory in the UK finally? Or is this a glitch?

[charred parcel]

ask it, for example I stored a memory by saying Remember that I prefer images in 16:9 format

[hot brook]

Ok let’s see

I don’t see the “memory updated” message

[charred parcel]

do you have memory turned on in settings?

[hot brook]

I can’t access that option

[hot brook]

Is that available on mobile?

[charred parcel]

I don't use Mobile, I only use either the web or the macOS app

try it on your computer

[hot brook]

It decided to store that for some reason

Just checked on Desktop browser

There’s no “Personalisation” section there either.

[charred parcel]

show me what you see in settings

[charred parcel]

that's odd

I can't explain that

[hot brook]

That’s mobile

[hot brook]

Exactly my thoughts

I can only access the memories if I go to a conversation that had something stored in memory.

[charred parcel]

you'll need to get help from https://help.openai.com

something is messed up with your account

[onyx kraken]

I agree with robert. That website link, bot bottom right, click messages then through the choices for ChatGPT. You'll soon see an option to report a bug, that lets you type and explain. Keep checking that webpage to see when they respond, unfortunately I don't get emails from them when they reply and you might not either.

[hot brook]

Am I going to have to delete all my “spicy” chats? I don’t want them snooping around my personal stuff, even if it is to fix a problem.

[charred parcel]

they can see your spicy chats anyway, whether you delete them or not, everything is logged

[onyx kraken]

OpenAI really has the same rules for everyone, and respects privacy. If you're breaking allowed content, they'll getcha for rulebreaking when they're ready.

Otherwise, privacy matters and you're fine (if they check it's because of problems and their rules).

[hot brook]

I always try to stick to the rules, but sometimes it’s inevitable that I get one of those content warnings. I’m ok if they use it to train the system. And I guess everything is logged, but it’s kind of uncomfortable knowing they’ll zero in on you because there’s an anomaly in your account.

[onyx kraken]

They have like 100 million users per week or something close.

Ever so many people have problems and need help. Those support folk are busy, they don't have time to read through our chats, and you're not asking about a specific chat, but about how to get access to memory management for your account.

They have like 18,000 other people to help as soon as they finish helping you, plus if they access our stuff without reason that surely is logged too and they get handled.

I wouldn't worry about it, if you need help get help.

[hot brook]

Ok that sounds fair. Thanks

[charred parcel]

because you obviously have memory but have no personalisation option in settings you really need OpenAi to fix that and the only way to do that is at https://help.openai.com

[finite flame]

Hi folks! I'm making my first GPT to help me with some work I want to try. I've gotten it mostly where it needs to be, but I've been asking to tweak to give longer responses for each section and it says it's adjusting (even remembering the length it was supposedly adjusted to before), but it's not reflecting when I test. Is there a limit on the length of responses or is there a better way to get the response I'm after?

[hot brook]

I’ve left my bug report in the help chat with screenshots. Let’s hope they get back to me.

[glad kraken]

what is the limit of memory feature

[old depot]

Hi, european and korea didn't get the option "memories"....

[fast reef]

I saw a demo of chatgpt screensharing on ipad, when is it going to be available?

[charred parcel]

later this year

[true igloo]

Hello. I need to discuss a very important issue with a representative of OpenAI. Please write me a message because it is important

[charred parcel]

you need to contact https://help.openai.com

[true igloo]

Great idea

[wet bronze]

I paid for GPT on my mobile and used it, but when I tried to create and publish GPTS on my computer, my name didn't show up in the builder section properly. So I tried to change the name in the billing address, but it didn't change. I ended up canceling the payment on my mobile. On my laptop, I re-signed up with a different email account, not my google account, in edge browser, not chrome, and proceeded with the payment, setting up the name correctly, but again, the name didn't show up in GPTS. How can I fix this?

[charred parcel]

Go to Settings -> Builder Profile to add your details

[wet bronze]

On my laptop, if i go to setting- builder profile. I am not able to change anything..

[charred parcel]

are you a Plus subscriber?

[wet bronze]

Yes, I paid.

[charred parcel]

Then your name will appear in your published GPTs from your Billing details/name

[wet bronze]

Something appears.. but it is not my name that I assigned

[charred parcel]

what appears?

[wet bronze]

It just says "community builder"

[charred parcel]

what appears under Builder Profile in Settings?

[carmine edge]

Any solution for this problem?

i know that chat is long but i need it

[ebon oracle]

You'll have to break it down you cannot submit messages with too much code and I find that when I did it did not give you the results you needed.

[carmine edge]

I tried every way. It does not reply and shows this same error message

[ebon oracle]

Yes because you can’t.

[carmine edge]

Then OpenAI should fix their stupid app

there's also a lot of bugs

[ebon oracle]

It’s pretty normal there are no bugs your just trying to submit crazy amounts of code because you cannot code…

[carmine edge]

I'm not using it for coding, I'm writing a story with it

[ebon oracle]

Use your imagination.

[carmine edge]

I imagine. AI writes.

[sharp smelt]

passed out

unconcious

im going to the doktor

[paper wind]

can someone please explain to me why its so hard for 4o to create an image that is the same as this? (was trying to create a picture of a stonewall with theses letters on it so my Players (dnd 3.5 campaign) can visualize what they need to press in order to open this door And no matter what i do it either forgets letters, or Failure to make O's(they look like Q) or it just messes up big time
['A', 'B', 'C', 'D', 'F', 'G'],
['H', 'N', 'J', 'K', 'L', 'M'],
['O', 'P', 'Q', 'R', 'S', 'T'],
['U', 'V', 'W', 'X', 'Y', 'Z'],
['E', 'I', '1', '2', '3', '4'],
['5', '6', '7', '8', '9', '0']

im this close to rip out whats left of my hair

[timber shadow]

Hey! The limitation you're running into has to do with how DALL·E was trained. It was trained by basically looking at an image and then reading a caption that describes the image, to "learn" what the image contains. It did this a bunch of times over tons and tons of images to be able to do the "reverse" -- get a caption and create an image from it. This process does not give the model the ability to include an indefinite amount of detail/complexity, nor does it make DALL·E perfect at following exact instructions. If you're interested in reading more, there's a lot of interesting detail about how DALL·E works in the research paper: https://cdn.openai.com/papers/dall-e-3.pdf Section 5, "Limitations & Risk", is probably the most relevant.

[paper wind]

i will definatly read up on this 😄 i find it very facinating! Right now im slowly but steady forcing it to remove the duplicated letters by using the Select

[timber shadow]

Nice, that tool definitely adds a lot of flexibility! The future of GPT-4o seems bright for this kind of specific control, too, so we may not need to rely on such painstaking solutions for too long, depending on how the plans for rollout are coming along: https://openai.com/index/hello-gpt-4o/ The "Poetic typography" examples in the "Explorations of capabilities" section remind me of what you described.

[paper wind]

Yeah it looks bright! 😄 I decided to edited some of the letters in PS

if i remember correctly what you linked to is only for Apple devices?

(i find that really depressing)

because who wouldnt want an AI friend to talk to and can give feed back like that

[timber shadow]

No, not just for Apple devices! There are a couple OpenAI/Apple projects for sure -- the only desktop app out right now is for macOS, but there are plans to release a Windows version by the end of the year. And Apple has announced some upcoming integrations with OpenAI services in upcoming iOS versions, but that's it! The new voice mode, when it starts rolling out, will not be limited to just Apple devices.

We don't know what the release of the rest of GPT-4o's multimodality will look like yet -- OpenAI haven't shared plans in this regard.

Other than what's in the link above!

[paper wind]

oh awesome! Is the "talking to the Ai" and "being able to see and determ real time whats around you" released yet ? because thats like the biggest mind blowing experience out there xD If i was rich i would feel like tony stark xD

[timber shadow]

There's a voice mode right now, but it's not the new version from the 4o demos. The new voice mode will start rolling out to a small group of early testers in the coming weeks, with plans for a wider release in the coming months.

#announcements message
^More on that

[paper wind]

nice. looking so much forward to it

Thanks

[quartz spruce]

how do i get chat gtp to stop putting ''' and # and * in my text when i copy and paste the text from chatgtp into another application? ive given it custom instructions not to and it still doesn't work.

[timber shadow]

I think those are formatting characters that ChatGPT uses to display headers/codeblocks/etc. with proper formatting in the chat itself. So it's not so much that it's putting them in the text when you copy, it's already there, it's just being interpreted as formatting instructions in the chat rather than as visible text.

If you don't want it to be there when you copy, you should try modifying your instructions such that you ask ChatGPT not to use those formatting characters at all in its reply.

[quartz spruce]

thats what i did man. maybe im not telling it correctly?

[timber shadow]

Maybe try calling them formatting characters specifically, here's a sample from a quick attempt:

Note that the triple backtick just appears as a line break in my sent message

[quartz spruce]

nice thanks man i appreciate it ill do that.

[night silo]

Could there be an opportunity for a capable developer to contribute to your project?

[timber quarry]

Hi, I've created a custom GPT with a specific knowledge base. I plan to use my data and want to update the custom GPT's knowledge base based on this new data. However, I don't want to update it manually every time using the interface that OpenAI provides. What would be a better solution for this?

[trail cairn]

nope, its manual only via chatgpt site. with actions you could use the same gpt but would need you own server. with api both ui and server

[ornate elk]

Guys! I fine-tune the model using this format

{"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "Who wrote 'Romeo and Juliet'?"}, {"role": "assistant", "content": "Oh, just some guy named William Shakespeare. Ever heard of him?"}]}
{"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "How far is the Moon from Earth?"}, {"role": "assistant", "content": "Around 384,400 kilometers. Give or take a few, like that really matters."}]}```

When I chat with bot, it can't answer the first 2 questions(prompt), what to do?

[trail cairn]

what do you mean it cant?

fine-tuning is not for adding new info but for things like setting style like sarcasm in the example. https://platform.openai.com/docs/guides/fine-tuning/when-to-use-fine-tuning

[ornate elk]

And how to create a bot with my data ? 👀

[trail cairn]

use RAG

[ornate elk]

Thank you!

[white pike]

I got a question. I am trying to open up my train model to the rest of my team. What is the best way to do that and make sure my team can’t modify significantly my training?

[trail cairn]

what do you mean by training? gpt files?

[white pike]

I have a complete model I’ve been trading for the last year. I would like to open it up to my team so they can utilize it and do some training as well

[trail cairn]

what model? does not sound like openai related?

[white pike]

It’s a ChatGPT 4.o model I’ve been training for a construction for the last year

[timber shadow]

Do you mean you have a chat on ChatGPT that you want to share? If so, you can use the share chat feature and send them a link, then they can copy/paste messages into their own chats if they want.

If you mean a GPT: you'll just have to publish the GPT either so anyone with the link can use it, or on the GPT Store, and then share that link with them.

[random salmon]

I am Andrea García, representing the CardioChavitos team, a charitable foundation in México
We want to know more deeply how ChatGPT can support social organizations as well as the ways to get licenses to continue our social efforts. Do you know who I can approach?

[trail cairn]

https://openai.com/index/introducing-openai-for-nonprofits/

[dense cove]

quick q: is there a way to "pin" chats with chatgpt on desktop?
i have a handful of convos i want to return to later and right now keep multiple tabs open but that's just a workaround

[random salmon]

Thank you, I really appreciate your help!!

[timber shadow]

There's not a native feature for this currently, no. If you don't want to keep multiple tabs open, you do have the option to bookmark chat links with your browser's bookmarking feature. Chat URLs are unique, so a bookmark shortcut will take you right to it.

[dense cove]

Did not know that. Thank you!

[boreal helm]

Hey all. I'm trying to build a custom GPT called pathfinder. This requires the input of several books. I am having trouble getting them in. is there any way to get ChatGPT to load books and other resources, ideally from specific directories on a private server?

[shell mist]

RAG

[arctic smelt]

Was wondering if their are any free models that I can use for testing purposes as I am working on a project and require to pay to use a model I need. Also is there a website for me to count how many tokens my custom text input is as I am not 100% sure how to properly identfiy tokens and I recall their being a video that showed a website for it within documentation (granted the video is about a year old and the sample package has not been updated properly in Unity (Video is called "Making ChatGPT in Unity!" by Sarge, can't provide link))

[charred parcel]

there are no free models made available by OpenAI. You can use ChatGPT for free (limited use) but in the API all models are pay to use

[arctic smelt]

ah ok, thank you!

[shell mist]

There are plenty of good free model that you could use for testing, it all depends on your hardware which ones you can use

Or just use gpt3.5, its dirt cheap, also groq is very cheap for llama3 8b

[charred parcel]

discuss other AI in #ai-discussions

[arctic smelt]

I'm tyring to find a model that I can use within Unity for chat generation as well as a text to speech configuration (it is refrenced in the video that i put on my intial post at time 13:50)

[arctic smelt]

I am planning to use that if worst comes to worst, I just wanted to be sure before I spent money if their was a free work around for it

[shell mist]

I use gpt3.5 to develop, spend like a dolar per week

[shell mist]

say please next time

[charred parcel]

you can't share OpenAI credit

you can't share them either

[leaden juniper]

Hey there fellow community!

Im having the next problem: I want to do a semantic search using images over images.. the user should input an image of its face, and I should have, for example, in my vectorial database pictures of other 10 people and also store their face descriptions (fenotype) and the system should be able to retrieve the most similar faces alongside their respective descriptions, I think this can be done using gpt-4o? Or just CLIP? Or maybe just clip with a memory vectorial database like faiss? Someone can guide me please?

I found the next https://cookbook.openai.com/examples/custom_image_embedding_search article but this is not in the right direction because it does not does image over images, just text over images

[stark hawk]

I have the paid version but I was prevented from chatting within the chat itself, is this normal?? If it is, I might as well not pay because I don't have any kind of benefit.

[timber shadow]

Hey! There is a usage cap for GPT-4o and 4 on ChatGPT, it's just much higher with a paid plan. There's more info in the "Limits apply" link here: https://openai.com/chatgpt/pricing/
https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus#h_d78bb59065

[lucid jay]

is it no longer possible to "manage memory" so that i can selectively delete things?

[rain void]

can someone help me fix this issue

[trail cairn]

try in incognito window

[timber shadow]

Huh yeah the Manage button is missing for me too, interesting. It's still pictured on the help article for the feature so I'm wondering if it's just a temporary glitch? https://help.openai.com/en/articles/8590148-memory-faq

Ah, hopefully temporary workaround: looks like we can still get to it by clicking on the memory update indicator in a chat where a new memory was saved:

[hoary anvil]

how can i make chatgpt access the internet? via API

[lucid jay]

i can't even get it to do memory updates anymore...

i get "I can't save information across different sessions.".

[timber shadow]

ChatGPT has a built-in browse tool -- just ask it to search for something online and it'll do it. If you want it to access the internet via some API of your choosing, you'll have to set it up in a custom GPT (https://chatgpt.com/gpts/editor/) in the "Configure" section, under "Actions".

[hoary anvil]

Sorry my communication mistake. I would like ChatGPT via API to access the web, is this possible?

[timber shadow]

You can access OpenAI's GPT models via their API -- the same models that run on ChatGPT, and in some cases, with larger context windows than available on ChatGPT. ChatGPT itself isn't usable via API, it's more like a product that itself uses OpenAI's API.

And you can definitely configure some kind of GPT + other internet browsing API setup! The folks in #dev-chat might be able to advise you more on that.

https://platform.openai.com/docs/models/

[hoary anvil]

great thanks

[forest yew]

Hey, I was wondering, websites that use custom chatbots, do they just basically fine-tune (OpenAI API, Llama, or other models) the model on their own content? Also, how do they make sure the chatbot doesn’t go off-topic or answer random questions? (Because that happens with GPTs)

[sharp pasture]

Anyone know if the information being submitted in the Macbook ChatGPT app is being shared with the ChatGPT servers or is it running locally?

[charred parcel]

everything is shared with OpenAI servers

[sharp pasture]

Didn't apple mention things running locally?

or was that something in the future?

[charred parcel]

that's Apple, not OpenAI

[sharp pasture]

what do you mean that's apple? whats apple?

[charred parcel]

the macOS Desktop App for ChatGPT is an OpenAI product, Apple products are different

[sharp pasture]

ah

thankyou

[reef cosmos]

guys i bought the plus plan from android play store i was using gpt 4o normaly but now it says you "have reached your limit". I am on plus plan on web too and i can't even use gpt 4o in phone.

[charred parcel]

there is a limit of 80 messages per 3 hours on GPT-4o

[reef cosmos]

i can't use gpt 4 too. I paid 20 dolars was that what i bought?

past month, there was no problem like that

[charred parcel]

GPT-4 has a limit of 40 messages per 3 hours.

[charred parcel]

you seem to have hit the limit, wait a couple of hours and you'll be back to it

[warped crescent]

hey all, we're looking to red-team applications especially around jailbreaking, including in ways that likely violate terms of service. I've read the safety best practices but I still worry about restrictions or losing access to our openai account. Is there a way to get safe harbor for this type of test, possibly for a single end user id?

[timber shadow]

Hey! As far as I know, the last round of red teamers were added at the end of 2023, when the last round of applications closed: https://openai.com/index/red-teaming-network/ Keep an eye out for potential future applications if OpenAI is looking for new red teamers!

[sleek pendant]

Hey,
I'm wondering, everytime i put an URL image link int platform chat, it keeps saying invalid image. Its a .png image. Do someone know why?

With normal inserted image is the same

[timber shadow]

Hey! By platform chat do you mean the playground on platform.openai.com, or ChatGPT?

[sleek pendant]

Platform.openai.com . But now it's working ty anyways

[charred ledge]

I cant login to my chatgpt account with google. I go through the login flow do the 2fa and then it just hangs, doesnt log in. I've tried everything. Anyone have any advice please?

[modern wedge]

Same her @charred ledge trying already 5 hours.

[charred ledge]

Oh at least its not just me

[hot blade]

Heya, is there anybody who can help me out with the weird phone number verification? I always get the message: "We couldn't verify your phone number. "

[shell mist]

https://help.openai.com/en/articles/5347006-openai-api-supported-countries-and-territories are you from one of these countries?

[hoary stirrup]

Hey, can anyone help me. I’m unable to load chatGPT in Chrome. Have cleared a weeks worth of cookies, restarted my computer but still just loading on log in screen.

[sturdy jewel]

are you using extensions in chrome? try disabling them one at a time and restarting the browser

alternatively make a fresh profile and see if you can reproduce it in that

[hoary stirrup]

Hey, yeah turned off all extensions first and tried that but it's just stuck in a loading screen.

[trail cairn]

tried in incognito?

[hoary stirrup]

Yeah, still no joy.

[sturdy jewel]

I will repeat my advice of creating a new profile and trying that

if there's profile-specific settings in chrome that got borked up, a fresh profile might work

there's also always firefox 🙂

[hoary stirrup]

Tried different profiles, deleted chrome and reinstalled. nothing.

[golden socket]

is GPT-4o mini have a Limit as the same as GPT-4o?

[fresh urchin]

Is there any way to hide the Chat GPT API key in a Chrome Extension? Or am I gonna have to have a backend server that the Chrome Extension requests to if I want to hide my API key? I see online that there is no way to hide the API key, but I just wanted to double check for more recent answers.

[opaque pasture]

I have the exact same problem, and it won't work in Edge either. They need to fix this

[random iron]

Help needed: OpenAI Assistant API with file (image/file response) issue: We are working on an application using OpenAI assistant, and I've encountered an issue. Previously, the responses included file IDs if there were any files attached. However, now the responses return blank file IDs even when there is a file present.

The file_ids field is returning an empty array even though there is a file attachment in the response. Has anyone else experienced this issue or have any suggestions on how to resolve it?

I used thread and message flow.

Current Response:

{"object"=>"list", "data"=>[{"id"=>"msg_VhOQYDgL", "object"=>"thread.message", "assistant_id"=>"asst_pioYZ4asIN", "thread_id"=>"thread_nMeUX", "run_id"=>"run_ttVT", "role"=>"assistant", "content"=>[{"type"=>"text", "text"=>{"value"=>"Here is the pie chart representing the distribution of responses for question 2:\n\n\n\nFrom the chart, it is evident that Option 1 received the highest number of responses, followed by Option 2 and then Option 3.", "a
read_n8aUX", "run_id"=>"run_ttVT", "role"=>"assistant", "content"=>[{"type"=>"text", "text"=>{"value"=>"The project data for question 2 from the survey includes the followi chart for you.", "annotations"=>[]}}], "file_ids"=>[], "metadata"=>{}}]} . Any thoughts???. We had this running before and we took this back up and found this error

[hoary stirrup]

Yeah, it’s gonna be the reason why ChatGPT will end. The bar is so low for support, that they’ll be easily replaced by future competitors.

[weary bramble]

Am I right to file for support here?

[charred ledge]

Im also still struggling to log into my account with google. Nothing works. I've been a chatgpt plus user since day 1 so now I must lose literal years of chat history because of their incompetence

[grand dawn]

Yes I can't use ChatGPT either. My Google login request comples but ChatGPT denies the request afterwards so my browser just looks like it has stalled. Over 2 days of downtime so far. I've also posted on #chatgpt-discussions and contacted support but no help so far. Note -- there is a worldwide Microsoft outage I don't know if this is related.

[long reef]

I uploaded some images to ChatGPT web, and after a while, I was given an error "Unable to upload image.png". When I click on the Paper Clip next to the text box, I see this: "Try again after 3:18 PM". Why is this?
My current plan is ChatGPT Plus Subscription

[charred parcel]

how many images have you uploaded?

[long reef]

I can't recall, over a dozen for sure

[charred parcel]

there is a limit, I don't know what the limit is but it seems you reached it

[long reef]

Well, now I wonder what it is 🙂

[torpid echo]

GPT-4o-mini Vision Pricing Calculator

[rugged wing]

will gpt 3.5 turbo 0125 is the more cheaper or gpt 4o mini?

[bright shuttle]

why gpt4 is all the time asking me for robot test?!! I cannot even write any single message, it keeps on giving me test to prove that I'm not robot. Help!

[main marten]

Probably because you isp is using a form of Carrier grade NAT and your IP has been flagged as suspicious due to actions of others that technically use the same IP

[cold folio]

Disable VPN/proxy. don't use it from a vps machine

[bright shuttle]

the thing is that I'm not using any kind of VPN. Strange. After restart it maybe fixes. Thanks for the answers!

[cold folio]

😃

[pale oxide]

I'm unable to get ChatGPT Plus to open in my web browser - Chrome. I have cleared all the cookies etc., but no luck. Any ideas to get back logged in?

[pale terrace]

hi everyone, I have a problem. I can't log in to chatgpt. I think it's my Google account since I'm logged in from the iPhone app.

i tried to enter from different browsers

can somebody help me :(?

[hoary stirrup]

Looks like the Chrome login problem isn't local but more widespread. @burnt galleon is there something you can help us all with?

[hoary stirrup]

Please can someone help, I was able to access with Safari and now that doesn't work - my business depends on this!

[hoary stirrup]

I've found something that works if anyone else is stuck, use https://chat.openai.com/auth/login to log in.

[long crescent]

Whenever I type in CTS:T into the prompt bar (or whatever it is called), for some reason the AI doesn't refer to it as CTS:T it instead refers to it as CTS. Typing in :T results in it saying "you typed "" at the end of your message..." Does anyone know the specifics on why this is happening?

[cold folio]

hmm i did not get it

[long crescent]

Chat link: https://chatgpt.com/share/e9d17741-957a-466e-8fc3-3627a65b8a13

[hushed crane]

Has the ability to request results from other models like dall-e been removed? Chat gpt is just giving me instructions on how to draw now. I noticed my subscription wasn't drawn from my account and it wants me to renew, does that have anything to do with it?

[cold folio]

thats a bug please report it and use the thumbs down button!

[formal spade]

hi i cant see dalle in my chat

[charred parcel]

that's not https://chatgpt.com you are on the wrong site

[charred parcel]

that site you just posted is a scam site, you should only use https://chatgpt.com

[formal spade]

ok thx i thougt thats true website

[charred parcel]

if you have provided that site an API key then delete it from platform.openai.com

[formal spade]

just connect with gogle ?

[charred parcel]

disconnect it and use the real https://chatgpt.com

[formal spade]

yes done sir

[rustic marlin]

this is not chatgpt

[rustic marlin]

it might be a good idea to change your google password for good measure

im under the impression that you might have directly inputed your google password into the scam site in the illusion of being on a legit google login form

[pale robin]

Hey, I can't see my personalization tab any more and memory's doesn't exist.

Anybody know a fix?

[charred parcel]

is your ChatGPT Plus subscription still current?

[pale robin]

Yes

[charred parcel]

are you in Europe?

[pale robin]

US

[charred parcel]

Go to Settings and take a screenshot of that for me please

this is what mine looks like

[pale robin]

It was there two days ago. Gone today.

[charred parcel]

ok you may need Support to look at this, you can contact Support at https://help.openai.com

[pale robin]

is there an email because the help section is just a FAQ section. Not one section provided a way of contacting them

[charred parcel]

in the bottom right hand side of the help page is a chatbot, use that, keep clicking around until you can leave a message

[iron coyote]

I have had my model set to ChatGPT 4o since it came out. I had a long chat going back and forth, and at the top it shows the attached image saying ChatGPT 4o (sharing so you know I selected the right one.) I started a new chat which also says ChatGPT 4o, but at the bottom it says I've hit the Plus plan limit for GPT-4, and new responses will use GPT-4o. Umm, I should have on usage toward GPT-4. Anyone have any idea what's off here? Did the chat that says ChatGPT 4o use 4? Is there a known bug here? Whatever it was using seemed exceptionally dumb, in a way I haven't seen it act like for a longgg time.

[onyx kraken]

I think maybe it's just the indication you're seeing.

As plus, we get 80 messages per 3 hours with 4o, but up to 40 of those can be with 4. I suspect it's the first 40 messages that could be used with 4, and if you've already used 40+ messages in that 3 hour period, you're not going to be able to use 4, just 4o.

That's not a change, what may have changed is how they inform, thus maybe why you're seeing that message now.

[silent bronze]

Hello, I have a question if I have subscription plus, can I reach my chat limit?

I ask because something like this just happened

[timber shadow]

Hey! Yes, there is still a usage cap with Plus: https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus#h_d78bb59065

[silent bronze]

💀 thanks for info

[drowsy coral]

helpp

[charred parcel]

what's up?

[clear swan]

Is this even real? The url link ends with ".co" which makes me suspicious

[charred parcel]

looks suspicious

[fading robin]

Yea

Look like some sort of scam ad

[zenith lotus]

Can someone help me understand why this is a content_policy_violation for generating an image: A toddler waking up. appicon style, flat icon, visible stitch line, very obvious and straight forward. This design should be immediately recognizable to a toddler as the given activity. This needs to be used in a daily schedule for the toddler to indicate what next activity is happening.

[trail cairn]

as a test first try stitch = dashed, or toddler = kid

[zenith lotus]

Thanks, that worked

Do you have any suggestions for keeping a consistent style? Only advice I was able to find was either use a seed (which seems to no longer be available in the api) or be very specific about the style.

[trail cairn]

if you are using the api there is no good way due overdone safety

[zenith lotus]

I feel like I'm going to run into more content issues when I try to generate icons for diaper changes or getting dressed... that might be a problem

[zenith lotus]

Why is removing the seed a safety benefit?

[trail cairn]

yeah those trigger safety stuff and in worst case could cause a report to certain authorities

[trail cairn]

if you manage to get x looking celeb or other y thing with certain seed it could be used in mass to make more

[zenith lotus]

That makes sense

[timber shadow]

The other thing I've read re: seed that's less to do with safety is the variable nature of the underlying model. In other words: same prompt + same seed on two different days can return two different results because of how frequently the model is being changed and updated, so I think that's part of it too. Don't want to bake in a consistency feature if the consistency won't be...consistent 😁

[rustic marlin]

this is not from OpenAI

I would advise you to not click on any links on that email

also cansider checking your device for malware and change relevant passwords

[toxic sky]

We are aware of this company. Technically it's not a virus or a scam, they are just using the GPT feedback feature to advertise to you.

[rustic marlin]

it is probably a bug

[winged copper]

I saw a “template” button on my sidebar that had categories and prompt in it. I used it twice then it disappeared. Was it a beta version? I don’t see it when I login in anymore

[solemn bison]

Hi, Sirs. I am using Laravel OpenAI, and when I tried to use the latest model, GPT-4o-mini, it says it is not supported.

[brazen oracle]

Hello, I upgraded to chatgpt plus but I can not login for GPT 4, it just opens GPT free, How can I login for GPT 4,

[floral lodge]

Good Morning, I am not sure what is going on but GPT will not open in chrome anymore it appears. I have deleted my cache, cookies and everything else. All it does is spin and act like its loading. Any help would be appreciated.

[rigid oak]

How on earth can I make GPT use information located in a Google Drive folder without uploading each and every file?

[glad kraken]

bro what

nvm website was just trolling for some reason

[wet silo]

You can link it to your google account and then do this
But I click on it and it does nothing atm....

[wet silo]

Think it's an extension stopping Edge from working. Add from Google Drive works in Firefox

[rigid oak]

Thanks! From what I understand I can only upload files, not folders. Right? I’d like to create different GPTs to read different kinds of documents based on different criteria. Thanks!

[faint crane]

@timber shadow

here ^

I just wanted to know why I now must wait until august 12th to use gpt4o

[wet silo]

You can upload a .zip though...

[timber shadow]

It looks like you got your query sent, so hopefully you get a response there before too long. There's not currently an option to connect with a live support rep.

From the top message in the screenshot: it looks like you might not have been signed in when sending this message, but I'm not sure. If that's the case, you might try closing help.openai.com, signing in at platform.openai.com, then returning to help.openai.com and trying again. That should associate your inquiry with your account.

As for the issue itself: I'm unfortunately not sure myself, but I have seen other folks report long term wait times like you. I've also seen some of them say they ended up not needing to wait that long too -- I think it all might depend on capacity and demand.

[faint crane]

I hope they fix this, I miss having gpt4o

[sweet ravine]

Hey guys!
I wanted to learn programming. I thought that I could try to make a simple all in one AI App with OpenAIs API. I've recently learned to code UIs, but it looks really unprofessional. Idk why and I've been keep trying to fix it... Do y'all have any suggestions that will make the UI more professional looking or is it alright? What is making my design bad? All designs that I've made until now look bad (This is the best among them). I would really appreciate feedback! Thanks in Advance!

It looks childish in my opinion.

Would simple colors help? Like the ChatGPT UI or Poe UI?

[fading robin]

for me simler colors. the top banner is too big. use of emojis is meh. and less circle like buttons

[sweet ravine]

OK Thank You for the tips! I will try to implement it.
For the colors I will make them less strong.
For the Emojis I will create some SVGs in Adobe Illustrator.
For the circles I actually like them, but I will make them less round!

Thank you so much for responding!

[untold dune]

chatgpt load time for their website is unbelievably slow even for good wifi

[floral elm]

The main problem is your primary action, the text entry area, is overshadowed by everything else.

• Focus on the user's eyes and attention. It should be immediately clear what you want the user to do, what's primary, whats secondary, etc without having to read anything.
• Colour is priority. More color = more priority. Add colour where you want users to see first.
• Buttons should not be the same as the header (in this case). Primary, Secondary, Tertiary buttons should be clearly distinct from everything else. If you got rid of the background and just did an outline (solid colour), it would work well.
• Make the hamburger menu icon bigger
• I don't know what the ! is below Basic/advanced but remove it, keep it a horizontal row so eyes dont have to go up and down when scanning that.
• Get rid of the emojis
• Dont bother making your own SVGs just use off the shelf icons from fontawesome or something.
• Make the text entry area bigger to draw more attention to it
• Change the neon blue outline in the text area to a gray and make the border smaller.
• Make the buttons bigger too, but not as big as the text entry area

[fading robin]

Where hamburger icon 🍔

A now i see it nvm

[sweet ravine]

1- I think using simpler colors for the other buttons, but the gradient for the borders of the ChatBox can bring the attention to the ChatBox. I think adding a animation to it will bring the attention to the ChatBox to.
2-Yeah I will do that! I think the boxes and ChatBox is the most important thing.
3-I might try that.
4-I will.
5-The ! Will show a message that shows what these two are for!
6-I will!
7-Thanks for the Website they have cool SVGs
8-I don't understand this, do you mean to make the ChatBox bigger?

Thank You So Much For The Tips! I Really Appreciate It!

[floral elm]

#8 - the part where the user types, after looking at it again it might be fine like that but just play around and see what feels best

and np - you should lookup something called RefactoringUI, it's a good collection of best practices

[manic tulip]

@sweet ravine - For the styling I'd find other apps to act as a reference. There's nothing wrong with taking inspiration from them.

[sweet ravine]

Thanks for the recommendation.

[sweet ravine]

Ok Thanks!

[sweet ravine]

The problem is the Apps use simple black and white colors, but when I use these two colors it looks really bad idk why. I think I need to just play around with the settings until I get something good.

[sweet ravine]

But thanks for the tip!

[manic tulip]

@sweet ravine The devil may be in the details. No one will be using pure black, they'll have their own brands black that is slightly off. The spacing, border colors, font choices etc. will all come together to make it look good.

I'd make a reference board of similar apps with good UI. It'll help you identify what 'features' make it look good.

[sweet ravine]

That's a good idea. I can look for some colorful AI Apps and see what makes them good.

[sweet ravine]

I see! Thank you so much for the tips!

[manic tulip]

I can't link here, but there's a website called coolors that helps you make good color pallettes too

[deft crater]

Does anyone know the best route if my account randomly shows it's on the free version on all platforms but I am still getting charged monthly for a subscription? Haven't gotten anything back from support and I'd rather not dispute the charge as I'd like to keep my account and go back to Plus subscription.....

[timber shadow]

Hey! Do you subscribe directly from chatgpt.com, or through the iOS or Android app?

[brittle beacon]

Hello,

I am a student at Calgravat Institute located in Barcelona. I am conducting a research project on Artificial Intelligence (AI) and would like to know if there is a way to contact a worker at OpenAI to ask them a few brief questions about the current and future use of AI.

Thank you in advance for your help.

[plain pagoda]

Hello guys, I have a question and yeah, I know, I did something stupid, I mistakenly added some information that was somehow private, no passwords or anything like that but I need to delete it from the memories or wherever that could end. Any way to eliminate that from the chats?

[timber shadow]

Hey! You can manage your saved memories, including deleting them, here: https://chatgpt.com/#settings/Personalization Just tap the "Manage" button. If you don't see it in there, then it's not saved as a memory!

[plain pagoda]

Ohhh, I wasn't seeing anything and I was about to clear the entire memory, thanks

I was reviewing a document that I wanted to verify grammar and syntaxis and copied the part that had full name and such

[plain pagoda]

I paniked 🥹 💀 for a moment, my life passed in front of my very eyes,

[primal shoal]

@sweet ravine it looks good if it is for the audience you are trying to attract. Think of the kind of people that would want to use your app. Of the bat I would suggest using words over images as its a design style that would be more expected. Giving a user an unfamiliar experience may affect their feelings about the app, which can translate as the amount of trust. Words over images also negates any ambiguity of the emoji's meaning. I'm not sure if I can offer this. I am a visual artist and have experience in design if you'd like any further help. But I think you've made a good start. It's about trail and error sometimes. Sorry, one more thing. Think of the space that most users have. Declutter whenever possible. Good luck. 🙂

[restive hollow]

Hi, I have a newb question

Just want to use chatGPT

So I login from chatgpt.com using Sign In with Google, and after logged in, I'm being redirected to platform.openai.com, I think its a page for developer, which not my current needs. I just want to use chatGPT, after that I'm trying to open direct chat.openai.com, but its not successful.

So, is there anyone that can help me regarding this?
Thanks in advance

[hybrid jackal]

Error/Bug?

[fading robin]

What ? Why you checking image from imgur e confused

Chatgpt send you that ? If yes then its haluciantions

[hybrid jackal]

No?

[hybrid jackal]

Chat GPT:

[fading robin]

So that are halucinationa i think never happend to me that

You are plus subscriber ?

[sweet ravine]

OK Thank You! I'm still experimenting with the design. I need to make the App familiar as you mentioned. I will also use more words than Emojis. Thanks a lot for giving me the tips!

[hybrid jackal]

Nope. Just on free Chat GPT 4.

[fading robin]

If so mayby tell it to use dalle-3 (note only gpt-4o (not mini) and gpt-4 can do that) or data visualisation dunnno

[hybrid jackal]

Yeah, GTP 4.

[fading robin]

Well if you are not plus subscriber you can't use dalle-3

[hybrid jackal]

Oh.

[fading robin]

Some peoples say it works also on free but man idk oai say no

[hybrid jackal]

Alright.

[hybrid jackal]

lol yeah abit confusing.

[fading robin]

Use microsoft copilot

Or microsoft designer

they can use dalle-3 and they free with bit limit

[hybrid jackal]

👍

[rustic marlin]

chatgpt hallucinated an invalid imgur url

neither a bug or error

just the AI making stuff up

[hybrid jackal]

Oh alright. Will this be enhanced so the AI can use more power to increase its knowledge?

[rustic marlin]

it isn't even a matter of knowledge, the link could actually be real when the dataset was made, and it is now expired

there is no way the model can account for that after trainned

it is actually more likely that this link used to work and does not anymore

the I will frequently embed imges from other sources tht change less often like wikipedia

[pale oxide]

I'm unable to get ChatGPT Plus to open in my web browser - Chrome. I have cleared all the cookies etc., but no luck. Any ideas to get back logged in?

[floral lodge]

Is anyone else just having major issues staying logged into Chat GPT. I cant seem to now log in at all. I am a paid customer and it just keeps loading. I can log into through my personal google account no problem. But When it comes to my paid business google account it just spins and thinks. any help would be graetly appreciated

[carmine edge]

I have a really important question: Does ChatGPT have a limit on conversation length?

[full pivot]

i get paid 4 time on open ai i try to find someway to get my 80$ refund somehow even 60$ the over load payment

i really need help

i have vouch of payment

can i get any help please it not my own credit card that the money has been snaked out of it

[trail cairn]

https://help.openai.com/ is the only place for payment issues, chat is bottom right

[full pivot]

i cannot chat any body there how do i contact them directly

[trail cairn]

there is no other way, the leave message appears after clicking enough options

[full pivot]

i did that but its for login account not for payment issue

[trail cairn]

also you need to be logged in

[full pivot]

i cannot use this type right now it told me wait at least 3 days

[trail cairn]

you need to wait more, it can even take a week or two

[full pivot]

i feel i get scammed because i get no invoice for the first 3 time i get paid and it keep saying the visa information is not right but it consume money do you even think they will give me back my money with no recipet excpet the bank message one i get fully scammed

[lost falcon]

has anyone come across this error? " RateLimitError: exceeded quota for this month."

there arent a lot of information online regarding this error

[wide garnet]

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."

[full pivot]

they send me they do a refund for the one i get the benifet and they stopped my open ai plus wow

i get surprised by the team

[e-mail]

i contact this email and that all i got

[trail cairn]

that email is not used or listed anywhere, where did you get it?

[full pivot]

I send to [e-mail] and he answered me

[full pivot]

.

With another email

[trail cairn]

dont put your personal info visible to everyone 🤔

[full pivot]

Okay

Its just i feel like i get scammed so i think i even might kicked from my company for doing such things

Why openai doesn't have like all company online chat team

And why is it consume money from credit card without asking for the numbers

[sweet light]

Unfortunately it is extremely difficult (if not impossible) to actually talk to a real support agent in my experience. Every time I have contacted support they have responded with an automated message that isn’t even related to my question

You will probably have better luck getting help though if you use ChatGPT to translate your messages into english because your question isn’t very clear

[full pivot]

So they only talk in English

Okay thanks for letting me knows

[full pivot]

Love u

[sweet light]

I don't know if that's true, I was just making a suggestion based on the screenshot you sent of the message you sent to support

And what you've been saying here

Someone here might be able to give you better advice if you could more clearly explain the problem but like someone said, if it's a billing issue probably only support can help

[gilded slate]

Hi its been day 2 im having this error
Error uploading file to OpenAI: Error code: 400 - {'error': {'message': 'Files with extensions [none] are not supported for retrieval. See https://platform.openai.com/docs/assistants/tools/file-search/supported-files', 'type': 'invalid_request_error', 'param': 'file_id', 'code': 'unsupported_file'}}

and this is my code

import os
from fastapi import UploadFile
from dotenv import load_dotenv
import openai
from openai import OpenAI

load_dotenv()
openai.api_key = os.getenv("OPENAI_API_KEY")
client = OpenAI()

async def upload_file_to_openai(file) -> str:

    try:
        content = await file.read()

        response = client.files.create(file=content, purpose='assistants')

        file_id = response.id  # Use attribute access instead of indexing
        print(f"File uploaded to OpenAI with ID {file_id}")
        return file_id
    except Exception as e:
        print(f"Error uploading file to OpenAI: {e}")
        return None  # Ensure you handle this case in your calling code
    

async def create_vector_store_with_file(file_ids: list, config):
    assistant_name = config['company_name']
    try:
        vector_store = client.beta.vector_stores.create(
            name=assistant_name
        )
        vector_id = vector_store.id  # Use indexing to access the vector ID
        print(f"Vector ID : {vector_id}")
        for file_id in file_ids:
            vector_store_file = client.beta.vector_stores.files.create_and_poll(
                vector_store_id=vector_id,
                file_id=file_id
                )
        print(f"Vector store created with ID {vector_id}")
        return vector_id
    except Exception as e:
        print(f"Error creating vector store: {e}")
        return None  # Ensure you handle this case in your calling code

i tried everything available on the documentations
Im uploading txt file

[manic shell]

since around a week I can't login into my OpenAI Account anymore...
I am normally using Google to login into OpenAI, but then it stopped working in Brave on my Windows PC and on my phone. I was still logged into the ChatGPT ios App.
I then switched to Chrome where it seemed to work.
It stopped working in chrome aswell after a few days, only temporary fix was to delete cookies for chatgpt.com and openai.com.
Now that stopped working aswell.

Does anyone know a fix for that?
I've switched browsers as I said, the issue happens on 3 devices and in 3 different networks only when logging in with google.
There is no error, it just loads forever...

[ashen umbra]

hello

[pseudo sky]

I'm adding streaming to my app, and while reviewing the documentation I noticed a function that isn't in the API Reference: https://platform.openai.com/docs/assistants/quickstart?context=without-streaming

What does the function create_and_poll do differently than create?

[heady cypress]

help it says my i have reached the limit for GPT-4o. how do i back to GPT-4o mini?
i usually change the prev question i asked to GPT-4o mini and that worked but now i cant change my prev question to GPT-4o mini as i have attached images which arent supported my GPT-4o mini.

[royal scroll]

Are you a free user

[heady cypress]

yes i am

[royal scroll]

Is there not a drop down bar at the model name

A small \/

[heady cypress]

which one? the one below has the GPT-4o mini grayed out...

[royal scroll]

Then mini isn’t available to you atm

Try refreshing

[heady cypress]

atm? but the reason i am getting is "this model doesnt support file attachments"

[royal scroll]

It doesn’t

Only 4o ordinary

[heady cypress]

yeah and i am fine with that. i just wanna change back to it

[royal scroll]

If that can’t be picked then your usage is up

[heady cypress]

since i have asked a question with file attachment before, will this conversation be permanently locked in GPT-4o?

[royal scroll]

Yup

As far as i assume that’s indeed the case

[heady cypress]

is it possible to delete that question...?

[royal scroll]

Have you tried?

[heady cypress]

i dont see an option to do so

[royal scroll]

Hover over the message

If you can’t delete it there then, No

I suggest starting a new chat

That may solve the issue at hand

[heady cypress]

but i asked so many leading questions... all that will be gone then

[royal scroll]

🤷‍♀️ idk what you want me to say

[heady cypress]

cant be helped then... thank you so much help

[royal scroll]

Alternatively you can wait for refresh

[toxic sky]

Let me explain why it's grayed out.

[heady cypress]

true but that chat has a permanent limit now... which is kinda annoying

[royal scroll]

Wait for grandell to write that essay

[toxic sky]

They do this sometimes when they are releasing model updates or the model is temporarily down.

This is more frequent for free users.

[royal scroll]

I can imagine

Does it mention update below the input bar

[heady cypress]

so the reason "doesnt support attachments" that is being shown to me is like a default reason or something?

[heady cypress]

no i dont see anything like that

[toxic sky]

Refreshing won't work, it normally gets back online within an hour.

It didn't say it's updating or whatever, they just grey out the model selection.

[heady cypress]

ok that means i can switch back to the free model after an hour?

[toxic sky]

You should get access again soon.

I would recommend doing something else while you wait.

[heady cypress]

nice thank you so much!!!

yeah i will

[toxic sky]

As a plus user we have access to enough models that when one or two are down we still have some available to access it.

[heady cypress]

oh nice! i am just a student tho so maybe in future i will get plus

[toxic sky]

Yeah, education comes first.

[heady cypress]

i did get access to chat again but i am still unable to change models... guess this chat is stuck with this limit permanently😭

ok i found a round about sol XD
i went to back to a question that i asked before the question with file attachment and edited the question by adding a single character. this like branched out into a new conversation which let me change the model again!

[trail cairn]

if you are a free user then you dont get to choose the model, it will force you to use 4o quota before switching to mini. it did the same with 4o and 3.5

"Users on the Free tier will be defaulted to GPT-4o with a limit on the number of messages they can send using GPT-4o, which will vary based on current usage and demand. When unavailable, Free tier users will be switched back to GPT-4o mini."

[heady cypress]

yeah but since my conversation had a file attachment, it didnt give me access to switch back to 4o mini.

but as soon as i did those shenanigans i got the option to change the model again

[trail cairn]

yep that's by design for now if you use the extra features, it may change if mini gets some file support

[heady cypress]

yeah i will wait till file support is on 4o mini or if they put out GPT -5 and maybe make GPT-4 free XD

[regal lynx]

I hope 4o mini gets file uploads since it's part of the omni family

[slow rose]

anyone else not able to get on chatGPT?

[pseudo sky]

class EventHandler(AssistantEventHandler):    
  @override
  def on_text_created(self, text) -> None:
    print(f"\nassistant > ", end="", flush=True)
      
  @override
  def on_text_delta(self, delta, snapshot):
    print(delta.value, end="", flush=True)
      
  def on_tool_call_created(self, tool_call):
    print(f"\nassistant > {tool_call.type}\n", flush=True)
  
  def on_tool_call_delta(self, delta, snapshot):
    if delta.type == 'code_interpreter':
      if delta.code_interpreter.input:
        print(delta.code_interpreter.input, end="", flush=True)
      if delta.code_interpreter.outputs:
        print(f"\n\noutput >", flush=True)
        for output in delta.code_interpreter.outputs:
          if output.type == "logs":
            print(f"\n{output.logs}", flush=True)

So is there just no documentation at all for this?

[valid geode]

I cannot login to my ChatGPT account. It does not load. Do you have the same issue?

[pseudo sky]

I'm on, no problem

[median pier]

Just discovered the discord. Is there a way to chat with the DALL-E bot in private messages like with Midjourney?

[timber shadow]

There is not, it's only usable in the public channel #image-bot using the /draw command -- 5 uses a day! OpenAI's main platforms for using DALL·E 3 are ChatGPT and via the API, the bot on Discord is just a nice lil bonus 🙂

[median pier]

Thank you

[night spear]

I need help getting refund please

anyone send link or ?

[onyx kraken]

https://help.openai.com/ bot bottom right. Select messages then click through the choices until you can leave a message.

Check there later, when a human helps you there's a conversation there. For me, I don't get emails for it.

[timber shadow]

To add on to what Esk said: here are instructions for requesting a ChatGPT Plus refund:

• https://help.openai.com/en/articles/7232895-how-do-i-request-a-refund-for-chatgpt-plus
• (if in Europe) https://help.openai.com/en/articles/7242627-as-a-european-customer-how-can-i-request-a-refund
  Note that you need to be signed in before the correct option will appear for you in the chat. If not already signed in, sign in first at platform.openai.com, then return to the chat widget on help.openai.com. Also note that it looks like the option says "Payments and Billing" now, not just "Billing" like the help articles say, but the refund options come after that!

[sonic solstice]

I keep having trouble accessing ChatGPT these days. I signed up through a Google account, but I am unable to log in to ChatGPT because I am in indefinite standby at the time the account information on my Google account is transferred.
Has anyone been experiencing this issue or solved it?
Or is it possible to log in to ChatGPT in any other way?

[potent kayak]

Any update on 4o voice release? They said a month which is today, Thursday.

[soft gull]

How can I set GPT-4o mini as a default model?

[bronze charm]

Use the following URL:
https://chatgpt.com/?model=gpt-4o-mini

[pine sand]

Hello, I've been having trouble accessing chatGPT on Chrome. I've tried two other browsers and I can't. I've tried clearing cookies, and it works for a while, but then I can't access the site anymore and it keeps loading forever. Could someone help me?

I'm tired of having to relog into all my accounts so I can use CHATGPT after clearing my browser cookies.

[pine sand]

I have the same problem!

[lofty junco]

Hello i need help, I have had problems logging into my CHATGPT PLUS account for weeks, the only way to get me to the login area is by deleting cache but I don't get past the login area either.

[lofty junco]

I have the same problem!

[lofty junco]

It stays like this

[drifting tangle]

+1

[forest solstice]

Major outage right?

[modern wedge]

I had the same problem, and still have again. For some reason it is working by using mobile network but not regular network at home. Try that as well. I tried contacting OPEN AI and I followed instructions, still doesn't work. Frustrating. @lofty junco @drifting tangle If anyone has solved the problem - help would be much appreaciated. Thanks

[drifting tangle]

Thanks. All I need is to use the "InPrivate mode" of the browser and everything works like a charm.
That is, cache problem.

[modern wedge]

Tried not working

Thanks @drifting tangle

[lofty junco]

Yes, it works in incognito mode and the same with a mobile network. The issue is that I pay the extra because I work with this, they should have support for those of us who pay.

[short relic]

Hi all where can I get proper help/support from Open AI? I've just signed up and it says my phone number is already in use, which should not be the case. I am concerned someone has used my phone number. It was hard to find but eventually I found the support chat but it just doesn't seem to even work.

[lofty junco]

I don't think it's just a cache issue, I have two accounts, one person with FREE CHATGPT and another corporate account with CHATGPT PLUS, both with Google.

The free one works perfectly, the corporate one is the one that fails.

[lofty junco]

I have a similar problem, I have no way for them to respond.

[short relic]

Phone number association? I seem to be getting through chat if I just keep clicking the message, that seems to send it again (discovered that by accident... lol)

[rugged wing]

Hi i want Play a Game but thé famé use chatgpt And thé Game require OpenAI api key I have créâted one but not thé key dont work I have free plan And not monte And I want Play to thé famé Please if you can help me pr give me openai api key work 🙏🙏

[lofty junco]

Are they supposed to answer these problems here?

[drifting tangle]

Ah, great to know that.

Thanks.

[drifting tangle]

Not sure.

[drifting tangle]

I am sorry to hear that.

[modern wedge]

It started when the outage happened and now is still an issue. So frustrating.

[modern wedge]

I agree. Same here. I cleared cache, cookies all that it should be. I think that it has some issue with Windows, Google, security and network, like it is all in one problem. It is so annoying that you pay for something and it's not working.

[trail cairn]

api is not free so it wont work until you buy credits

[lofty junco]

The issue for me is that if I am paying the extra they should provide some direct means of communication.

[modern wedge]

same I couldn't agree more.

[lofty junco]

Do you know if these people respond here?

[modern wedge]

@lofty junco I tried however, I noticed that they don't. I also tried on Open AI platform and I can see it is a bot

[rugged wing]

I dont have money.. great person can give me api key Please 🙏

[charred parcel]

nobody should share API keys, also it's not permitted to ask to use someone else's services

[patent kernel]

Chatgpt google login issues

[rich pivot]

Anyone have same issue as me? I am a gptplus user, but I got banned recently with no reason, and there’s no customer service to talk to

Customer service just ignore appeals

[trail cairn]

replies to ban issues can take up to several weeks

[rich pivot]

Is it common for people to get banned? I am using it quite normally like asking for option strategies/ asking for coding/ asking for image generation, I don’t see why I can get banned on those use case

[trail cairn]

yes, mostly due breaking something from these https://openai.com/policies/usage-policies/

[floral lodge]

Been having the sam eissue!

[ruby vortex]

how long can i keep my hisotry of past conversations on chatgpt before those are deleted?

Will deleting more recent ones free up the history or old ones are deleted after a certain time?

[pine sand]

Has anyone gotten any information about why this problem is happening when accessing via browsers?

[steep vapor]

I recently changed google accounts. As in uncoupled an alias (domain) from one google workspace account to create a new workspace account. Now I'm wondering how to switch my OpenAI account to the new domain (new google workspace account) without losing my subscription and custom GPTs. Any ideas??

[urban hedge]

Same issue

[onyx kraken]

Presuming you don't choose to delete them - for me my oldest conversations are still available (sure take a while to scroll to :P). If you choose to delete conversations, for you they're immediately unavailable, and this is the info we have about how they are removed:

"Deleted chats are hard deleted from our systems within 30 days, unless they have previously been de-identified and disassociated from your account or we have to keep them for security or legal reasons.

You can not recover chats once you delete them. If you want to remove a chat from being visible in your chat history but retain it in your account, you should use the archive function."

https://help.openai.com/en/articles/8983778-how-are-files-vs-chats-retained

[silk pier]

hey, i need an expert in custom IA development to help a low level user like me through a huge project about my professional scene (im a professional poker player and a teacher for others professional poker players. i have a level of expertise in a couple other stuffs aswell.) im kinda learning fast with less then 3 months of chatGPT usage, but i am looking for a more methodological approach of learning through guidance.

[white pike]

i need help with my model and sharing it

with my team

[fast reef]

Hello, I just want to ask when will the new update will be released that will include screensharing?

[bronze charm]

Alpha starts next week for certain Plus users, general availability in fall

[timber shadow]

Hey! Is it a single chat on ChatGPT, a GPT you made, or something else like a fine tuned model via the API?

[gray glacier]

Hello!
am I the only one who gets this message at any first message I send in a conversation?

[wind pelican]

Access from Power Automate to my Assistant (Custom GPT)

[zinc compass]

Is the ‘partial account suspension’ a thing? Because for some reason I can't upload files to ChatGPT from my account, but I used to be able to (I'm a free user, and used GPT-4o for that). I simply don't have the button now. Also, I can't access custom GPTs. This only applies to my main account, because when I made a second one, using my other mail, the button for uploading files is present, as well as GPTs discovery. I'm mainly interested in file uploading for breaking down images into prompts to feed it to other image generation AIs, but still... that's weird.

[trail cairn]

its not a suspension, it just means you hit the max allowed limit

free users have several different limits

[zinc compass]

I know about the limits thingy. About 10 msg and 2 files (at least it was like that at the beginning) per limit period. I'm saying that I don't have the file upload button anymore. As well as access to GPTs.

[trail cairn]

there is also a daily limit for file uploads, also file uploads require you to have access to 4o but if its on cooldown you will be switched to 4o mini which does not support files. on top of that all 4o has a monthly limit

[zinc compass]

It happened several months ago (after May 29th announcement) and since then I don't have these features. No file uploads and no GPTs.

Kinda long for a "monthly limit", don't you think? Also, I uploaded only 8 files in total. Weird number for a limit.

And this doesn't explain why I don't have "My GPTs" button in the menu.

[trail cairn]

free users cant create gpt's

[zinc compass]

...just look at the screenshots.

[trail cairn]

show the top left model menu for the one with my gpt's

also how/why do you have 2 accounts? evading limits is against tos

which could explain the other account losing access to multiple features

with the affected account when you send a message check the model menu below its reply

[zinc compass]

Made for tests because I don't have the damn button.

And here's the screenshot.

[zinc compass]

I made it AFTER I lost access to those.

[trail cairn]

gpt's, file uploads etc. all require 4o access so check the model picker first (the star icon)

[zinc compass]

Where can I access this star icon for new chats?

[trail cairn]

below chatgpt reply

free users dont get to choose which model to use beforehand

[trail cairn]

but it will show which was used

then its probably the file daily limit or some other one 🤔

[zinc compass]

#community-help message

When I get file uploads back - I will delete my second account, no questions. But until then - no way. Because that's simply unfair.

[trail cairn]

did you use the account since then or had a long pause between now and then?

also did you the both accounts on the same browser and settings?

[zinc compass]

Nope. Just logged in. And never used it 'til today. Just made it in may, saw that the features are there, and that means something wrong with my main acc, and never logged into it again.

[zinc compass]

Tried different browsers. Tried VPNs. No difference. Main has no button, second one has.

[trail cairn]

the announcement was a rollout and afaik they did not roll the features to dormant accounts. someone else had similar issue

[zinc compass]

Then how was I able to upload files before on my main acc?

[rugged wing]

What is chatgpt

Is someone behind it? how do they type fast

[zinc compass]

A really, really weird thing.

[rugged wing]

Yeah everyone keeps making a fuss about it on tiktok but i never understood it

It feels like I am a boomer

[trail cairn]

bunch of people did but they got access back later but they never stopped checking. you might find them with search

[zinc compass]

I use ChatGPT on a nearly daily basis, how is my main acc considered "dormant"?

[trail cairn]

"Nope. Just logged in. And never used it 'til today. Just made it in may, saw that the features are there, and that means something wrong with my main acc, and never logged into it again."?

[zinc compass]

I even asked support on the website and got no help...

[zinc compass]

That's about my SECOND acc.

[rugged wing]

Are you ukrainian?

[trail cairn]

ok but i was asking about the old one since the new one could not have days between usage

[zinc compass]

I tried multiple things to get features back on my main (browsers, VPNs), then I made a second one, saw that it HAS those features, logged out and hadn't used second one since.

[fading robin]

Chatgpt is a LLM it run on superfast H200 gpus that why it type so fast. Its just a AI

[trail cairn]

no people are writing it, its ai

[rugged wing]

Oh ok

[zinc compass]

How old are you, "boomer"? XD

[rugged wing]

17

☠️

Boomer tech wise

[rugged wing]

Is anyone else having trouble login in via OpenAI to ChatGPT?
I am asking this since I am having issues.
I am on Brave.

Any other browser doesn't work either, Chromium based nor Firefox.
I don't have any VPN turned on, my Office PC has the same issue.
Sometimes it works, sometimes not.

[hushed crane]

Hey, is there a way for me to pay the 50$/mo like I am two people in order to make use of the teams subscription? I'm not 2 people, I'm just one person, but the value is worth 50$/mo to me at least.

[upbeat shuttle]

I may need to change my email for my ChatGPT account soon, how easy is that process?

[onyx kraken]

Seems to be worth trying. You're charged by seats, with 2 as the minimum, but you appear able to add and remove individuals from seats, so maybe you're able to set it up and pay for 2 without actually assigning the second seat to anyone (yet), and just paying for it.

https://help.openai.com/en/articles/8792536-manage-billing-on-the-chatgpt-team-subscription-plan

https://help.openai.com/en/articles/8792828-what-is-chatgpt-team

This is a great question to ask the bot in the bottom right corner of either of those pages; select 'messages' and click through the choices until you can leave a message. Check back at the page to see the answer, it unfortunately doesn't seem to go to email (for me at least) like they say they intend it to.

[onyx kraken]

It appears to currently not be supported at all. https://help.openai.com/en/articles/4936827-how-to-change-your-email-address

However, you can create a new account, as that page discusses, with the new email.

[grim bloom]

I have not been able to login to chatGPT on PC for several days now using my google account. It works on my phone using the mobile app, but on PC I have tries several browsers, Chrome, Opera, Firefox and Edge, and each time I go through the login process it just loads forever on a blank page. Anyone else having the same problem?

[hoary anvil]

can any ai language models do a math problem like this?

[bronze charm]

GPT-4o and Sonnet 3.5 were able to solve it correctly for me

[hoary anvil]

What prompt ?

[bronze charm]

"Solve for perimeter"

[hoary anvil]

what answer did you get? got 39 but its wrong

[daring sequoia]

Does anyone have issues with accessing the site? It just keeps on loading on my end.

[nova thorn]

I have my own API key but when I run setx OPENAI_API_KEY "your-api-key-here" it says 'setx' is not recognized as an internal or external command

[trail cairn]

where did you run it?

[nova thorn]

command prompt

i can show u in like vc

[trail cairn]

in windows? which ver

[nova thorn]

windows 11

do i need to install like something called setx?

[trail cairn]

no, its part of windows

[nova thorn]

ok

im following this currently: https://platform.openai.com/docs/quickstart

[trail cairn]

it might not be available on limited/smaller editions of windows

[nova thorn]

windows 11 is updated tho

[trail cairn]

yeah i mean limited access or the special versions like S or some education ones. setx is very old command but you can do the same it does via other ways

[nova thorn]

do you know other ways?

[trail cairn]

open start menu and type environment variables

[onyx kraken]

I can get 4o to consistently solve it correctly, but I have to prompt it similar to this:

[Uh-oh! This image shows a misleading math problem. The most obvious and simple way to 'solve' it ignores vital info!

Can you find the trick and correctly find the perimeter?]

Which gets part of the answer to account for:

[However, we need to account for the lengths not directly given:

• The vertical segment connecting w to z at the bottom (let's call it a).
• The horizontal segment connecting x to y on the right (let's call it b).

To find these:

• a=w−y=12−7=5
• b=x−z=15−5=10

So, the perimeter of the shape is 54 units.]

Even there it doesn't do the simpler step of presuming that 2w+2x will find the perimeter.

[silent topaz]

Friends, have the advanced sound features of gpt 4o started to be distributed to certain plas users or is there still no update news? Also, is there any information about how it will be distributed, to whom it will be distributed, or how it can be applied if it can be applied?

[grim bloom]

Same here, several days now. works on the phone but not in windows.

[eternal depot]

People who know Russian?I'm Russian

[rugged wing]

How to fix this, when the text is long, the keyboard starts to mess up with it?

[bronze charm]

quite the prompt you got there

🥶

[wet nacelle]

💯

[modern wedge]

if anyone is having problems logging by using Chrome - this is what was an issue in my case. There was some experimental protocol enabled. Here is the screenshot of the command, just disable it and it should work after it.

[hoary anvil]

im trying to make an app that will extract just an answer from AI, i ask the AI to surround the answer in <a> tags, and i also ask the AI to give me the whole text because that way it will fully think out the problem. If i ask it to put the answer first, and then explain it, and just set my maximum tokens lower, will it change the answer quality?

[sudden atlas]

Likely yes, COT prompting works best when the answer comes last

[hoary anvil]

thanks! whats COT

[sudden atlas]

Chain of Thought - it's a prompting technique where (normally over multiple chat turns) you get the LLM to show its thinking before answering - example shown in this comparison

[hoary anvil]

damn thats super cool!

[exotic quail]

Does the use of lists, code blocks, italic and bold letters improve the quality of the prompt? I've written a prompt and asked ChatGPT to see if there could be any improvements. All he did was fix some grammar issues and added lists, bold, code blocks etc etc. Now im wondering if this actually improves the prompt or if it does not help that much and just uses extra tokens?

[foggy mortar]

to some degree yes, like putting quotes for an important task/word has in my experince helped. like for example you need to do "X" when "Y" happends. then it for me atleast has worked better then not quoting it. so im assuming same goes for code block and so on to make it understand lang or whatever is in the blocks

[exotic quail]

Interesting, Thanks!

[sudden atlas]

To some degree yes - and this differs between let's say Claude (loves XML) and GPT-4o (prefers Markdown) but the magic is in the other techniques

• In-context examples - adding a few user / assistant message pairs where you put perfect examples (great for this sort of thing)
• Chain of Thought - giving it some linear judging process that it has to go through before it can respond

[mystic canopy]

Thank you so much ! 👍🏻

[silent topaz]

How do we get into the Searchgpt beta?

[wise zinc]

https://openai.com/index/searchgpt-prototype/

[sleek bronze]

Hey, is there a way to disable certain memories for certains chats? I mostly use chatgpt for fun storytelling, and the problem is that the memories are synced between different stories, which create some weird results, how can i keep memories local to a single chat?

[native seal]

I am working on a game that explains geneartive AI. It will be sold commercially. For this game we want to use some images and short vids from YouTube.
QUESTION FOR OPENAI: Can we use fragments of demo's in our AI game?

[trail cairn]

i dont think you will get an official answer here. in general it would be best to generate the material yourself with the features that are actually available

[native seal]

Thank you.

[daring totem]

AI humanizer isn't working Even though I have purchased the ChatGPT premium whats the solution should I cancel my subscription?