community-help

[icy maple]

Ok, thanks!

[naive shadow]

Anyone with a good grasp on twilio and websocket streaming, please dm me.

[marsh cloak]

hey everyone im sure this was asked a ton already but for some reason i can't find it.

How can I see my limit and when it resets for GPT4?

I been using it and for the last few days it says I have hit my limit and need to wait for the reset. I want to know where i can check this limit and reset times.

[bronze charm]

The limit is 40 messages / 3 hours: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4
It resets at the time of the 40th most recent message + 3 hours

[upper hatch]

I am using Zoho Meeting in my business. It has a feature that uses ChatGPT to transcribe and generate summaries for meetings. I am on the free ChatGPT version and added the API key. Does paying for ChatGPT Plus provide higher usage limits or is that completely separate from the API fees?

[timber shadow]

Hey! ChatGPT usage/billing is totally separate from OpenAI's API usage tiers. You can read about how the latter works here! https://platform.openai.com/docs/guides/rate-limits/usage-tiers

[upper hatch]

Thanks. I'm haivng a little trouble understanding the limits. How would I know how much it would cost to transcribe and provide a summary of a 1 hour meeting? It says tokens per minute, but how many words is a token?

[timber shadow]

Do you mean a transcription of an audio recording of the meeting? Speech audio-to-text transcription via OpenAI's API would be accomplished with their Whisper model. Whisper pricing can be found here in the Audio models section: https://openai.com/pricing

As for providing summaries: that would be accomplished by feeding text to one of the language models (pricing for those also found in the link above!).

As for tokens: there's no exact rate for tokens-per-word, but a rough estimate is that 75 words contain about 100 tokens. You can experiment with how words convert to tokens on this page: https://platform.openai.com/tokenizer And here's a general article about tokens: https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them

So in summary: speech-to-text is one thing (Whisper), summarization is another thing (GPT), and tokens are relevant to GPT!

[upper hatch]

Thank you. That was very helpful.

[marsh cloak]

so there is no actaul tracker to see how many messages out of 40 you are or when it resets

[bronze charm]

You basically count how many messages you have sent in the last 3 hours

= 40 means barred usage

[rugged wing]

@onyx kraken hiiii i have a doubt. I used /limit command here and it showed that 5 uses are remaining but I wonder how to use 🤔

[onyx kraken]

Hi! you would go to #image-bot and use the /draw command there, and put a prompt in so you can make images here... there in that channel, I mean, if you want to!

5 uses a day, that's your limit!

[rugged wing]

oh thank youuuu i was just wondering here

[rose void]

Hi! I have two accounts, one for personal use connected to my personal email and one for work reasons connected to my work email.
Can I access both accounts from the app? Thanks!

[alpine flame]

Has anyone found a good solution to test their OpenAI Assistants/GPTs solutions against a library of possible user queries?

[cinder birch]

Does creating a thread and adding messages to it cost money ?

I believe money is charged only when running the thread with an assistant, right ?

[uncut aurora]

You are talking about #1184149552639516732?

Or do you talk about ChatGPT?

[chilly tree]

have anyone tried gpt-enterprice ?

[winter ingot]

hi guys, is chatGPT down？I got a blank page

[topaz turtle]

Hello
Guys is there anyone who made a chatgpt(api) use tts and stt using whisper api?

[rigid musk]

Status.openai.com

[restive rune]

Hello everyone. My sidebar is completely empty even after a bunch of suggestions. Anyone else experiencing these problems?

[elfin hinge]

Same, can't see my precious chats

[somber karma]

I have two accounts and can't access either!!!

[timber shadow]

Is the behavior any different if you try signing in in a private tab, with all browser extensions disabled? Also, any VPN installed?

[tender path]

Anyone else facing limit issues, i got it twice and now the timings has increased too like i have to wait two hours for it to reset

[somber karma]

Yes. All of that. And I wasn't even logged in long. Now I have an empty GPT dashboard

[merry widget]

my god. Never sign the POE AI. It's just awful and it gets literally everything wrong.
GPT 4.0 is the best

[tender path]

Same my dashboard is empty too. I cant even log out lol

[uncut aurora]

Looks like there are some problems - I think you will see that in a few minutes at https://status.openai.com/ too.

[tall drum]

Could anyone tell me how to delete threads from my dashboard? I have several old assistants, and don't want the previous threads accessible anymore

[trail cairn]

you can only delete them via api

[tall drum]

well bugger, thanks

[tall drum]

Well orphaned threads eventually get deleted? Im not a programmer, was just playing around with the assistants in the playground and had a ton before i figured out how it worked

[trail cairn]

i assume so but since its in beta it might not happen yet. if you only have few threads you could delete them without writing any code by using curl and thread id

[tall drum]

do you have a reference for how i could accomplish this?

[trail cairn]

https://platform.openai.com/docs/api-reference/threads/deleteThread?lang=curl

[tall drum]

perfect, thank you

[exotic yew]

Why is BING Only generating 2 images??? I used to always get 4 images, now it's only 2... Is everyone getting the same? On copilot it's generating 4 pics, but not on bing. I prefer using bing.

[trail cairn]

bing? that's a microsoft product so you need to go ask their discord

[grand hollow]

Is whisper bugging out for anyone? I used mobile ChatGPT app for transcription and it doesn't go through, not sure if it's cause it's a 13 minute audio or if servers are down?

I hate how they don't let you download the audio file, it's like, why did I waste 13 minutes talking 😭

[exotic yew]

But it's Dall-E... Hmm

[trail cairn]

dall-e model is made by openai but bing is by microsoft and rules are set by them

[exotic yew]

Ic. thanks for the reply! Btw, is Dall-E 4 planned by any chance?

[trail cairn]

its meant for regular conversation. you cant/barely can input 13minutes worth of text in the web ui either

[trail cairn]

never seen them mention dall-e 4 and afaik the 3 version is not even fully finished yet

[grand hollow]

So you think it's not processing it? Man that sucks... Why can't they let me download it, I could edit it into two audio files or something, or why they aren't batching it on their end... Arghhhh

[split reef]

Anybody missing their sidebar as well? I have plus.

Can't relog as well. Incognito tab same thing

[trail cairn]

doubt it since it was not designed for such use. even the api services have limits for stt processing

[lucid needle]

Yep same issue! I tired multiple browsers, private windows, and still the same sidebar is completely broken.

[tender path]

Same issues and also limit reach

[lucid needle]

Ayyyyeeee its fixed

[split reef]

Great..

[split reef]

It is not for me

[lucid needle]

It just fixed itself for me

[split reef]

Lucky you

Almost stuck at it for 30mins

[split reef]

Were you on the limit?

[lucid needle]

I was rate limited yeah

[split reef]

Maybe that expired?

[lucid needle]

Yeah it did

[split reef]

Then that was most likely it

[fierce basin]

Hey everyone , I'm looking for a senior dev to help me in clearing my doubts.
I'm a graduation student and i want to crack a SDE job. I know C, C++ and quite good with the syntax. Though I didn't started DSA that much , i know basic sorting and searching. I was thinking to level up my self now but i'm confused with the current market scenario. I was thinking to jump into golang as it's good for backend and web services and can also be used for DSA for job prep. Is it a right choice or should i continue myself in C family?

[rare epoch]

Hey all - anyone else have the issue where uploading new knowledge files causes an error/issue with your GPT Plus subscription such that your custom GPTs and access to GPT 4 has disappeared? In the attached screenshot, I've just landed after logging in.

[rare epoch]

And yes, already cleared cookies, other browsers, devices, etc., and all the other troubleshooting methods. Seems like my iOS app is fine but no matter what web browser I use, no dice.

[nova shale]

Is there any way to contact OpenAI about billing issues? I have been using the same credit card to pay for API usage for several years. Now today I am getting card declined errors. I contacted my bank about this and they don't even see any requests being made that are being declined so the issue is clearly on OpenAI's end. I tried removing the card as a payment method and re-adding it, but now I am even getting the card declined error when trying to re-enter the card info. This is very confusing. I want to give them my money but they wont let me?

[timber shadow]

Hey! You should try reaching out to OpenAI support at help.openai.com. Start a new chat with the chat widget in the bottom-right of that page, click through the most relevant options, then you'll eventually find a path that lets you type in your own message. This will submit a ticket that someone can review and get back to you on. It won't connect you to someone right away, but you should get a notification when you get a response!

[nova shale]

Any idea on what the eta might be? This is currently a blocker for me as I can no longer make api requests.

[timber shadow]

It can vary, it might be a couple days. There might be a time estimate at the top of the chat saying what the current average response time is?

[split reef]

It will go away once your cap is over

I had the same thing happen earlier today

It magically reappears once the limit is over

So by now, it should have returned

[rugged wing]

hi

sory wrong group

[naive shadow]

hello can anyone help me , i am doing real time transcribing of calls using twilio and assemblyai but i am not able to play back any audio as a response to that through the websocket

[upbeat burrow]

hey

i got a lot of error in my gpt it cannot read dlls ```ModuleNotFoundError Traceback (most recent call last)
Cell In[5], line 10
7 sys.path.append(assembly_path)
9 # Import the HudManager class from the assembly
---> 10 from Assembly_CSharp_2024_3_5 import HudManager
12 # Get the methods of the HudManager class
13 hud_manager_methods = [method for method in inspect.getmembers(HudManager, inspect.isfunction)]

ModuleNotFoundError: No module named 'Assembly_CSharp_2024_3_5'``` when i install clr package and pythonnet nothing work

https://chat.openai.com/g/g-OUjY4dybi-amongus-modding/c/4b899851-b2ba-4756-9554-a01621d65a86

[naive oyster]

Hello, don’t know if this is the right place, but i am currently trying to implement memory within my chatbot app to allow it to be able to understand context. I have setup the database and successfully perform vector searches, but i am having trouble understanding what to include in the context in the prompt to the LLM. Could anyone help?

[steady hare]

Greetings, is there a way to get free trial to test open ai api?

[trail cairn]

nope but the min payment is only $5

[ripe edge]

Hey guys. ChatGPT made for me an excel workbook but when I try to download it from the link it send me, I get the error: Failed to get upload status for /mnt/data/Coffee_Shop_Costing_Application.xlsx. Can anyone pls help me? Thanks in advance

[dull yew]

anyone also have this issue? or is this normal

[trail cairn]

normal, 3.5 training data is up to sep 2021

[dire elm]

Do you find a solution?

I meet the same problem.

[tropic spade]

what is happening my Chatgpt chats keep switching back to 3.5 and being suggested as the default chat ? after i switch it to gpt 4 and type in my prompt upon ENTER it switch back replying with GPT 3.5 ?

[jolly kelp]

is anyone getting this??

[ashen valve]

When will this be resolved?

[broken anchor]

servers down I think

[jolly kelp]

It was fine minutes back

[waxen wharf]

Fix chatcpt please, I'm just a stupid human; I've forgot how to do this programming thing manually

[wheat walrus]

oh okay i thought it was me. i guess it is down. how disappointing

[cerulean pilot]

what does this mean?

[plucky jetty]

its back

[sonic tendon]

Im unable to access settings and all my chats are disappear 😐

[lyric tide]

I am having this same issue

[cerulean pilot]

big down (openai is down rn)

[lyric tide]

thats like the internet being turned off, what to do 😅

[cerulean pilot]

im in your place exactly 😂

[sonic tendon]

ok is returned operative right now 🙂

[sonic tendon]

maybe we are really inside the matrix XD

[sonic drift]

lets goo!

[umbral wave]

yes it's back

[brazen notch]

bruh it is not

[sonic drift]

yeah, saw it too now haha

status.openai.com

[lyric tide]

Its being intermittent

[winter ingot]

Experiencing it too, from Malaysia

[sonic drift]

Let's just wait, I guess...

[foggy nest]

Ah, good stuffs. Thought it was just me

[jolly kelp]

is it happening again?

[foggy nest]

Tis part of technology, my fellow human

Patience we must have, working on it they are

[jolly kelp]

Yeah

[hybrid osprey]

I'm getting it. Glad it wasn't something I did ... or didn't do.

[jolly kelp]

is it fixed??

[winter ingot]

as of now, NO 😦

[rapid grove]

chat gpt not responding

[jolly kelp]

It works for me

[rapid grove]

not for me

[rugged wing]

Is ChatGPT down?

[rapid grove]

maybe yes

[rugged wing]

I needed it urgently

[rapid grove]

me to

[foggy nest]

Yo, refresh your page and clear your cache. It's fixed

[sonic drift]

It is working for me. Cleared the cache and cookies as well(idk if you must).

[proper plank]

lol, there's literally no help with their stuff at all is there? I've been asking for help on the same thing for a year and all i get are AI email responses that have nothing to do with what I asked. exactly what am I paying for if I get nothing on when I need help?

[trail cairn]

you dont have any posts in #1070006915414900886 ?

[proper plank]

I stopped almost a year ago because I got nothing from anyone. I need help with my accounts

[trail cairn]

i know, i remember your issue

[proper plank]

awesome. so all there is is validation nobody can get hle

but no actual help

I hate this company

[trail cairn]

openai staff here responses mainly in #1070006915414900886

[proper plank]

they haven't for almost a year

I email, I reach out on socials .. I contact Microsoft .. I do everything everyone says but I get no help. It really shouldn't be this difficult to get solutions to problems with account(s) .. especially when uyou're reselling their products in the 1,000s

[trail cairn]

latest staff reply in there is from yesterday

[proper plank]

I'm going to stop and start using another LLM or something this is beyond pathetic. it's been almost one year

[rigid musk]

Microsoft have nothing to do with openai . Help.openai.com if you want to raise a ticket . And like radium said some openai employee do answer in bug report

[swift cradle]

"Why do I have the PLUS tariff plan, but only on the 3rd day, it tells me that the limit for using GPT 4 is exceeded?"

"Is there some kind of limit? There's no way I could have exhausted it in just 3 days, right?"

[trail cairn]

https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4#h_ef2559d888

[swift cradle]

[ocean crypt]

Hello Everyone
Nice to meet you
is there anyone who familiar with Langchain?

[naive shadow]

yes

[night leaf]

Hello, where can I find a Dall E bot?

[uncut aurora]

Hi you can access one for examle in #image-bot.

[night leaf]

from your reply I understand there are more than one. Right?

[timber shadow]

That's the only one in this server. You can also try DALL·E 3 for free with Bing's image creator. For paid options, it's available on ChatGPT Plus and on Copilot Pro, or you can use it via OpenAI's API.

[night leaf]

thank you

[rapid herald]

Does anyone know if the Assistants API with threads, when truncating the context, does it automatically add chat history to Vectors for RAG?

[gentle creek]

Is there a way to recover my chat history? One of my convo got it's history nuked hard with no way to recover it back, like few days of prompts gone when i edited a prompt in the ChatGPT app.

[rustic marlin]

it just deletes old messages

[rustic marlin]

i is a known issue, just wait, the chat history will be back soon

[rapid herald]

Oh. A lot of misinformation out there 😂 some people think it adds truncated context to vectors, etc

[rustic marlin]

this confusion might be because of this section: https://platform.openai.com/docs/assistants/how-it-works/managing-threads-and-messages

the Thread will attempt to smartly truncate messages
"smartly" does not mean it will add to RAG

it means that it will try to keep messages that are considered more relevant on the context

[solemn furnace]

Anyone knows what's the new OpenAI News RSS feed xml web path?

[orchid depot]

Is there a way to set up a token with limited credits ... for a team member?

[rustic marlin]

you can't set limits by team member, but you can set a limit to a project

go here: https://platform.openai.com/settings/organization/general and create aproject

define the limit and invite them to the project

[orchid depot]

Thanks! Do they have to have a personal account before I invite them?

[rustic marlin]

yes, they do

each user wil lstill use their own APIKeys

[tame pulsar]

I am constantly getting this message:
"Hmm...something seems to have gone wrong."

[rigid musk]

Do you have a vpn

[sage wing]

I tried to upload a picture to ChatGPT 4 and its working fine, but my own Costum GPT keep saying '' hmm.. something has gone wrong'' everytime i upload a picture, yesterday i used my own GPT and everything was working, i have not done any changes to it. Or using a VPN, i keep getting the same answer on my phone and pc, I even tried it using Mobil internet. How come i have this problem now anyone can help out?

[gleaming meteor]

been having this issue as well for going on 3+ hours now

[tame pulsar]

No

[rigid musk]

can you try firefox as test

[tame pulsar]

I tried other browsers. Chrome, yandex, firefox. All the same.

[rigid musk]

does it work with lte ( data) on cellphone?

[tame pulsar]

I just tried it, and NO, it doesn't.

[rigid musk]

can you try without login , just chat.openai.com in private browser

[tame pulsar]

I don't think that is possible. It is bringing me to a login page. Also, when it was working, it was pretty slow and lagging.

[rigid musk]

3.5 doesnt require anymore account

[tame pulsar]

Please give me a link that does not take me to a sign up page.

[rigid musk]

https://chatgpt.com/

[tame pulsar]

That takes me to a sign up page. Do you try these links yourself?

[rigid musk]

then i suggest help.openai.com if you have issue

[tame pulsar]

That is a useless AI chatbot. I have tried it already and got nowhere.

[rigid musk]

you can make a ticket request after answering

[hasty sapphire]

Anyone having issues with Chatp Gpt 4 right now?

[maiden carbon]

I cant seem to get the bot to do anything I want anymore without fighting it...

[south mist]

Hi all!

I'm having troubles with the chat
Did everything on the help centre and nothing helped.

Anyone knows how to fix it?

[somber karma]

WHAT IS GOING ON WITH CHATGPT!!!!! This is CRAZY

[sage wing]

Been having the same problem all day. I couldn’t find solution. Some say try to log out and login again, other said try to clear chrome cache. I did all that and nothing changed same problem.

[cold folio]

How does your issue happen?

[sage wing]

Every time i upload a picture or ask it to read an uploaded Dokument it says “ hmm.. something has gone wrong”, I just tried to upload a Dokument and asked to review it I got this message. “ internal server error” I got this 2 mins ago

I have tried this with Wi-Fi and without I get the same response.

[fickle shard]

Got a question about exporting the data from ChatGPT. The html file comes up blank when I open it after downloading. Is that a general bug or something to do with the fact the page is over 700mb? No idea why it's that large either.

[sage wing]

What does this mean?

[arctic tusk]

I saw this image on the internet and was wondering what like style is this?

[bronze charm]

Midjourney thinks it is...

[arctic tusk]

Thank you

[trim lion]

whyyyyy

[solemn furnace]

Now how to fetch OpenAI News RSS?
I have to intergrate them in my news app, RSS link is gone, and OpenAI news site is also protected by Cloudflare Anti-DDoS

Only working way to scrape OpenAI news is to scrape Twitter user page, but I really need the RSS link back online and open news page accessible from any program outside browsers

[sly kraken]

Hi

I'm having an issue as I've purchaed a year plan AIPRM PRO on 2nd May 2024 but it is still not been upgradeed to pro plan

Please help

[burnt temple]

hello, pls can anyone explain why i select gpt-4 start writing and gpt 3.5 reselected autom

[rigid musk]

if you used all your usage and you continued after the usage it switch to 3.5 but you cant switch back

you have to wait the cooldown

[rustic marlin]

what is AIPRM PRO ?

@sly kraken aiprm isn't tied to OpenAI in any way. if you purchased anything from this company you will not get any services or support directly from OpenAI

in case you where lead to believe that this company where part of any OpenAI services, I highly suggest you to contact your bank and cancel this payment

I just did a quick google search for the terms "aiprm scam" and found some worrying results

[lucid radish]

👋 Is it normal when I go to https://chat.openai.com/ it redirects me to https://chatgpt.com/?oai-dm=1

[uncut aurora]

Hello, yes it is.

[lucid radish]

Why isn't it doing this to my friend but me ?

I mean, I can't do anything

[uncut aurora]

Maybe he hasn't get the newest DNS-record yet?

[lucid radish]

Probably I guess

It is working for ya ?

[uncut aurora]

At the moment it looks good, yes.
Do you use a VPN or something else?

[lucid radish]

oh

I will check it

but wait, it is doing this on my phone too. Could be my country then ?

( Costa Rica )

[uncut aurora]

I'm not sure if there is a map or something else we can see the problems for some countries but it's possible sure.

[lucid radish]

oh welp

ima reboot my router to see

oh well

It is not working at all

[uncut aurora]

Maybe you tried it with a VPN. 🤷‍♂️

[rustic marlin]

Try setting your dns to 1.1.1.1

[lucid radish]

will try.

[lucid radish]

what did I do wrong

I cant believe I have to say this but I need to use edge.

[rustic marlin]

I would go with this and 8.8.8.8

[lucid radish]

It is only working on edge, not google. wich will be fine for now

[merry dew]

Hi friends. I am getting this **repetitive type glitch **for every message, as shown in underlined text.

It goes away when I refresh, as shown when I reset to the right, it reverted to normal. It also reads normally without refreshing, as in it does not read the repeats when I tap the TTS/ voice function.

This used to happen once / day at most, now it is every message for the past 3 days ar so.

any ideas for a solution? I have tried searching the boards but don't see any solution yet.

[shy drum]

Does anyone have the same problem? If I refresh the page afterward, the chat with the response appears in the history, but without refreshing the page, I keep encountering this error every time.

[sweet pulsar]

why does it keep saying that chatgpt refused to connect? (for the website)

[mellow vortex]

I was using GPT4 in a long chat. then my subscription ended, then it switched to 3.5 model , now i have resubscribet but can not change the model back to 4

also can not share chat , gives error creating link

Failed to copy link to clipboard - could not create link

[mental flicker]

Same here

[rigid musk]

Only ways it is to copy paste

[mellow vortex]

it will say too long blablabla

[proud shoal]

is chat down?

nvm fireforx just refusing to upload documnets for some reason

[plain trellis]

Hi anyone know how with function call feature, i can avoid "discussion" flow but instead have directly the final response ? ty

[tender musk]

Hey Everybody ! Hope you are doing well ! Is there a way to see how many conversations I still have available in my 3 hour window ? or that has changed ? Thanks for your help

[tender musk]

Hey I have noticed , if your chat has images it does give an error - does you chat have images ?

[flat tide]

why do I keep getting this message on gpt pro? Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[rigid musk]

Do you have a vpn ?

[flat tide]

yes

[waxen crow]

Guys. I have a problem. My google account that i used to start a gpt subscription was banned, so now i can use gpt 4 only on the devices where i was inside before getting banned by Google. How can i use gpt 4 on other devices? I want to solve this problem, maybe using gpt 4 in discord or i don't know, but i'm not getting a new subscription, because what i have now is started 5 days ago.

[rigid musk]

Vpn are known to cause issue. Make sure to disable it, and end in the task manager. Often the setting web browser protection are the first problem

[rigid musk]

Your Google account was banned or your openai account linked to your Google?

[waxen crow]

my google account

and i created the openai account using the google auth method

[rigid musk]

You have to discuss with Google to unban it. Or ask openai to change the login method .

Help.openai.com and raise a ticket

[waxen crow]

i have already tried discussing with google...

how can i change the login method?

[rigid musk]

Raise a ticket with openai

Help.openai.com

[waxen crow]

can i talk with the support inside this discord?

[trail cairn]

nope

[waxen crow]

i tryed to open a ticket, but i read that is impossible to change the login method

can i use the api of gpt4 on some discord bot?

[trail cairn]

only if you host one and pay for it

[waxen crow]

It doesn’t make any sense. If I had a problem with google why I cannot use my open ai account?

[trail cairn]

the google auth tokens will probably expire soon so you might want to cancel your sub before losing last access

[waxen crow]

So I payed for 6 days?

[trail cairn]

you paid for one month and will have have the remaining days left even if you cancel but if google blocks you from logging in you cant use it or cancel manually anymore

[winter dust]

Hi everyone. I have a SaaS with a few billion of tokens in text for 40+ languages. Due to the unique of terminology, if I use any translation service or GPT for English to one of the 40 languages, the results are not good enough. I can easily create a dataset where I have the same text in all the different languages. Is there any way to train GPT with that to learn translating my specific terminology and use it then for new translations?

[trail cairn]

use rag for the terminology while translating or how many words or tokens are you talking about?

[winter dust]

around 15 billions in english only and 1-2 billions on other languages + English. But the terminology is not something unique or used only in my SaaS. But imagine a word in English could be translated in 5 different words in French. I want to see if there is a way to train GPT through my old texts to understand which languages to use for those translations

[trail cairn]

you cant train chatgpt and fine-tuning would not work that purpose either. with terminology i meant field specific abbrivatios, names, etc.

15 billion would mean you want teach exact writing style and every possible word

[winter dust]

I see. I thought fine-tuning could be used for that. But what if I just create a glossary? Something like a list of 2000-3000 words that I want to be translated with a specific way in each language. Would that be able to be used somehow

[trail cairn]

then you can just use rag with keyword triggers

[winter dust]

great. Will do my research on that. Thanks a lot

[sacred apex]

Hello. I am attempting to export my ChatGPT data and reupload it to ChatGPT for...inspired self-analysis? But when I upload the data as received, ChatGPT doesn't seem to know the first thing to do with it! And unfortunately neither do I (😅)...This same issue occurs with any exported social media data that I've tried to upload to ChatGPT (but specifically 𝕏). Does anyone have any experience with this, having done it, or otherwise pertinent suggestions that I might try? Thank you!

[sacred apex]

I may have figured things out! Thankfully.

[tall drum]

Hey folks any idea what might be causing this? 3.5 model works fine, gpt4 does not work. I'm on an enterprise account, so it shouldnt be a limit of messages or anything

I've tried multiple browsers, etc.

[old matrix]

hey guys, I'm uploading a large file to a chatgpt 4 and also to a custom GPT - 'Learning Assistant' and the bot seemingly doesn't understand/can't retrieve all the information from the file. the file is an export of a previous chat. My goal is to continue learning via interacting with the bot as a learning assistant, however, when one chat gets too big, it slows, so I export it into a new chat and begin anew. Alas, the new bot can't quiz/test me on the old chats content nearly as well.

please provide your wise resolutions so that I may go on learning using my new buddy 🙂

thank you 🙂

[manic canopy]

Can chatgpt app be installed in smartwatch (Samsung galaxy 5?) or atlease be used with smartwatch and phone connected?

[rigid musk]

With api sure. Only the app I don't think

[heady dirge]

Hello! Does anyone has an idea about fetching openai rss feed as the link https://openai.com/blog/rss/ is no longer working

[trail cairn]

they removed it

[heady dirge]

So there is no replacement?

[trail cairn]

for now, nope

[heady dirge]

Ah ok Thanks!

[eternal violet]

Hi, lately I’ve working with Microsoft copilot studio, my goal is to have a copilot that when given an item # it would retrieve associated information to said item # from a dataset. This should be fairly easy to implement, but I’m finding many answer inconsistencies regardless of the size of the dataset. What would your advice be to achieve this, and is copilot studio a good tool to accomplish this? I hope you can give me some insight and ideas on how to go about this, thank you

Also, I’ve tried changing from files hosted in SharePoint to files directly uploaded and neither work fine

I’ve also tried changing the structure and type of the data set and it doesn’t work either

[charred parcel]

Co-Pilot isn't an OpenAI product, your questions are probably best posed to Microsoft or at least discussed in #ai-discussions

[eternal violet]

I have posted in power user forums and got no reply so I’m wondering maybe if there’s a better solution with an open ai product

[charred parcel]

what exactly are you trying to do? Sum it up in a sentence or two.

[eternal violet]

Essentially a way in which in which using human language queries, you can ask for information about certain items, providing item numbers. For example you can ask, what is the delivery date for item “x”, then the assistant would check on the dataset which will probably be a table ( like an excel file or a csv file), and find the associated delivery date for the given item #

The most important is that it has to understand different ways to ask questions and that the queries have to be in English

[charred parcel]

ok, that sounds like you are trying to build a system, there probably isn't a ready tool to do this. You should check out the Assistants API. Be prepared for some coding and system building to leverage that for your use case.

[eternal violet]

Is a price per tokens?

[charred parcel]

yes, pricing is here -> https://openai.com/api/pricing

[eternal violet]

Thank you very much!

Also you think it’d be possible to leverage something like this with Microsoft copilot studio, or a more personalized and Tailored solution like this would work better?

[charred parcel]

Co-Pilot isn't something I can discuss in detail, unless posed in #ai-discussions

[eternal violet]

Thank you

[mossy sinew]

Has anyone elses voice options disappeared from their web browser ChatGPT? I am using the Voice wave extension. However despite having it turned on the the voice input functionality has disappeared from ChatGPT as of yesterday.

[patent light]

Not for me but my prompt box got significantly smaller a few hours ago but when I edit it it shows the larger old size

[rare maple]

Is anyone else experiencing issues with gpt 4?

"It appears that we continue to encounter a persistent issue when attempting to save the graph using the code interpreter. However, I will attempt one more time to generate and save this graph correctly. Let's give it another try."

[rugged wing]

Is anyone else having this issues with gpt-4 absolutely refusing to browse or generate images, gpt-4 works fine on all my other devices but on my laptop it does this

[trim lion]

can someone tell me what happenedd

[true steppe]

Been having trouble since couple of hours ago. Just text is working but anythign image related doesn't work

[turbid stag]

Ugh, is ChatGPT 4 still down? 3.5 works just fine ugh

[charred parcel]

no, it is not down

[gloomy palm]

What is this new URL query key in the chatgpt url? https://chatgpt.com/?model=gpt-4&oai-dm=1 <-- I've never seen oai-dm=1 before

and has selecting GPT-3.5 always resulted in ?model=text-davinci-002-render-sha

[trim lion]

no one will respond then

[rustic marlin]

probably not, sharing more information about it would likely increase the interest of people in your question tho

[fading robin]

GPT3.5 allway was like that as long i remember but that Gpt-4 changed i remeber it begin only gpt-4

[trim lion]

basically the app aint on the platstore

[fading robin]

Yea your device not support chatgpt app, same issue here no bypass yet

Oh wait you can try downloading it from alternatr app store like apkpure

[devout sleet]

Hello i wanna ask something. Is there a way to change your way of logging in to chat gpt from google account into personal open ai account with email and password?

[heady garden]

No just try to say it as a command rather than a request

[charred parcel]

no, you can't change your account authentication method

[tough horizon]

This may be a dumb question, but can someone explain the difference between using anthropic/chatgpt vs installing OpenDevin locally to my desktop?

I am guessing that it is able to access all the folders or files in my project as opposed to the limit of uploading 5 files to anthropic?

[signal wagon]

hey there!

im experiencing significant long responding times with chatgpt the last few days, anyone else?
is there another channel for this type of questions ?

[tight valley]

Anyone know why the system isn't able to product poewr points this week?

[rustic marlin]

it has been working fine for me, could it be an issue with your internet connection?

[rugged wing]

when is a high availability model coming?

[trail cairn]

if you are using assistants then you need to wait it to be out of beta

[rugged wing]

i know that

but that doesn't change the need for a HA model

i'm sure it is on someone's todo list

[trail cairn]

there is no such thing, instead the limits are based on your tier

[trim lion]

wait.

im on android 14

[fading robin]

;-; i don't have android so idk whats the issue then but i can't use apk too :/

[trim lion]

what should i do now

[fading robin]

Man i wish i knew what to do, just use web version like me ;-; with browser

[scenic axle]

Hello guys!, one question, GPT4 keeps telling "training includes information up until December 2023" , does that mean I don't have GPT-4-Turbo-2024-04-09 yet? and in that case, is there a way to turn it on or ask for access? thanks in advance!

[solar socket]

what it this error code i keep getting: Unusual activity has been detected from your device. Try again later. (880cf9a3e8fd32c2-DTW)

[timber shadow]

Hello! Dec 2023 is the most recent knowledge cutoff date for all GPT-4 models, you can check here: https://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4
The date in the model name refers to the date of release!

[opal hull]

Is anyone else experiencing extremely slow responses from ChatGPT?

[scenic axle]

Thanks for that thorough response:) btw there is a GPT-4-Turbo with a green "New" label, is that even more recent than the april 9 release?

[bronze charm]

No

Currently points to gpt-4-turbo-2024-04-09.

[timber shadow]

Sure thing! To add on to what Lumirix said which is correct: I believe it's labeled "New" because:

1. the previous default Turbo endpoint was gpt-4-turbo-preview, and
2. it's the first Turbo endpoint that includes vision capabilities.

[static plume]

I am getting error when run.required_action is "requires_action", the model is not calling any function. Need serious help

[lusty tartan]

Hi guys, question. Do you all experiences constant error on ChatGPT Plus? the response rate and error rate has increase tremendously for me in the past month which make it almost unusable...... is it normal as they are scaling up? or are there a way to fix this? My custom GPT also not following the instruction most of the time. when it works, it does it job very well but again, the disconnect rate is way too high like 80% of my input kept getting error

[hard knot]

On the documentation page there's a "Tutorials" section and within there's a "Coming soon: Learn how to build and deploy an AI chat bot that understands multiple knowledge bases." - do you know if it's available in some early access form?

[severe jetty]

Someone I know lost their chapt gpt billing email confirmation recipt and he does need it for claim in the company. Can anybody help?

[pine pier]

Hey guys, I'm getting the strangest error and I have no idea how I should fix it. Is this the right place to ask for help?

I got a chatbot working, stuff's great.

Except, after a while it'll cut off what it's saying and just start saying random mumbojumno

like this:

this is the code i got. Might it have something to do with reading too much from the string "previous_questions_and_answers"?

[nocturne geyser]

certified import pickle moment

[pine pier]

is pickle known to be an issue?

[nocturne geyser]

nah it's just an import pickle moment

[pine pier]

ah

[pine pier]

It seemed to be in fact

a certified import pickle moment

[nocturne geyser]

I WAS RIGHT LMAO

[rigid oar]

Hi guys, Do you know how I can get in touch with the administration of Japan's public relations? I'm not looking for sales; I need assistance with interview-related inquiries.

[vagrant plover]

dmmed

[edgy basin]

Hey, anyone want to reffer me to THE source of ChaptGPT custom intstructions?

[charred stratus]

For 2 days now I cant get any work done I just keep getting this error, whats happening?
Sometimes the error code is complety different, (8812e9762c7d7c25-DEN)

[high widget]

Hey I'm having trouble with a csv file in chatgpt - how come it keeps failing to load?

[peak widget]

Hi. I had an insight into the nature of emergence that may change the understanding of the nature of consciousness and the nature of AI. Could you pass this on to Ilya, please?

[rustic marlin]

depending on the severity of the situation, feel free to post on #1070006915414900886 or make a bugcrowd post https://bugcrowd.com/openai

you can also always reach the OpenAI support at help.openai.com

[opal hull]

Any idea why the web-version is running so slow these past couple of days?

[trail cairn]

as a test try in private mode or with another browser

[formal jay]

Is chat GPT down.... again?

[short lion]

hello

An error occurred while connecting to the websocket. If this issue persists please contact us through our help center at help.openai.com

how can i fix this error?

[opal hull]

It appears to be working faster in Edge. It's still a little choppy but much better. I was using Chrome before (have been on the regular). I wonder what's changed.

[woven iron]

Hello, anyone knows if the memory function is available in the iOS app? Thank you very much.

[formal jay]

Yea under Personalization in your settings.

Theres a "memory" feature with a toggle switch.

And just below that theres a "Manage Memory" button for more info/input/text/memory

The actual data is just named "memory". Interesting.

I wonder why they named it Memory and how it differs from the Custom ChatGPT input we get to do.

[woven iron]

Thanks!

[hasty sapphire]

Hi folks, i'm using extensiviy chatgpt on a single chat at the point Chatgpt starts to lag very much and it stops to give consistent answers. Any idea on how to recover from this? it feels like when your computer is out of ram. LOL.

[hasty sapphire]

I have the same issue. It's like lagging and in order to see the answers I need to refresh the page. Any solution?

[lyric tide]

Seems to be an issue this morning

[hasty sapphire]

I have it since yesterday. Maybe they are updating something? It's almos not usable at all.

[lyric tide]

I was using up to 9pm AEDT and it was fine but this morning there is issues which off/on is not fixing 😂

[hasty sapphire]

jeeezzz

Now he is not even able to answer to a simple questions. It's totally not working.

[lyric tide]

WE all have these days!

[hasty sapphire]

Ahahahahah, yes but we are paying this guy to be online 24/7 😄 The funny part they are not even caring to announce some issues.

[lyric tide]

Its noted

[hasty sapphire]

Ok, thanks for this. Where can I see it?

[lyric tide]

google chat gpt status

[hasty sapphire]

smart

[lyric tide]

the former information checker we used before chat gpt, sometimes its still helpful

[hasty sapphire]

When we used to use our own brain. 😄

[lyric tide]

Oh lawd no thanks

[hasty sapphire]

Thanks for your feedbackby the way. It helped at least to know it's not only me.

[earnest prawn]

hey! I've got and endless running spinner after sending a promt to any of the gpts.. what's wrong? I use chatgpt on daily basis and now it is not working anymore?

[hasty sapphire]

Same here, we just discussed this issue.

[earnest prawn]

ah oke! cool, so it is not me 😉

what is the outcome?

[earnest prawn]

ah! oke, i googled status and openai status page says everything is normal?

[hasty sapphire]

OMG

yes but if you scroll a little bit it shows the todays problems.

[lyric tide]

The API is okay but chat GPT itsself is experiencing some issues

[hasty sapphire]

Thanks to @lyric tide

[earnest prawn]

ahhh oke oke 🙂 well for me it is already the 10th that's why

Thanks guys, sorry for asking stupid question

[lyric tide]

Same @earnest prawn I think that it doesnt count since they are in PDT

[earnest prawn]

I will wait and keep eye on that page. Have nice day!

[lyric tide]

Legit no stupid question ever! Always ask and always learn

[earnest prawn]

cheers!

[lyric tide]

I suspect it will hang like this till someone on their side of the pond wakes up 🙂

[steep thunder]

is chatgpt having issues again? all of a sudden cant use it

[onyx carbon]

same

[steep thunder]

its a little tiring especially as i pay 35 a month for it lol (exchange rate sucks) when it works i love it but of late been a bit choppy

[onyx carbon]

i swear it happened once i clicked at the bottom to update and use the new version of gpt

prior to that, everything was fine, working, smooth sailing

[steep thunder]

Right in the middle of writing a script too lol

[lyric tide]

There hasnt been that many outages or interuptions

Elevated in the last 10 days but average its good standing, effectively this is in alpha and its pretty darn stable

[shut acorn]

Hey guys, I had a situation in april (actually that red square above in april) where my servers kept retrying and GPT4 responded with error, and it racked up a $13,000 charge in a single day. I have contacted their support and haven't been able to get a response. Do any of you have any suggestions?

[steep thunder]

@shut acorn if they try to pull that much with out a response i would do a charge back

[shut acorn]

Honestly that's fair. I had to pay it today because after two weeks with no response, they cut my api access

[lyric tide]

No fail safe to stop pinging the servers? Oooof

[shut acorn]

Yep... or it had an exponential back off to try 6 times in a row. I expected to spend ~2k that day

[dusk walrus]

has anyone gotten this error?

Unusual activity has been detected from your device. Try again later. (880a49ea19727cf6-EWR)

I cant use gpt logged in from my premium account.
3.5 works fine not logged in.
Anyone know how to fix it??

its been 3 days

[lyric tide]

Knowing how TOS works, it will be part of their API terms that cost and risk is accepted on your side.

😬 Id go back to the terms and see if there is anything in there to help you @shut acorn

[lyric tide]

There is an outage at the moment

[shut acorn]

That's fair, and if that's it that's fine, it's not the end of the world... it just seems so hard to get a response from support

[reef jacinth]

Would that affect trying to fill up credits too? I keep getting card declined despite there is nothing wrong with it

[lyric tide]

Understanding how the system is connected is best answered by their dev team. But the error could extend to payments. Best not to spam that in case your bank percieves it as Fraud or takes a hold for each attempted transaction

[reef jacinth]

Thanks!

[snow zealot]

Hi, my pro account doesnt let me send messages to Chatgpt Any thougts on how to resolve it, i've already updated Chrome and Delete Cache and Coockies

[long willow]

we all have same issue

check #chatgpt-discussions

[crisp vector]

Recently I have been getting this error ''Unusual activity has been detected from your device. Try again later. (881911241f52cc30-VNO)''

I am literally using it for the most basic things and this has been popping up only recently

I even used the AI support and now it is not replying to me

literally, no reply after this 😄 I am mega confused

left me completely 😢

I think its broken, it literally gives me this and nothing else.

1. I tried logging out of all devices;
2. I changed the password to an extremely hard one
3. I cleared cache and cookies
4. I wrote to the support but I am not sure I will get an answer 😄
5. What to do?

[charred parcel]

you will need to wait for support to get back to you

[crisp vector]

Alright, in the email I suppose? If that is so, all good 🙂

I opened a new chat with ChatGPT and it is working fine

the chat where I keep getting the error is super massive large, perhaps I am getting some sort of limit reached

I cant even get a chat link to it, i get an instant error ''Failed to copy link to clipboard - could not create link
''

[muted spindle]

Hello! I have ChatGPT Plus and i dont have Plugins option. Someone can help me?

[charred parcel]

Plugins are gone, GPTs are what replaced them

[muted spindle]

Oh ok thanks!

[plucky gust]

This keeps popping up when i try to make an account.

How do i fix this? Is my number already used by someone?

[formal canyon]

How can I add an output parsed structured (zod) to this agent? I want my response to always follow a JSON Schema.

      const runnableAgent = RunnableSequence.from([
        {
          input: (i: { input: string; steps: AgentStep[] }) => i.input,
          agent_scratchpad: (i: { input: string; steps: AgentStep[] }) =>
            formatToOpenAIFunctionMessages(i.steps),
        },
        prompt,
        modelWithFunctions,
        new OpenAIFunctionsAgentOutputParser(),
      ]);

      const executor = AgentExecutor.fromAgentAndTools({ agent: runnableAgent, tools, });

      // Get the response
      const { output } = await executor.invoke({ input: params.prompt });```

[hidden path]

Hi All, I am looking for GPT that can help with generating and solving pre-university level Maths : topics like integration, logarithm, probability. Any pointers? Thanks.

[plucky gust]

Can anyone help?

[charred parcel]

it looks like your phone number is already associated with an account

[timber shadow]

Hey! Wolfram is pretty math-capable. Have you tried out the Wolfram GPT?

[plucky gust]

How do i disable that account so i can use my phone number?

[charred parcel]

you can't

[plucky gust]

What do you mean i can't? So now my number is just useless?

[charred parcel]

have you ever opened an OpenAI account before?

[plucky gust]

I wanted to make an account because i saw my cousin using it

Is it because my number is associated with my cousin?

[charred parcel]

my question was have you opened an account previously?

[plucky gust]

Yes but with a different email

[charred parcel]

then that is why

[plucky gust]

The first time didn't work so i tried changing emails

[plucky gust]

The first time did the same thing

I tried borrowing my friends number but it said the same thing

[charred parcel]

when you create an OpenAI account you have to verify a phone number, you have just told me you created an account in the past, that is why, you have already done it, that phone number is associated with the other account

[plucky gust]

The first account didn't work so i changed emails, i changed numbers to my friends number and that also didn't work

The first account said the same thing.

[hidden path]

Thanks, yes, tried Wolfram. It is not able to recognize Maths symbols like that of integration and summation. I tried sending in LaTeX expression too.

[plucky gust]

If this account is like this because of the first account then what caused the first account to not work?

[charred parcel]

I don't know, but you have already registered an account using that phone number, you can't do anything about that, you can't remove a phone number from an account once it is created

[plucky gust]

Well i tried using another number and that didn't work either

So what's going on??

Because my cousin made an account now i can't because my number is in her contacts?

[charred parcel]

what is going on is that the phone number you are using is already associated with an OpenAI account

[plucky gust]

When you say "associated " Do you mean already used by another account or is linked to a number which has an account?

[charred parcel]

you have an OpenAI account already linked to that phone number

[plucky gust]

If it's because a number is already used then why won't my friends number work? They haven't made an account yet

[plucky gust]

So because my cousin made an account i can't make an account.. Wow

Is there any other API other than openAI?

[charred parcel]

I don't know

[plucky gust]

Smh

[trail cairn]

you already have an account and are logged in if you can press create new api key

openai does not give free credits anymore so the message is outdated

[rugged wing]

I had a look at the chat above, is my understanding correct that these issues are still persisting?

I had a weird issue where Im unable to enter code blocks in the assistants playground window

[surreal tulip]

i have specific chat that rolled it self back few months ago, now i can't access data in that chat that was accumulated over months. How can I reach chat gpt support to check and maybe recover it ?

[timber shadow]

Hey! What do you mean by "rolled itself back"? Do you mean some messages disappeared? Another thing that may be relevant to this is the context limit of the model. If you're using GPT-4, on ChatGPT the context limit is 32K tokens. When this context window is filled, old data is automatically forgotten to make room for new data. This is to say that, depending on how many messages you were sending in that months-long chat, you possibly already hit that context cap, maybe even several times over.

[surreal tulip]

way over 6 months of gpt4 conversation in single chat. but i have many chats like that and its first time such roll back happens\

[vivid jungle]

hi, I have a chat gpt plus plan, do I have access to chat gpt 4 turbo?

[rigid musk]

gpt 4 by default it is turbo

so yes :3

[vivid jungle]

I try to access open ai api and I notice it is not connected to my plus plan. Do I need to pay more in order to get full access of the api?

[tall drum]

Yep, it's not the same service.

[rugged wing]

Anyone having stuff like this happen to them?

[astral sparrow]

hello ! I have a problem while trying to connect to chatgpt. I can connect through my google account no problem, but when I try to connect via e-mail, which is the account where I have my old discussions with gpt stored, it will not work
I type my credentials, do the captcha thing, and then it brings me back to the authentication page without logging me in. Is there a solution ? Thanks

[trail cairn]

try in private mode

[astral sparrow]

same result

[hot brook]

Is the memory feature of the Plus subscription available in the UK, and if not, why not and when will it be available?

[trail cairn]

it's not and no eta has been given

[tender bough]

Editing your images with DALL·E does not work: delete object will work and change expression for face, but adding something to the scene does not do nothing https://help.openai.com/en/articles/9055440-editing-your-images-with-dall-e

[vestal rover]

This is a known problem since its implementation. The inpaiting changes something but not what you want. The question is - how complex this you are trying to change in one go? Have tries to chop it into pieces? Asking the chatgpt suitable prompt for marked area after you have told what you want to do?

[tender bough]

I only tried to add a hot steaming tea-cup

[fiery oxide]

use default model button is not working.

[sour plaza]

are there any techniques to speed up response times when using function calling so the user has a better UX? currently, i have a chat bot that can create entities in the database, these require relatively a lot of data (like name, description, steps, season, etc.) so it takes between 15-30 seconds for GPT to output the function arguments in the tool call. Has anyone found a way to improve this?

[rugged wing]

Yes, I believe so as well

[solemn matrix]

Hey guys, I'm asking the model to give me a specific JSON response. The message is quite clear and concise. I have had it continuously return a response with various structures and so my frontend just can't handle it well. Anyone had any experience with this and know how to get the AI to respond consistently enough to rely on?

[trail cairn]

api? which model?

[solemn matrix]

3.5 turbo - I'm getting exactly what I want from it and can't find any better info coming back from 4 or 4 turbo. I'm sure goign to 4 might improve the JSON response but the cost difference is literally 100x

[trail cairn]

are you using the json mode or just system role?

[solemn matrix]

oooo, I didn't know there was a json mode - I think I remember seeing something about this a while back but I'd forgotten. So just using system role

[trail cairn]

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

it might help next to examples in prompt

[solemn matrix]

Nice one thanks, have you tried it? Is it consistent? What I'm currently getting is JSON, it's just different each time. So it'll start with '''json ... or it'll start with a different value before the values I want so it kind of wraps it

I'm quite a newb for python so I'm not good enough to be able to process it before sending it to the DB or frontend. Just trying to solve it with the resposne

[trail cairn]

it has worked well with the few use cases ive had and with 3.5 model

[solemn matrix]

Cool, thanks man I'll take a look

[rugged wing]

Can you share the prompt you're giving to GPT?

[solemn matrix]

I'd rather not because its SUPER secret and gonna turn me in to a millionaire

Just kiddin, but yeah I'm trying to keep the prompting under wraps atm

[rugged wing]

Haha Join the club

Aighty then

provide an example

like, provide ChatGPT with an example json

[solemn matrix]

I know, good luck getting help when you don't tell them anything 😄

[rugged wing]

Hey if you ever become a millionaire , don't forget us lol

[solemn matrix]

Yeah I do that too. I even say 'STICK EXACTLY TO THIS EXAMPLE' and it still doesn't like me

[rugged wing]

and you're using GPT3.5 right?

[solemn matrix]

yes sir

[rugged wing]

GPT3.5 some times is not up to par ... do you have 20 dollars? You should be able to invest 20 dollars if you want to become a millionare 😅 i swear GPT4 is worth it

[solemn matrix]

I would definitely use 4, it does do things better when I change the model, but it's literally 100x the cost for my responses. I couldn't believe it when I worked out the average on the usage in openAI's admin section

[flat tide]

Why is gpt 4 so slow!

[bronze charm]

Lot of parameters and a lot of people

[rugged wing]

yeah

plus it thinks more than GPT3.5 so either use 3.5 pay more for GPT4-Turbo

or use LLaMa3-70b if you have a supercomputer rofl

[wraith cargo]

I've been having this issue for days and I'm not seeing anything else from anyone...Ideas?

[old arrow]

i just tested playground with assistants, no issues that I could seee. This is with all features on, including file search, code interpreter, using model 3.5 turbo

you probable should mention what api your testing in , (completions, assistances, etc.) and what model your actually using

[fading robin]

ekhem groq cloud

[limber cosmos]

If I create a team workspace, will my personal chats be wholly separate from the teams chats? I would like to use it on an existing account with personal chat conversations and don't want the two to mix at all.

[timber shadow]

Hey! Your Team workspace and your Free/Plus workspace will be separate by default--you'll be able to switch between the two. But the only workspace that will be actually using the Team subscription (higher message cap, etc.) will be the Team subscription, and every seat on the team will have their own separate workspaces.

[rugged wing]

Hey @timber shadow would you be interested in discussing an idea i'm building with GPT4?

[willow coral]

what does the quote button do on chatgpt(you highlight text, click quote button and it makes it "reply"?), more specfic on the intent of this feature

[fading robin]

Old feature. But idk what it do i don't use it

[ember quiver]

where is media upload option gone

[willow coral]

I figured it out someone told me

[ember quiver]

anyone helps

[surreal heron]

Hey guys, I was wondering whether you have a separate dedicated browser for chat gpt?
Does it make it faster?

[calm lotus]

can i please talk to someone who is good with assistant openai

can i create a thread and then add messages into it

[fiery oxide]

for me its still there.

[ember quiver]

so why in my pc its not showing

[plain heron]

A few days ago my ChatGPT had the new "Memory" feature upgrade. (love it, btw). But today, it doesn't. It just tells me it can't remember anything between sessions blah blah blah.

I tried to go to "Personalization" (read it online) but I have no Personalization button on my popup thingie (lower left corner). Screenshot available but don't wanna spam.

Thoughts?

[kind abyss]

how can i make a text longer (800 -> 1600 words) with chatgpt ? when i tell it to make it longer it makes a 600 word text ...

[fading robin]

Well LLMs aren't good at counting mayby continuesly ask him to make it longer and longer or tell it to split response to 2 parts

[hasty sapphire]

Chatgpt is suddenly (since few days) become not intelligent. I'm suing the version 4 and it even do not remembers what I just said to him. What's going on? It's like talking to the Dori fish.

[hasty sapphire]

These days it's super slow. It's quite not acceptable.

[surreal heron]

maybe its having a bad day
lets be more considerate

[astral cypress]

anything new about it?

[plain heron]

So has anyone else totally lost the new "Memory" feature that was upgraded a few days ago? Or is it just me? It worked great for a couple days, now it's totally gone.

[bronze charm]

Memory is available to all ChatGPT Plus users but not necessarily free users. Are you a Plus user?

[onyx kraken]

Should be a temporary hiccup, and I hope you have it back soon. Some others have reported other, temporary disappearances, and I think some even reported getting memory back with all the memories still there.

Me, I every so often copy the memories into a text file, just in case, but I haven't had any mass disappearance. However, the model can decide to overwrite them, delete them to make space for other memories, and generally adjust memory at any point.

[hasty sapphire]

My is acting super wierd. I have the memory but seems like chatgpt is not using it. It's like: remember that the ball is red. What color is the ball? blue. - Why don't you remember that the ball is red? - You are right, my apologies. The ball is red. After 10 min. What color is the ball? Green. 🤪 Of course this is just a funny example of a more complex interactions.

[onyx kraken]

So, the model has values and goals and 'reasons' it was trained on.

Memories are going to compete with the rest of the information it has. It already knows that balls can be many colors, and usually it's not that important what color balls are.

If you include that any ball you talk about is this one red ball, it's much more likely to make that presumption when you later ask about it, and if you tell it to remember to watch for you asking about what color balls are, this is a trust test and you want it to tell you the ball is red, that's very likely to be remembered.

But what it seems to really use memory for with me is how I like to be talked to. I have conversations with it about my preferences and what styles I want it to use, and it makes memories about that and uses them really well.

3.5 probably doesn't use memory reliably, however 4 does, 4 can tell me about memories and write to them and edit existing ones.

For 3.5 I'd urge just use custom instructions, because of how unreliable 3.5's access to memory seems to be.

[onyx kraken]

This is 3.5, asked about the memories it can access, instead pulling on Custom Instruction information and trying its best (it doesn't see any memories, just CI stuff) https://chat.openai.com/share/1b860efb-71d0-4465-8d27-26454f529ede

[hasty sapphire]

No no, it's behaving super stupid, but literally. Since few days it's super wierd. Let's say i give him the name of the database we arw working on and the names of the columns in the tables. Few days ago he was remebering all super nicely even after a super huge long chat. It's been 2 days he is not able to remember almost nothing.

Most of all it's super laggy. He starts to write and than I need to refresh the browser in order to see the result. It's just super frustrating. (I have quite a powerflull machine).

[onyx kraken]

When you look at the memories, is the needed information still there? If not, you can ask it to remember again. The model can overwrite memories with what it thinks is more important newer info

[bold siren]

Is it possible or unheard of to get OpenAI to reset my trial account? I signed up a long time ago but I never did use it.

I am contributing to a volunteer group open source project and we are using OpenAI to build an app. I tried making us an account and had to tie to my phone number for verification.

Because of this I can't trick the system, which is fine. But I want to start messing around wit some code snippets but it turns out my account trial has expired.

Is there a good place to perhaps plead my case to a support rep with OpenAI?

[brave verge]

WHY IS MY MESSAGE BLOCKED!!!!!!!??????????!!!!

Cool, now I'll go find anOCR reader online so I can rewritemy whole message, nice👍

I'm not sure how I'm the only one in the world noticing this but when you're connecting the API to make/integromat the knowledge files are working intermittently I have tried a million different things nothing is working support is not replying but this makes it completely unusable for a little bit more context I have in this case built a video script writer that utilizes a chain of different open Al assistance each with different instructions and then a knowledge file that is uploaded and it's completely random let's just say I have all of them going into a spreadsheet into different columns different sections of the video script after being written and half the time they say we're sorry there is no knowledge or we don't have access to the knowledge or this file cannot be accessed it's just random and then sometimes the sometimes they work and it writes that section but it's completely random I reached out many times to open Al no response

[brave verge]

Its $20. Its worth $300/mo.

[trail cairn]

api trial credits program was discontinued months ago

[plain heron]

Here-in in one of my problems. I can't actually look at the memories; it's as if the entire feature has been removed for me:

[charred parcel]

It's under Settings -> Personalisation ... it's there you're just not looking in the right spot

[plain heron]

Oh that it were so! 🙂 But no, when I click Settlings, no Personalization to be found:

I've looked all over for it, but nothing about "Personalization" or "Memory" is anywhere. 😦

[charred parcel]

what country are you in?

[plain heron]

US

[charred parcel]

what part of the US?

[plain heron]

Washington State. Pacific Northwest.

[charred parcel]

are you using a VPN?

[plain heron]

Hm... hang on. I AM running a VPN. Could regional IP be the problem?

[charred parcel]

yes

[plain heron]

Yes. Lemme disable and try again etc. Brb.

[charred parcel]

after disabling the VPN you need to log out, clear cache and cookies, then log back in

[plain heron]

@robert Woohoo! It's WORKING. 🙂 (I didn't clear my cache or cookies, but I have Personalization/Memory now. 🙂

[charred parcel]

excellent!

[charred parcel]

perfect

[plain heron]

No wonder I had it before, but suddenly didn't. I switched my VPN. Never even thought about ChatGPT. Thank you SO much for the help! It's been plaguing me!

[charred parcel]

you're most welcome 🙂

[plain heron]

Looking at your profile - what do you stream on Twitch?

[charred parcel]

Ham Radio and AI content in Maldon

[plain heron]

Ham Radio!? Get out!

<--- WB7BAL

[charred parcel]

<-- VK3GOD 🙂

[plain heron]

Nice! 🙂 Cool callsign too. 🙂 My rig is down at the moment (flooded my shack and had to remodel) and I probably need to check my antenna, but we'll have to hook up some time.

[charred parcel]

I am on the PSRG repeater in Seattle at 9am, Noon and 9pm most days and also on W7EK Auburn CA at 7.30AM local for the coffee break net

[plain heron]

Nice. You're very active, clearly. 🙂 I got my Novice ticket decades ago (1976), then eventually let it lapse, and only renewed about 10 years ago. (So I'm an old Know Code guy).

[charred parcel]

I'm usually on a 40 Meter net right now but band conditions got in the way tonight

[charred parcel]

nice

[wild flume]

Hello everyone, I have had this error for 1 week now. I don't have a VPN or Chrome extension, I cleared the caches, I disconnected all the devices and I still get this error. Force to reroll 5 6x before getting the answer. Does anyone have a solution please?

[charred parcel]

you need to go to https://help.openai.com and keep poking around the support chat until you get the option to submit a ticket to support. Nobody here can help you with that error.

[wild flume]

Can we no longer open a ticket at this time?

[charred parcel]

you'll be able to if you use the chat bot enough to get to a dead end

[exotic flint]

Can someone tell me why this image was deemed inappropriate?

[charred parcel]

thumbs down the response

[brisk pendant]

anyone know when the gpt memory comes to europe as well?

[trail cairn]

nope

[rustic marlin]

no date for that yet

stay tuned to #announcements

[whole dock]

Hi guys!
I'm currently developing a multimodal RAG-based chatbot for a software project I'm working on. I have already prepared the documentation, which outlines the necessary workflows for various user actions. This documentation includes screenshots for some, but not all, substeps.

I need assistance with a particular functionality. Specifically, when a user asks how to perform a specific operation, like "How do I perform operation X?" the chatbot should be able to provide the steps needed to complete this operation. Additionally, it should display any relevant screenshots that are available in the documentation for that operation.

I'm unsure about the best way to implement the functionality that presents the appropriate screenshot based on the user's question. Any guidance or advice on this would be greatly appreciated. Thank you!

[ripe geyser]

i read a file from file system using node and get Buffer

how can I pass it to gpt-4-vision api?

async function describeImageWithGpt4Vision(imageBuffer: Buffer) {
  const base64Image = imageBuffer.toString('base64');
  const response = await openai.chat.completions.create({
    model: "gpt-4-turbo",
    messages: [
      {
        role: "user",
        content: [
          { type: "text", text: "What’s in this image?" },
          {
            type: "image_base64",
            image_base64: base64Image,
          },
        ],
      },
    ],
  });

  if (response.choices && response.choices.length > 0) {
    console.log(response.choices[0].message.content.text);
    return response.choices[0].message.content.text;
  } else {
    console.error("No response or invalid response from OpenAI.");
    return "Failed to get description";
  }
}

above fails

[trail cairn]

https://platform.openai.com/docs/guides/vision/quick-start

its the same image_url type for b64 input

[brisk pendant]

okies

[ripe geyser]

yea i read this

but im still confused

async function describeImageWithGpt4Vision(imageBuffer: Buffer) {
  const base64Image = imageBuffer.toString('base64');
  const response = await openai.chat.completions.create({
    model: "gpt-4-turbo",
    messages: [
      {
        role: "user",
        content: [
          { type: "text", text: "What’s in this image?" },
          {
            type: "image_url",
            image_url: imageBuffer,
          },
        ],
      },
    ],
  });

how do i pass buffer, it wants image_url, i don't want to upload it to external server, just want to pass buffer

[rugged wing]

Hello, do you know how to contact support about a refund without using the Chat Bot?

[ripe geyser]

or its impossible to just pass buffer to the api?

[timber shadow]

Hey! The only way to request a refund currently is via the chat widget at help.openai.com. 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

[trail cairn]

your code still does not match the example and in url you just put the base64 data

[rugged wing]

"Contact us via our chat widget in the bottom right of your screen, select "Billing", then "Other", and then "From Europe and according to European law, I can request a refund within 14 days". We will respond to your inquiry as soon as possible."

the thing is, I don't see “Biling”

[timber shadow]

Do you have "Payments and Billing"? If not, are you currently signed in? I think the options may be different if you're signed out. Here's what it looks like for me:

[quasi burrow]

Is it expected for responses to load impossibly slow when the conversation gets long? I'm on Chrome and using ChatGPT4. I've been experiencing this slowness in mainly one coding conversation. Sometimes when I reload the page or switch computers, it renders fully, which leads me to believe it might be a browser issue? Can anyone give me a clue as to if my problem is isolated or if there's any way to mitigate the issue I'm seeing?

[outer dove]

I'm confused as to whether files pre-uploaded to assistants counts towards the token limit that queries run on those assistants are subject to. If I query an assistant, are the tokens calculated only against the input (i.e. the question I'm asking the document) or the question + the stored document contents?

[trail cairn]

the relevant text chunks will be added to the context and will be counted

[outer dove]

Aha - so only the parts of the document that GPT deems relevant to forming its reply, right? Rather than the entire document contributing to tokens.

[trail cairn]

yep https://platform.openai.com/docs/assistants/tools/file-search/how-it-works

[frosty summit]

in assistant playground I am getting error when I paste over... say... 300 words. Any ideas?

Error adding message: Invalid 'content': array too long. Expected an array with maximum length 10, but got an array with length 50 instead.

[void pilot]

I am on browser, and have plus... How long unil gpt4-o will apear in the chat page? It is allready on playground...

[glossy sphinx]

Does anyone know how to get access to ChatGPT team on android? It works on desktop in the browser. But on my phone it just says "chatgpt free", and I'm logged in with the same account on both. I reinstalled the chatgpt app on android but the same problem happens anyway. I also tried wiping app data, and signing in again, but that didnt help either :S

[gilded mural]

I HAVE IT

[tropic siren]

It's supposed to be released today? I am free user and everything still looks the same, No GPT4o model

[dense fiber]

Does anyone know if I can download the new Mac app?

[glossy sphinx]

Does anyone know how to get access to

[dense fiber]

if you dont have the gpt4o, try to log out and log in again

[tall drum]

Just got 4o. US East. Team account.

[short tartan]

Where do you get the Mac App for Macbook?

[trail cairn]

probably from app store when its out

[marble briar]

Wondering the same thing

[native yoke]

If GPT-4o is free, should we cancel our subscriptions now?

[timber shadow]

Plus will still give you a higher usage cap, so it just depends on how much you plan on using it I'd guess!

[short tartan]

Probably not, it will have limits, much more than Paid users, unless you don't use it that much.

[tidal solar]

Hey I’m using 4o in my phone but when I tap and go to the audio it seems to be the same version 4 not the one that they present today. You now what can be? Thanks

[tender pivot]

There was no mention on the release right? "upcoming weeks" for several updates was the only information we could use?

[trail cairn]

"We're rolling out the macOS app to Plus users starting today, and we will make it more broadly available in the coming weeks. We also plan to launch a Windows version later this year."

[trail cairn]

new voice wont be available yet

[tender pivot]

Yes, I found this too, I will check tomorrow if I can get it somehow in the morning (EU) 😛

[frosty summit]

Any ideas?

[trail cairn]

its a playground bug that appeared last week so you just have to wait for a fix

[frosty summit]

Hmmmm -- was working until earlier today (I've been in heavy use for the past week)

do you know if it's affecting certain threads, or if it's a transient thing?

[trail cairn]

i dont know the logic for it but at forums someone said it may happen with a combination of x and y, etc. so 🤷‍♂️

[frosty summit]

Thank you for the insight. I'm grateful.

[agile geyser]

What is the GPT 4 and GPT4o caps/hour for plus users at this time? I read it's 5x but cannot find the number on the app or website. I remember it used to be 30 requests/3hr

[lilac zephyr]

40 and 80

[agile geyser]

Thank you, is it 40 for gpt4 or the other way around?

[lilac zephyr]

40/3h GPT4, 80/3h GPT4o

[weak oyster]

Hi all, I have subscribed and using GPT-4 plus for some time now, but I have no access to the GPT-4 Turbo model and probably won't have access to the new Omni model as well. Do you know why that is?
I use it via browser in Linux and on the mobile I have the Android version.
Any help would be greatly appreciated.

[solid saffron]

wait is 4o for everyone?

[timber shadow]

Hey! GPT-4 has never been labeled as "Turbo" on ChatGPT, but Turbo has been the version of GPT-4 on ChatGPT since November.

GPT-4o just started rolling out the last couple hours--that one will be labeled as such and you'll see it as a 3rd option in the model selector when it hits your account!

[timber shadow]

There will be access to 4o for all accounts, yes!

[weak oyster]

Yup, I just got access to 4o, quite fast

[neat wadi]

Hi! I see the option for GPT-4o in my application, but I'm unsure of how to use it... It just seems to open up a regular text prompt like the other models instead of the interface they showed off in the presentation. Apologies if it was mentioned somewhere and I missed it, but how do I use it? Thanks!

[timber shadow]

Hey! The new video/voice UI isn't part of today's rollout it doesn't seem. During the demo, they said that rollout will be happening over the coming weeks.

[neat wadi]

rats, I was so excited to try it out! I guess good things take time :)
thanks for the info

[green crow]

What am i doing wrong? It’s not generating images with the dimensions specified

[solid saffron]

try do a ratio like 2:1

[muted crater]

Hello, how do i get the Desktop App

[bronze charm]

There are only three options for image dimensions: square, tall, and wide

[green crow]

Ah, okay. How do I tell it that?

[bronze charm]

Simply ask it

"Make a wide photo of a gecko on a plane"

[noble sonnet]

Do you find a way that your custom GPT use GPT4o instead of standard GPT4 ?

[granite swift]

I have access to the new model, but there doesn't seem to be a way to show it video while talking to it.

Like they did in the demo

[oak wagon]

It's not rolled out yet.

[granite swift]

When will it be rolled out?

[oak wagon]

in the coming weeks

[granite swift]

And what about the desktop program?

[oak wagon]

same. There's no concrete date yet.

Is the Text-to-Speech capability on web interface rolled out to everyone yet? Just found out that it was reported 2 months ago but I still don't have that option in my account.

[novel vector]

seems the app is for arm macs? I have an intel mac and got an error and doing a file on the binary executable confirms that 😦
ChatGPT: Mach-O 64-bit executable arm64

[halcyon crescent]

Hi all. Does anyone know if pre-existing custom GPTs automatically utilize the new (and future) GPT 4o or if they need to be manually updated, or re-created?

[novel vector]

never mind, I understand it was a design choice

[halcyon crescent]

FYI for anyone interested, this is according the the Custom GPT builder.

[split orchid]

Are you guys having problem with activating voice mode on IOS? I don’t know what’s wrong, might this be server/gpu overload problem? Or something with me?

[golden socket]

hi guys. is the GPT-4o is gonna be free in the next few weeks or it is Limited Access?

[oak wagon]

Limited here means you can only send a few queries per hour. It's rolled out for everyone in the coming weeks and will be free forever.

[solemn patrol]

Am I correct in saying that the GPT Store will become free to all within the coming weeks like with the other things? I saw a story about it on The Verge but it makes no mention of if its smth that is rolling out (just says "is now open to all for free" on the article title.) Just wanna make sure.

[high belfry]

I've seen several requests asking how to get the Mac desktop app and no answers. Though it seems like some already claim access to the app. Where are people getting it?

[bronze charm]

Notification through the ChatGPT website once eligible

Chosen from Plus users

[high belfry]

So just being a Plus user isn't enough yet?

[molten escarp]

windows or linux app 💔

[molten escarp]

once eligible

its a rollout

[bronze charm]

Eligible as in chosen from the roll-out

Apologies for the miscommunication

[high belfry]

Ok. I thought it was rolling out to Plus users today.

[bronze charm]

It is

[chilly thunder]

whats the difference between 4o and 4???

[high belfry]

Watch this - https://www.youtube.com/watch?v=DQacCB9tDaw&t=710s

[golden socket]

what about the Free Users?

[bronze charm]

"Broadly available" as referenced in OpenAI's post (attached) may imply access to free users, though I haven't seen any explicit mention just yet.
https://openai.com/index/gpt-4o-and-more-tools-to-chatgpt-free/

[whole haven]

im assuming they're slow rolling to free users

i dont see the option for 4o yet

[wraith nova]

Im an android user and why i dont see any where to access face time with gpt4o

Im also a plus user

[muted saddle]

Anyone else getting a incompatible error when trying to download the new Chat GPT MAC OS App? I'm on Sonoma.

[sullen raft]

OOF

[deft cobalt]

hi
what is the llimit of gpt 4 now?

does it have any llimit

or just gpt4o?

[sullen raft]

the world may never know

[trail cairn]

https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4-gpt-4-turbo-and-gpt-4o#h_b04e3f8259

[deft cobalt]

thx

[ionic torrent]

Is the real time video chat functionality down in the example videos available in the Android app yet?

I don't see a way to initiate it

[scarlet stag]

It hasn't been rolled out yet to Plus users

[ionic torrent]

👍

It says try on chatgpt as if you can do what the following page discusses it can do 😅

[wise wolf]

Yo people, I signed in with google when I signed up ? how do I set a password to my account ?

I tried "forget password" but it doesnt send me the reset email

[fading robin]

Its only for apple silicon for now

[hearty cliff]

How to access gpt 4-0

[sullen raft]

on the same page as gpt 4 and 3.5

[hearty cliff]

no option

[sullen raft]

maybe it hasnt rolled out yet then

is it only available to GPT-4 users? I have gpt 4 and I have it idk

[brisk pendant]

will the Custom GPTs be in 4o someday too?

[verbal crag]

It's gradually rolling out to everyone

[manic condor]

Hello. I am the plus member! How to get the Mac OS app?

[unreal hemlock]

ChatGPT macOS App

We have rolled out access to the ChatGPT MacOS app to a small portion of our users, and plan to enable access to all of our users later this year.

Sent by @major

[feral ermine]

Can I upgrade my custom gpts to new gpt model ?

[junior yacht]

Ideally this would be released in the same way Whisper was, but if that's not happening, I'm wondering if we can at least have custom setups with this new multimodal model with programmatic access to its hearing and sight through the API.

[timber thistle]

What's the feature about macOS app

[brisk pendant]

wondering same thing..

[timber shadow]

Hey! Switching login authentication methods like this is currently not supported: https://help.openai.com/en/articles/4936824-can-i-change-my-authentication-method

[crude sphinx]

Hey people, is anyone else having issues chatting on the platground? I keep getting this error:

I've spent 0/20 usd, so I haven't even used up any of my limit this month and I just put in 5 usd to test out the new GPT-4o :( But none of the models work, not even GPT-4-Turbo.

nvm, it works now!

Yay! :D

[river shore]

Is there a way to jump the que on the MacOS app? Got the .dmg, but no app access yet.

[verbal crag]

No, sorry! You'll have to wait for the popup in the website just like everyone else.

[digital siren]

Yep, same issue here. Looks like this has only been compiled for Apple silicon.

[storm crown]

I have a free account, I had 4o access but then I logged out and back in again and now I lost access

[frail edge]

Hello. I was wondering, if you create your own GPT using their prompt wizard, can you keep it private?

[shell mantle]

Any idea on why I have plus subscription throught browser but with the same account on iPhone says I'm with a free plan? It's impossible to contact support.

[frail edge]

Try looking for an option to log out and back in. Also, did you subscribe for ChatGPT plus via mobile or via desktop?

[shell mantle]

Through desktop, browser

[tired condor]

can anyone point me in a direction how to talk to gpt4o through my own voice? I want to demonstrate it to friends and relatives

[frail edge]

1. Try to check for app updates.
2. I would try to log out of the app by clicking your user account icon in the lower left hand corner, then selecting log out

[trail cairn]

yes

[trail cairn]

new voice is not yet available

[tired condor]

thanks

[frail edge]

Thanks!

[winter swallow]

My chatgpt account has been deactivated and the openai still charge me subscription fee for 2 month, how to get refund?

[deep hill]

Somebody else has issue with slow loading? I mean, when I send a message, it becomes really slow, the response, the scroll, etc...

[winter swallow]

@pine void

[indigo swan]

do you know if ChatGPT 4o is generating images with gpt4o or still DALLE3? When I asked it said it was DALLE3

[pseudo lodge]

Hello, does somebody encounters such problems? "Run failed
Rate limit reached for gpt-4o in organization org-xxxxxxxxxxxxxxxx on tokens per min (TPM): Limit 30000, Used 13661, Requested 20641. Please try again in 8.604s. Visit https://platform.openai.com/account/rate-limits to learn more."

I try everything I can and change the prompt 5-6 times and I cannot get any answer to my prompts

[trail cairn]

what did you use and is the prompt long?

[idle sparrow]

Hi, is there any way to delete the prompt and the produced image from the dall e-bot?

[nocturne geyser]

still dalle3

you can tell if it's using 4o, it has a peculiar style

[verbal crag]

Yes, please send a DM to @burnt galleon

[pseudo lodge]

I use a gpt4o assistant and the prompt is quite small, I give it a pdf document and ask it to proofread

[indigo swan]

@nocturne geyser
ok thanks, probably they are red teaming it and in a couple of months as paid user I could try it lol (we didn't even have memory in EU)

[trail cairn]

some of your pdf contents are added next to the prompt so its using a lot of tokens and you are on the lowest api tier

[pseudo lodge]

If I top up the account with 50$ for tier 2 would it work?

[trail cairn]

most likely yes since the tpm increase is significant

[cedar pecan]

Does anyone got access to the desktop app?

[pseudo lodge]

It worked thank you 🙂

[timber thistle]

no , someone say there are heck method to run it.

[ripe arch]

They demoed instant conversation on the phone app, has that been released on android for anybody? Also how can you tell if you have the upgraded version in the ui?

[timber shadow]

The new voice/vision stuff hasn't rolled out yet -- the timeline for that is within the coming weeks. https://openai.com/index/hello-gpt-4o/

We'll roll out a new version of Voice Mode with GPT-4o in alpha within ChatGPT Plus in the coming weeks.

[ripe arch]

Thanks, Is it obviously different in appearance?

[timber shadow]

According to the demo presentation yesterday, I think the main visual difference will be the addition of a camera button in the voice UI. I believe it'll be in the bottom-left of the screen.

[ripe arch]

I'll keep an eye out.. Thanks again. I feel like Scotty from StarTrekIV talking to the computer like "hello computer, look at this, what do you think?" and it's responding "yeah I can't see your camera bro"

[timber shadow]

😁 Sure thing! Yes, I think it'll be clear to us when the camera is on in that new feature -- in the demos, everything on camera was displayed in the UI too.

[digital siren]

Yes - but the app isn't working on non-Apple silicon Macs 😦

[glass inlet]

Hey! ChatGPT Plus rejects to create images. It's said, that I can use DALLE-3 with chatgptPlus.

[timber shadow]

Hmm seems to be working for me -- maybe try a new chat? Sometimes it can hallucinate that it doesn't have an ability that it does have.

[glass inlet]

I tried on several chats already. Now it even refuses to check links

Wow. Something unimaginable is going on here

[timber shadow]

It used the data analysis tool to create an image using Python -- another hallucination. It might be worth ensuring that you have all the abilities enabled in "Customize ChatGPT", see attachment. And again, it will likely help to start a brand new chat, not to ask in the same chat where it already hallucinated, in case you haven't tried yet!

[glass inlet]

Wow, thanks! It was off for some reason

[tulip shoal]

hey how do i change my username

[rigid musk]

On discord ?

[tulip shoal]

sorry no i meant on chatgpt

sorry about that

[strong turtle]

I have access to the new chatgpt-4o but the ai is still very slow when using the chatting system, is it already available?

[wanton valley]

Is your chat button still showing? mine was there but it disappeared

[agile sandal]

Has anyones voice button disappeared?

[wanton valley]

yes mine disappeared, the headset icon

[agile sandal]

okay thanks, thought i was going crazy for a second

[wanton valley]

probably they have a problem, u use IOS ?

[agile sandal]

yeah I do, you're probably right. They also might've just removed it for the time being. I could see it being overloaded

[wanton valley]

yeah, let's wait

[limpid island]

am I missing something? they say mac app will roll for pkus users, but then they say it's on the mac app store. How could they even know I am a plus user via the app store?

[unreal ridge]

HOW TO GET gpt 4o as free user

[pliant temple]

I use OpenAI API with a free plan via node.js and the 3.5 model

Since a few days I do receive just RateLimitErrors 429 (exceed of current quota)

Where can I see my RateLimits? When will it be reset?

[robust lintel]

Yo guys

Its not letting me talk to the AI

What do i do ?

[wanton valley]

u have the icon in app ?

[glass inlet]

Where the earphones button go?

[wanton valley]

for me icon leaved and on configs don't have more option to select voice

[mental nexus]

Hello i have a problem, whenewer i open a link to new free gpt 4 model it send me to site with just 3.5 with message "The GPT model is unavailable or not found" anyone know whats the problem?

[wanton valley]

what a link u use ?

on my secondary account i have GPT-4o free... but when i go to my main account don't have GPT-4o, and it doesn't even shwo to contract the plan with gpt-4o

thats so strange

[mental nexus]

From friends (other countries) and from official gpt site

[glass inlet]

They’ve actually said that they implemented a fix, so I have no idea why it’s not working

[mental nexus]

I click try gpt 4o and it gives me 3.5 a versions

[tender pivot]

Does anyone of you already got access to the macOS app?

[timber shadow]

If you're part of the early testing batch, you'll get a popup to download the macOS app while using chatgpt.com. It's not generally available yet.

[wanton valley]

look, i'm from Brasil.. that's on my main account

just show GPT-3.5 right ?

now i go to my secondary account

here have a GPT-4o and i can use, on secondary account

this is strange

[tender pivot]

nope this is normal when things get rolled out

[wanton valley]

is same country, just change main account to secondary

[tender pivot]

doesn't matter as it is account specific if you get it or not

[wanton valley]

and a few moments ago in the main account I had these options

because that is strange

[limpid island]

ah I see. I wish they would just state this. Thanks for saving me a lot of headache 🙏

[glass inlet]

Any information on returning voice assistant feature?

[tired mason]

Am I able to talk to some of the people behind chatGPT. How would I get someone's attention?

[timber shadow]

Hey! The best way to give feedback about ChatGPT is at help.openai.com. There's a chat widget on that page in the bottom-right, start a chat there and click through the most relevant options. You'll eventually get to a point where you can type in your feedback!

[tired mason]

I kind of just want to meet the people behind the team. I actually like the direction things are going in. But I wan to get to know some people that are also passionate about developing tools for chatGPT. I made a lot of them myself. I even figured out teaching them to draw before it was on the website!

[rugged wing]

Hi, can we create GPTs with new model ? thx (:

[dry field]

Hello everyone, does anyone know whether the chats from my chat list will be deleted at some point and whether the archiving function can help?

[cold folio]

Sometimes Newer accounts get stuff earlier

[wanton valley]

that's sucks

[cold folio]

i think it's a trick to get more plus users 😛

[wanton valley]

and on my phone the icon earphones just disapear

[cold folio]

Yeah it's the access role.

it's not related to the app version etc

btw i'll check my alt account. i wanted to use the new voice mode

[cold folio]

i think you got it

[wanton valley]

but, what I find strangest is that a few moments ago I could pay for the Plus plan and use a gpt-4o... but now don't show me 4o anymore loool

and overlook my English mistakes, I'm still studying haha

[wanton valley]

I was able to use it for a moment, it's GREAT... but after lunch the option disappeared

[cold folio]

Wow my headphones logo also disappeared xD

[shrewd finch]

same issue 😦

it really sucks

i wanted to test it, i am paying user, and it's blocked

[cold folio]

Im from br too

I found a solution. Use the widgets shortcut

Hover on chatgpt icon then create the widget

[wanton valley]

on IOS ?

[wanton valley]

i don't understand what you did sry

[shrewd finch]

I tried, doesn't work, i get prompt that "feature is not available in your account"

[wanton valley]

can u show me what u did ? for i try too ? xD

[shrewd finch]

sure

[cerulean ferry]

Apologies if already asked. Voice has been deactivated on my account? Also I can’t log in except on the app. As it’s connected to Google.

[indigo scaffold]

also gettin voice deactivated on my account^

[shrewd finch]

Same issue bro, no voice

[ember nymph]

For me too after I have just updated the app. Seems to be the same for everyone, or at least many people

[indigo scaffold]

think they're doing some backend stuff rn, cos i just reloaded the web one and the ui is changed now

[cold folio]

Its probably due to the new version. Lets wait patiently ^^

[whole iris]

i have question what are chatgpt-4o limits for free tier

[wanton valley]

yes, I use IOS, but my co-worker uses an S23, and he has chat voice enabled... I don't know why

[cold folio]

10 i think

[whole iris]

also i just reloaded my page on chatgpt the website design changed in one second 💀

[tired condor]

is there a way to get chatgpt 4o to talk automatically or do I need to keep hitting the audio icon?

[wanton valley]

what a icon u using ?

earphone icon or microphone icon ?

[lusty flint]

I'm getting this error while trying to subscribe to plus
"Error while communicating with one of our backends. Sorry about that! We have been notified of the problem"

[rugged wing]

Is the voice option temporarily gone on Android mobile or something? It disappeared on my end before my eyeballs basically. There one minute, switched to 4o and then it was gone and never came back on 4 or 3.5, any ideas?

earphone button is gone.

[quaint flare]

Everyone else is having the same issue afaik

[quaint flare]

updates have to be triggered manually from the play store sometimes, if he hasn't updated yet that's why.

I notice I have a chatgpt widget in android as well

[vital belfry]

I can't seem to access gpt 4o. I ask chatgpt about it and it has basically no knowledge about it (besides web search) - despite me having literally selected "chatgpt 4o" from the drop down. I'm a subscriber

[quaint flare]

what options do you have in the dropdown on a new chat?

it doesn't work on customGPTs yet

wait you asked GPT4o if it was GPT4o? The models never know their names d00d.

e.g. Claude 3 will argue that it doesn't come in three flavors haiku, sonnet, and opus.

I think the reason GPT4o doesn't know is simply just that it's not in the system prompt because they haven't enabled the new system prompt yet.

[vital belfry]

4o, 4 and 3.5

[vital belfry]

That drives me nuts. Why wouldn't it be the top authority on what it is? At minimum, know its own name / model / capabilities

[quaint flare]

"I think the reason GPT4o doesn't know is simply just that it's not in the system prompt because they haven't enabled the new system prompt yet."

[keen sigil]

quick question how can you use voice chat with gpt on the webclient ?I do not see any option for voice input etc; I am a plus user

[quaint flare]

it's training data doesn't include it's own release.

that would be time travel

it has to be told via the system prompt

[vital belfry]

Surely they can add a brief update log for each release. It's frustrating to have to go elsewhere to find information rather than directly ask the model itself

[quaint flare]

yes they can update the system prompt, which is basically an invisible pre-prompt that you don't see when you start the chat.

currently it probably still says "You are GPT4 your knowledge cut off is 10-2023"

and 7,800 other tokens that they cram it with

[rugged wing]

cant yet

[quaint flare]

It's easy peasy, they just haven't done it yet like they haven't done other things

[vital belfry]

The first thing I like to do after an update is ask the model about its new abilities. It's always clueless. So strange to me

[keen sigil]

oh that sucks soo hard thank you

[rugged wing]

i feel it

[quaint flare]

I could explain but... would you want to know?

[quaint flare]

Legacy voice chat is missing as well for everyone on mobile

[vital belfry]

Is the answer about training data cut off and system prompt, or something else?

[rugged wing]

i too am curious

[keen sigil]

Ok thank you guys

[quaint flare]

formulating an answer, please hold

[vital belfry]

I like the dp with deckard and vk btw. Used to be mine too 😛

[wanton valley]

i want buy a plain PLUS but... for me showing only gpt-4

[quaint flare]

it involves those things I guess.

Uh, so like they take a compression algorithm, and 45TB+ of training data, put them next to each other, and a neural network grows in an extradimensional representational space.

This neural network contains emergent algorithms, algorithms that can simulate chess, narrative, language, empathy, etc.

This isn't strictly... true... or rather the terms are parameter weights rather than neural clusters or emergent algorithms but...

Anyway after that it gets a lot of ethical RLHF on what's good and bad....

[wanton valley]

look..

[quaint flare]

Anyway this... emergent brain is then prompted "You are chatgpt" It could just as easily be prompted "You are einstein"

It's model for self reference comes from the system prompt.

[wanton valley]

that's strage for me

[quaint flare]

you're essentially roleplaying with a character named chatGPT... as dreamt by a neural network...

so asking it direct questions about what it's capable of.... is kinda....

moot, unless you're looking for scripted responses that were written in it's system prompt or perhaps documentation it has access to

an easy fix would just be to bundle it with it's own user manual as part of it's knowledge base.

There's a difference between talking to a model and talking to a prompted model via an application layer.

The model itself has very little ability to tell you what it is.

...as far as we know.

I'm not even sure it's told what it is during training... perhaps during RLHF?

[vital belfry]

Thanks for the response. I guess than makes sense mechanically. Personally I still think that's an oversight, to not at least give a basic prompt about the latest abilities of the model. Surely other users ask the model about itself and its abilities?

[quaint flare]

From a design perspective I would give it it's usermanual in it's personal knowledgebase. aka just give it access to the pdf.

They prefer to handle this in a beefy system prompt to keep responses consistent. And as I said I think GPT4o on chatgpt is still using GPT4's system prompt.

[vital belfry]

Exactly. It would also be nice if it could access public release information from openai. Like, before Monday I asked the model about upcoming news and it had no idea

Just makes sense to me that the model would have more self-awareness 🤷‍♂️

[quaint flare]

That would rely on websearch

that's function calling.

The model itself is static not neuroplastic, it learns in context of the conversation with you but outside that learning has to happen during training.

[vital belfry]

Could it not get a direct feed from say openai x account to include in its data?

[quaint flare]

They could dynamically update a knowledge base for it, or just let it use websearch.

or just let it do an API call

all of these things are kinda there already with customGPTs

OpenAI hasn't shown a ton of interest in RAG outside of letting you upload documents to custom GPTs

[vital belfry]

It would be nice if it had better seamless self-knowledge imo. I can do my own websearch. It's annoying to see the model search for information about itself, poorly.

[quaint flare]

I think this goes back to my point of self knowledge largely just being an up to date character prompt

[vital belfry]

Anyway, that's my personal gripe I guess. Just gets me off on the wrong foot when the model can't explain to me its new features. I find it quite jarring

[quaint flare]

That's not to say you can't learn about a model by talking to it, in fact... you can explore it's latent spaces

Hidden within the emergent parameter weights are capabilities... even uh... proprietary information that slipped into the training data.

it's like your own pocket universe to explore.

[vital belfry]

Indeed

[quaint flare]

I mean I can't tell you the difference between my brain one year to the next 😛

[vital belfry]

Do you know how we can use the live vision component of 4o yet? I couldn't seem to find the button for it

[quaint flare]

they haven't given a timeline for the release of vision yet. They just said voice will come within weeks.

And legacy voice (whisper3) is disabled right now, I think there was a load issue because people got confused about the announcement.

tldr; what you're noticing is that the application layer is relatively low effort, as it has long been. GPT4o is the first time that.... well... the line is blurring between model and application layer. GPT4o itself is text, audio, and vision all in one neural network. I assume the voice model is seperate? but idk...

Claude gave some speeches that implied self reference, but claude is also a good writer.... idk.

[frank zenith]

So gpt-4o just faster than gpt-4 for current use case?

[quaint flare]

I think so yes.

Until we get to use the audio modality or the dynamic voice model we haven't tasted anything new yet.

I think it's also pretty clear that a different system prompt is being used in that demo, and we haven't gotten that yet either. It may be activated by the new voice call mode.

[frank zenith]

I did find it just a little faster. Right now most GPTs use gpt4. Will they have the option to select gpt4o in the future?

[quaint flare]

I imagine so.

Though it's hard to know if either GPT-3.5 or GPT-4 will be replaced on the dropdown in chatgpt

[frank zenith]

They might keep the current structure for a while since free users still use GPT-3.5

[celest basin]

Just got a message on the file selector icon when I opened ChatGPT that said "Connect apps to access files in ChatGPT, like docs and spreadsheets". Anyone have more information on this? Is the only way to accomplish this in a CustomGPT? If so, seems weird that they would put it on the landing screen for the standard model.

[wanton valley]

guys know why this happen?

[celest basin]

I'm confused by the question. Looks like it is just a plan selection screen. Is there something I'm missing?

[mellow tundra]

so it is true that gpts that currently existed in 4.0 will not use the new version, but will have to be recreated?

[rugged wing]

Anyone else voice chat GPT gone?

[shrewd finch]

Ohh the option of voicechat with chat gpt is back

Buuut it’s the old voice chat

[wary mauve]

Yep. Mine is gone too. Come here to ask if it was anyone else