community-help

[vivid mortar]

Highlight old messages/conversation into a new chat and have it summarize it there

[rugged wing]

How do I that

[rugged wing]

How

[vivid mortar]

Copy and paste on your phone/computer

Highlight the text and copy it

[rugged wing]

All of it

Ok one sec

[vivid mortar]

I'm not sure what your question is, if your conversation back-and-forth with the AI is more than 50+ messages it's not going to remember what you initially talked about anyway. There's only a certain context window of memory it has

[bright basin]

Yo

[vivid mortar]

So you'd have to summarize it every once and a while for it to remember what you first talked about

[vivid mortar]

No, not pasting it here, into a new chat and have it summarize it

[vivid mortar]

Yoyo

[bright basin]

I wanted to say, GPT, best AI. Absolutely enjoying the product.

[crimson sluice]

Ehhhhhhh...

It's not the best, it's decent.

[bronze charm]

What is the best?

[crimson sluice]

Writesonic.

[bronze charm]

Different use-case 👀

[crimson sluice]

yeah true true

[bright basin]

One google check. Writesonic is powerd by GPT3.5 and 4 plus it uses Google.

[crimson sluice]

DAMN

[violet terrace]

Will Chatgpt Plus users automatically get Dalle-3 or?

[rigid musk]

Starting in October in chatgpt chat for plus user

Don't know yet the limit of it

[violet terrace]

Sounds pretty good, Midjourney is pain to use.

[desert apex]

I have been using the ChatGPT 4 for over a month

[bright basin]

Guys, had you have the same problem, that "normies" are evading to use AI but cry about their own obsolecence all day? Honest question.

[desert apex]

It's nice

But limited seems to be up until 2021

[bronze charm]

Jan 2022 as of a few days ago

[desert apex]

Have anyone used the web version of it

[violet terrace]

Yup. And whenever you try to explain to them, they are defensive and hostile towards AI.

I use chatgpt like I use my laptop, all the time for almost everything.

[desert apex]

@bronze charm you are right

[vivid mortar]

Glad I can volunteer and help everyone learn to use it and excel creativity with these incredible tools :)

[desert apex]

@vivid mortar I missed out. Can I get a demo again

I want to go more in depth

[regal lynx]

In depth with what?

[desert apex]

Ai

[bright basin]

m2 ever day all day.

[regal lynx]

The server staff team doesn't work for or represent OpenAI 😅

[violet terrace]

When you have massive emails to read and you just drop it in GPT to get the bulletpoints

[bright basin]

When you just paste in the Readme and GPT tells you everything about it.

[violet terrace]

Yes, That's what I do with guides. I just put in entire article or guide and let GPT tell me the instructions.

[bright basin]

Well, nice to meet you, Tiberius. 👍

[violet terrace]

Nice to meet you too

[analog shore]

Yo people still call "normie" to people with different opinions?

💀

[rigid musk]

No since we all agreed to be respectful in this discord #server-rules

[undone owl]

Is dall-e capable of creating stl model files based on a description?

[tribal talon]

Look into shap-e

[rugged wing]

I can't copy it all it is to long

[vivid mortar]

Copy it in sections

[rugged wing]

Oh ok

[vivid mortar]

150 messages? Summarize 15 message chunks, then take all the summaries and summarize that.

I'm not sure what you’re trying to achieve though

[modern scaffold]

Good Day, I am currently having issues with the loading any document for the analysis to review:

Finished working
Show work
It looks like there was an issue while processing the file. Let me try again to read the content for the evaluation.
Finished working
Show work
I apologize for the inconvenience. It seems there was an issue with accessing the file you uploaded for evaluation. Let's try to read the file again.
Finished working
Show work
I apologize for the inconvenience; it seems there's an issue with the code execution environment. Unfortunately, I am unable to directly read the file you uploaded at this moment.

Can anyone offer any aid?

[sturdy robin]

Which (if any) of the ChatGPT plans allow you to upload sensitive info, where no one can see your data and it's not used for training (API, Plus Plan, or none?)

[bronze charm]

API and Playground requests will not be used to train our models.

[sturdy robin]

By playground do you mean the Plus plan?

[bronze charm]

No

https://playground.openai.com

[sturdy robin]

Ok, so even if someone is paying for Plus, they shouldn't upload or paste anything sensitive. I would have thought that should be a benefit to a paid plan, that it's more private

[bronze charm]

You can turn off chat history and training in settings

[sturdy robin]

Ah ok thanks

[feral jasper]

Dalle3 can create text to the images?

[timid hawk]

Where can I find the top custom promp instructions created by the community?

also how can I provide images to ChatGPT to ask questions about it?

[primal robin]

Hi guys sorry about this newbie question but does the gpt-3.5-turbo-16k-0613 support function calling or that’s just the non 16k version

[rustic marlin]

it is better at it than previous models and even better than other AIs

[feral jasper]

Yes yes, but can it generate text correctly?

[rustic marlin]

it can more often generate correct text

it is far better compared to dlale-2

but it will not make proper text 100% of the time

[storm summit]

Hopefully they don't heavily censor it like snowflakes

[regal lynx]

???

[steady cedar]

Is GPT-4 available on the Playground or is that solely for API calls?

[sturdy roost]

why your writing english texts as a native German?

[granite topaz]

Hey, i installed the python Open AI API, and have the issue of not adding a payment. Since i am underage, i do not and shouldnt have access to any type of debit cards or credit cards, so, i have a pre paid card, which the Open AI billing accepted, but im worried about which account it will charge me on... like the money on my pre paid card or the bank account, which my parents have access to and wouldnt be too happy

so, yeah, will it take money from the prepaid or the bank account?

if anyone knows, let me know please

hello?

[granite topaz]

anyone available?

[chrome prawn]

Why is it putting out these weird parantashes things

[trail cairn]

only cards are accepted so it will charge the card. only real credit cards are officially supported and anything else may not work or will be denied

[fervent pendant]

Hello. Is it possible to talk to ChatGPT over the API, with a plugin enabled?

[trail cairn]

api does not have plugins

[rugged wing]

How do you know?

The reason that I am writing all my texts in chatgpt in english is because I feel like I get better responses rather then when I am not writing english

[sturdy roost]

Your settings are german, so its obvious. Ah okay now i understand

[granite topaz]

so it will use the money from the pre paid card? it wont touch the bank account?

[granite topaz]

?

[solar kelp]

hey is anyone getting this error today whilst using dall-e? i tried random prompts and no idea whats wrong

[somber phoenix]

Hi guys, what function do I call in node.js to list all available engines in the openai API?

[rustic marlin]

I don't think it is a method of the API, but you cna check them at this page: https://platform.openai.com/account/rate-limits

[primal shoal]

New here, I'm wondering why all of my images are shown extremely close up. The images come out showing a very small section of flowers or portions of the top and bottom of a drawing of a seahorse.

[vestal basin]

Hi, where/how can we experiment and test with the new Dall-E 3?

[trail cairn]

"DALL·E 3 is now in research preview, and will be available to ChatGPT Plus and Enterprise customers in October, via the API and in Labs later this fall."

[vestal basin]

So Dall-E 2 users will be able to use it?

(Labs)

[trail cairn]

yes as long as you have credits

[vestal basin]

thanks!

[sudden delta]

How can it say it'll be available to gpt plus users if it says it needs credits..
(I tried)

[tribal talon]

ChatGPT and labs are separate platforms. DE2 over labs now uses credits, or you can use it in API for $$. ChatGPT plugins for plus users is the $20 a month subscription. They're different things.

If DE3 is on chatgpt, you can use it with whatever limits it has for plus users. And if it goes to labs, it likely will need credits.

[pseudo violet]

Anybody know what the ballpark price would be to fine-tune gpt-3.5-turbo on a 10GB dataset?

[trail cairn]

pricing is token based so count the token size for the dataset

[pseudo violet]

The dataset is called SHP on Huggingface. I tried to post the link but I got put in timeout

Not sure how I can calculate the tokens

Oh, it says "up to 10.1K T5 tokens"

What is a T5 token

[trail cairn]

you would need to convert the material to the format openai requires to be used with fine-tuning

[primal robin]

Hi guys sorry about this newbie question but does the gpt-3.5-turbo-16k-0613 support function calling or that’s just the non 16k version

[brave sun]

I need some help - ChatGPT won't work on Chrome or Edge on my computer but it will work on Firefox. I have restarted, cleared cache, deleted all browsing history, changed privacy settings..... I need it to work in Chrome - anyone have any ideas?

[trail cairn]

do you have nordvpn, avast, etc.

[golden ruin]

maybe vpn?

[brave sun]

I don't know about VPN. I picked a free one and added the extension and it still didn't work. Then I watched a video that said not to have VPN on it

[trail cairn]

your computer has something installed that affects chatgpt on all chromium browsers

[brave sun]

How do I find what it is?

[trail cairn]

only you know what you have installed so you either start removing apps or just use firefox

nordvpn for example installs web protection service that is always on and breaks it

[brave sun]

Ok, I guess its time for hit or miss - just start deleting apps 1 at a time and trying it....

Is there a chatgpt app i can use on my desktop vs the website?

[trail cairn]

nope

[brave sun]

OK! So then I can't add the app and then have my firewall approve it

[mossy wagon]

hello. i have a question about the api and its costs, maybe someone can help me. i want to use the api for my software, but this software should be usable by several people at the same time. do i need a separate openai account for each user and have to insert their key, or can i combine this by one? thanks for your time!

[vagrant plover]

Would it not be easier to charge other users for your software?

Also, I think you may find better help in #dev-chat or #1037561751362863144

[mossy wagon]

that's right. thanks for your answer. and sorry that i used the wrong area for my question

[vagrant plover]

No worries! Happy to help

[marble ravine]

Do you know if they are planning to release this to non-enteprise customers. We are still on the waitlist for enterprise but want to start using Dalle-3 instead of midjourney for our SaaS as soon as it is available!

[tribal talon]

The release post says API and labs and chatGPT plus as well as enterprise. So of course?

[marble ravine]

got it, it was just the wording that made it seem like it was only going to be released to enterprise. Looking forward to the Labs release!

[jovial brook]

Hello, If I understood correctly Dall-e 3 will not be accessible outside of ChatGPT? I use GPT-4 via the API or the playground, I don't use ChatGPT, so I would never have access to Dall-e 3?

[tribal talon]

Later this fall it is said to come to labs and API. Chatgpt and enterprise is just the first wave in October.

[jovial brook]

Ok thank you very much, with my bad English, I didn't understand everything. I'll be patient then ^^

[fossil tusk]

a 600kb dataset cost me 15$

it doesn't 'retain' data in my experience, only response types.

[rustic marlin]

10 GB sounds like a lot, not so fine tune, i guess

with that much data it might be easier and cheapper to code something that can search on this dataset and insert that as context on the AI reply

[storm summit]

Hey everyone does GPT 4 seem more heavily censored than before

To the point in where it actually deletes the AI response which is frustrating?

[rigid musk]

They update it so people dont break the tos

[worn marsh]

Hello, is there a feature where I can log out of all my device accounts for GPT4? I have changed my password several times and someone is still using my accoutn to write random blog posts. I have tested this and some of the devices I have dont actually get logged out when I change my password. Any idea what to do in this situation?

[smoky pilot]

Is there anyone who can help me?

How can help me to understand fine tuning of chat-gpt?

[buoyant drum]

I see a link (see image)
But it launches into the most bizzare multi-step captcha I've ever had and I can't seem to get past it.

[formal dock]

That link goes to https://chat.openai.com. Try going to the website directly if you are having trouble clicking that link.

[buoyant drum]

Thanks man. I realized it actually says DALL-E 3 is not available until October on the splash page anyway. Guess I got caught. "It's a trap!" LOL

[formal dock]

No problem, anytime!

[faint temple]

Hi

[smoky pilot]

Hello. Who knows embedding in openai?

[vivid mortar]

Here is some more information about embeddings:

https://platform.openai.com/docs/guides/embeddings/embeddings

[smoky pilot]

Thank you.

I read all about it. But I'm not sure what it is. Can you explain in one word?

[regal lynx]

One word: relatedness

[somber phoenix]

Hello, when I tested with one of the models, it gave me back weird responses, what do I do?

The 'How can I help you?' message is set up by me, not generated by the model

[smoky pilot]

Thanks. Is this a good idea? I have more than 10G of database. And I need a chat bot that reply for this database. And then, it's impossible to fine tune the model, because it's huge. So, my plan is to analyze customer's request, and then send a query to database, and then make reply with the query result. How's this?

[trail cairn]

fine-tuning is not for adding knowledge anyways but with embeddings you can do it and have no size limit

[rugged wing]

When you hit rate limit error or context limit reached error, in open ai gpt api. Do you get charged in either cases?

[storm summit]

But now it's stupid I can't even have it write me romance novels because apprantly even something as simple as neck kissing or kissing in general is apprantly harmful

I can't believe am getting babysat as an adult to not be able to do stuff like that. I see why people prefer open source local llm now

Well am more excited for lama 3 every update OpenAi manage to make me loose faith in there vision. Guarantee they have a model where it's less restricted that they use. AI safety my bum

[rigid musk]

That's their rules . They focus on a product rated G .

[storm summit]

And you're fine with that being baby sat?

Well I guess no one forced me to pay for plus

[rigid musk]

I use it for rpg giving idea and suggestions . Filling part i want by writing my own stuff .

So yes I'm fine with it. Working great

People force it so they cut that part.

[storm summit]

....

[rigid musk]

If you don't like the product try other ai .

You can always thumb down a prompt result and give suggestions.

For other ai you have #ai-discussions

[storm summit]

Thank you will check it out

[stone bear]

Hi i have been having issues with chatgpt I have subscribed to the plus subscription yet gpt 4 is still locked

[vivid rivet]

Guys, when you buy a tariff on a new account, the prepayment mode automatically turns on, or can you somehow change it to monthly payment?

[heavy adder]

Help me some one please. I have added 10$ last week and used playground just fine. But now it say You've reached your usage limit.... I check the balance was 0, and i added more 10$ but it still show an error. It say my limit is 10.. but actually in settings is 120...

Nvm, i should have more pation... works 🙂

[trail cairn]

yeah was about to say there is a delay after adding more

[heavy adder]

Thanks 🙂

[surreal mica]

I'm a ChatGPT Plus kinda guy, and it sounds like my subscription will give me access to Dalle-3. Am I understanding this correctly? Or will there be an additional charge?

[sudden delta]

Knowledge cutoff September 2021 huh?

[rose glen]

Based on the announcement, it seems it will be available as part of CHATGPT Plus

[rose glen]

It seems ChatGPT had an update recently through January 2022

[sudden delta]

Mmm idk about that
I tried asking how it knew about that storm and it said that what it said was wrong and how it's knowledge is sep 2021

Now it's not perfectly accurate but if that's hallucination thats damn good

It says like it reached peak strength Oct 30-31 In reality it was 27th

[trail cairn]

gpt-4 is 2022 jan but for 3.5 its still being told its cutoff is 2021 sep

[sudden delta]

Yeah this is 3.5

I used a political event since those effects last a long time and can get ai to list a timeline

Its hallucinating very convincingly

Irl that last one happened in Sept 2022 but still the fact it knows he got indicted at all-

[somber phoenix]

Hello, when I tested with one of the models, it gave me back weird responses, what do I do?

The 'How can I help you?' message is set up by me, not generated by the model

[rapid flax]

yo guys i think ChatGPT is broken
i started dialogue and ask how much 1+1

[rigid musk]

With the api ?

[trail cairn]

you need to use the share feature and not cropped image as trolls have attempted the same before

[somber phoenix]

no

I put that manually in a chatlog. I was following a tutorial that helped me build a chatGPT clone-prototype and that's what was done there.

[rigid musk]

Try in the #1037561385070112779 #dev-chat

Since it isn't chat.openai.com

Maybe people know more there

[rapid ermine]

Does OpenAI do fractional billing? For example, a request to GPT 4 is .03 per 1k tokens. If my request is ~150 tokens I get charged a penny. Is that accurate, that 1 cent is always the minimum? Just want to make sure.

[fickle abyss]

Guys how do i make GPT summurize books from a video in tabels or rows
i saw someone do this command

[rustic marlin]

is this running on OpenAI API?

in case it is, which parameters you used for the request?

[somber phoenix]

It is

[somber phoenix]

you mean this?

Don't worry about the comment.

[outer kraken]

Hey I’m thinking about purchasing chat gpt 4. I want it to be able to have access to the information of new released books. Ik chat gpt 3 have book information from 2021 n below. If not would a plug in work??

[rustic marlin]

the screenshot you sent look like chat messages, try using the chat completion model

[rustic marlin]

GPT-4 has the same knowledge cutoff date

[outer kraken]

So a plug in wouldn’t help?

[rustic marlin]

plugins that can access the internet can help witth that
im not sure if there is a plugin just for checking information on newly released books though

[outer kraken]

You know which plug ins would help?Im trying to come to a discussion weather to pay for the upgrade if it has this feature

[somber phoenix]

how would I do that

[rustic marlin]

by using the openai.chat.clompletions.create() instead
check the examples:
https://platform.openai.com/docs/api-reference/chat/create?lang=node.js
https://platform.openai.com/docs/api-reference/chat

[smoky pilot]

@rustic marlin, expert, can you help me?

[rustic marlin]

sure, and thanks for calling me an expert =P

[smoky pilot]

Thanks. I have a large database of more than 10G. And I have to use it for chatbot. Can you tell me any guideline?

[rustic marlin]

there are a few ways to handle that, all of them will require some knowledge in coding

doing a fine tune might worth it if the system is going to be used a lot, because a fine tune on a 10GB dataset will get pretty expensive

[smoky pilot]

But fine tuning has limit.

[rustic marlin]

yea, I htink it is lower than 10GB, actually

[smoky pilot]

At most 50 million tokens.

[rustic marlin]

at this scale, it might worth more to have a consistent way to make searches on that database and use function call to figure what the user wants, search the database and inject the reply from the search as context to the AI

[smoky pilot]

Do you know about spaCy?

[trail cairn]

for new knowledge you use embeddings

[smoky pilot]

So, what's the limit of embedding?

[trail cairn]

there is no limit

[smoky pilot]

Really? Then how about price?

[trail cairn]

https://openai.com/pricing

[smoky pilot]

Thanks. By the way, can you explain about embedding?

[trail cairn]

https://platform.openai.com/docs/guides/embeddings

[smoky pilot]

I read the guidlines. But can you explain in one word? What's the input, output, and what's it for? ...

[trail cairn]

that page has link to openai cookbook which has examples

[smoky pilot]

I've already read it. But I'm not sure. What's the input of it.

[trail cairn]

Question_answering_using_embeddings.ipynb shows everything you asked

[smoky pilot]

How about this? From customer's request, we develop analyze code for making queries for database.

[trail cairn]

you just need to test it

[smoky pilot]

So, embedding is trustable?

[trail cairn]

yes if you make good quality embeddings and combine them with good prompts + gpt-4

[smoky pilot]

But the size of database is more than 10G

Is it still possible?

[trail cairn]

there is no limit for them other than your own storage size for them which is usually a vector database

[smoky pilot]

Do you have any experience?

[trail cairn]

yes

[smoky pilot]

Can you introduce me any open-source on github? or ...?

[trail cairn]

the openai embeddings page lists them

[smoky pilot]

Thank you very much.

For embedding, do you need those pytorch framework or other frameworks?

[trail cairn]

depends about your platforms and other things but generally no

[smoky pilot]

OK. Thanks. So, can you give me the flow of embedding? Just in one word.

[astral ridge]

When you train a custom model, is the data fed into all models for query (ChatGPT 3.5, 4)? Or how can one query these models?

[trail cairn]

there is no single word other than itself

[solid sedge]

How good is ChatGPT3.5 at translating Ukrainian to English vs say Google Translate. I need to translate my MIL’s stories for a book for her grandchildren who do not speak Ukrainian. Alternatively she can write in Russian if it is better at that.

[trail cairn]

probably a lot better than google translate, deepl is also good

[smoky pilot]

So, I can do with only GPT-4?

How about gpt-3.5?

[solid sedge]

I’ve not heard of deepl. I will look into it as well

[smoky pilot]

@trail cairn, embedding, so what's the result of it?

[trail cairn]

3.5 works too but if your task is code, queries etc. then gpt-4 is still superior

[trail cairn]

check those cookbook examples about sportss and wiki

[smoky pilot]

Where's examples related to embedding?

[trail cairn]

#community-help message

[smoky pilot]

In embedding part, there's no such examples.

[trail cairn]

there are several in the examples folder

[smoky pilot]

Can you give me the exact url?

[trail cairn]

no, bot does not allow external links

the embeddings page has the cookbook link to github

[smoky pilot]

https://github.com/openai/openai-cookbook/?

Is this right?

[trail cairn]

yes

seems it allows them

https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb

[smoky pilot]

Thanks very much.

Did you receive my DM?

[trail cairn]

i dont use them on this server as i get too many

[smoky pilot]

Wow!

Let's keep in touch.

[trail cairn]

you always ask in #dev-chat which is for devs

[smoky pilot]

And then, can you tell me the experience in embedding?

For example, for 1G, you took how many days? or like this ...

[trail cairn]

you limited by your accounts rate and payment limits so you cant compare

by default you are capped to $120 per month which is not enough for large usage

[harsh crow]

Is AI Free and why sometimes it Updates?

@trail cairn

[smoky pilot]

So, how long did you take for 1G?

[trail cairn]

website is free with some limitations

[harsh crow]

Like?

[smoky pilot]

Do you think embedding's accurate?

[trail cairn]

you dont measure anything in data size when it comes to openai so there is no answer for that

[trail cairn]

like how many messages you can send per hour

[harsh crow]

is 100 limted or more?

[trail cairn]

its not a public number but it will tell you when you hit it

[harsh crow]

ok thanks, is there a support link?

[trail cairn]

https://help.openai.com

[wind creek]

anyone here use claude?

[harsh crow]

thansk

[smoky pilot]

@trail cairn, so, it's really accurate?

[trail cairn]

yes, but everyone has their own data and use cases

[smoky pilot]

Very thankful. With AWS, how long does it take to finish embedding, do you think?

[trail cairn]

you need to read the examples first

[sick wyvern]

Is anyone getting this error when trying to log in?

[smoky pilot]

Actually, I have a project related to that. But I cannot guarantee it for client.

[trail cairn]

you first need to read and understand how they work because after that you would know the aws question makes no sense or needs rephrasing 🤔

[smoky pilot]

OK. Can you tell me when you have free time?

Because, after reading book, what time can I call you?

[trail cairn]

you cant 🙂

[smoky pilot]

So, how can I have contact with you?

[trail cairn]

i only help here randomly

[smoky pilot]

OK. Anyway, thanks for your courage.

Can you introduce me an agent of you?

[trail cairn]

nope

[smoky pilot]

I'm really happy today.

[dense gyro]

hi, can someone recommend me a plug-in which is the best alternative to bing search? As is is disabled for now within gpt.

[trail cairn]

#dev-chat is the best place if you later have questions and someone there might also be able to answer

[trail cairn]

webpilot

[dense gyro]

thx

[smoky pilot]

@trail cairn, how about my plan? Getting query from customer's request? Isn't it a good idea?

[trail cairn]

impossible to answer as you did not give any details but in general its possible when all things are done right

[smoky pilot]

But I heard that getting queries are not so accurate.

[trail cairn]

not with 3.5

sometimes not with 4 either without extra work and combined api usage

[smoky pilot]

So, do you know about spaCy?

[trail cairn]

nope

[smoky pilot]

If you combine with apis, what kind of apis do you use? Only chat-gpt api?

[trail cairn]

yes openai ones

[smoky pilot]

What you mean, by combined api usage?

[trail cairn]

dont try to get finished result in one call

[smoky pilot]

OK. I see.

[trail cairn]

alternatively use langchain and customize it

[smoky pilot]

So, in langchain, those template, prompt, functionprompt, system, can you give me some detailed explanation of it?

[trail cairn]

langchain is too broad subject but they also have their own discord and big community around the net

[smoky pilot]

So what do you think? Embedding is more accurate than my idea?

[trail cairn]

you can only find out by trying

[smoky pilot]

Are you an AI expert?

I think so.

[trail cairn]

🤷‍♂️

[smoky pilot]

Then, do you know about those GAN models? For generating video from an image.

[trail cairn]

yes

[smoky pilot]

Really? I'm going to generate a video from an image. And the video is just the character is reading a book.

The problem is reading part, and cloth part.

[trail cairn]

nothing does videos very well yet

by using a lot of time you can do with stable diffusion and million extensions and tweaks or shortcut some with runway gen 2

[smoky pilot]

How do you think the price?

How much can you expect?

[trail cairn]

only you can calculate it

[smoky pilot]

OK. Are you an AI engineer or data scientist?

[trail cairn]

neither

[smoky pilot]

So, can you tell me what do you do?

[trail cairn]

maybe c# fullstack and also game dev

[smoky pilot]

How many years have you been working on AI?

[trail cairn]

started using openai api in 2022 spring so probably randomly since then

[smoky pilot]

I'm sorry.

[trail cairn]

?

[smoky pilot]

So, are you openai master?

[trail cairn]

no

[smoky pilot]

So, are you working on fixed job?

[trail cairn]

yes

[smoky pilot]

Can I call you anytime? I'm growing to be an expert now.

[trail cairn]

nope, best you can try is to ask here

[smoky pilot]

Can I?

[trail cairn]

you can always ask here and see if someone answers 😛

[rigid grail]

The login page say my account has been blocked anyone had experience with that?

[rigid musk]

Make sure vpn is off

[rigid grail]

I dont use vpn

[rigid musk]

Can you try firefox as test

[lilac venture]

does anyone have login issue ? report " Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts. ChatGPT August 3 Version"

[rigid grail]

You mean firefox?

[rigid musk]

Yes sorry typo

[rigid grail]

Okay will try

But it said it has been blocked

Not suspection

And the support guy is seeing my messages and doesn't reply

So do you think it will work?

[rigid musk]

Try just to compare

[rigid grail]

Ok

[stone cipher]

That is the guardrails and that has always been the case; you see this happen a lot with Bing Chat. Try talking to that bot about normal sex-ed stuff and it will censor itself in real time if you engage with it enough.

[steady cedar]

Can you use GPT-4 in the playground on OpenAI's website like you can with GPT-3.5-turbo, or is GPT-4 only for API calls?

[nova salmon]

Is there any way to check if someone else is on my account? I'm seeing a bunch of new chats in ChatGPT in Chinese, but have no idea where they are coming from or why.

[rugged wing]

Does anyone else get that annoying mobile autoscroll. Like I want to start reading responses before chat finishes but it autoscroll while writing until it's done.

[rustic marlin]

yes, you can, just select chat models at the meno on the right

[rustic marlin]

it is very likely you account has been compromized, I strongly suggest you to change passowrds
also, it is known that some browser extensions steal chatgpt credentials, so, it is good to uninstall them if you have any

[steady cedar]

I don't have that option, I never got GPT-4 access because I never use the API for it, Ive always just used ChatGPT

[rustic marlin]

oh, in that case, no, you will not be able to use it

[echo cave]

I would like to know how I can access the GPT-4 API key. I have ChatGPT4 Plus which I pay monthly

Anyone know the best way to do the custom instructions I work alot with Software Development

[regal lynx]

https://platform.openai.com/account/api-keys
this is where you can view your API keys

[echo cave]

how can i verify i am using GPT 4 API

[regal lynx]

if you're using the API, you have to set a model. Assuming you use the GPT-4 model parameter, you're using GPT-4

[rugged wing]

anyone know how to design an ai from the ground up that can be trained to write in a specific style, mainly rap music

[regal lynx]

why do you ask?

[rugged wing]

cuz id like to make a fanmade kendrick album

and post it to youtube cuz people like ai music

itd be a fun side project

[rugged wing]

could you help?

[hazy lily]

Why is chatgpt flagging every little thing? It's frustrating.

[regal lynx]

I can't help with that

[regal lynx]

Maybe starting a new conversation would help? Also clearing you cache and cookies might help

[junior needle]

Where can I go to ask for help with my GPT-3.5 talking to itself

I didn't ask the second User question, that's why it's on the same line...

[rugged olive]

I have a question. I been using ChatGPT for a while with helping with writing and learning to do better sentence structures.

However I been doing this for a hot while, but out of no where today. I been getting spammed with this.

This content may violate our content policy. If you believe this to be in error, please submit your feedback — your input will aid our research in this area.

Yet my writing style hasn't changed.

[charred parcel]

thumbs down the response and provide feedback

[rugged olive]

So I need to do this each and every time?

Because it blocks where I wrote and what it types because it will also flag itself.

[charred parcel]

what do you think is triggering the content warning?

[rugged olive]

Honestly I have no idea. I am doing a fight scene, I have done several of these. I have never been flagged before. So today I been getting hit with it each time and the scenes I wrote before were more detail of injury and such. This scene I am working on is minor compare to some others. I honestly don't got a clue.

[charred parcel]

maybe adjust your prompts to not include content which may be construed as violent

[rugged olive]

So... Why out of months of using it would it just now start?

[charred parcel]

I don't know.

[rugged olive]

If its getting upset over violence. I mean... its a fictional story of a adventurer. Gonna have some battles against beast and bandits and such. Its just baffling because its making it hard to keep working on this.

[charred parcel]

violent content is probably always going to bump up against content standards, you need to adjust your prompts so you don't cross the line

[rugged olive]

Oh... How do I adjust my prompts?

[charred parcel]

write a new prompt that doesn't include anything that may become violent

[rugged olive]

...give me an example

Because I am not following

[charred parcel]

can you share the last chat that had the content warnings?

[rugged olive]

Sure, its horrible but I can do this.

A swift, slashing motion closed in on the young half-elf's face. Upon impact, the razor-sharp claws embedded themselves into his skin, beginning just above the bridge of his softly glowing blue eyes and tearing down his right cheek in a jagged, painful gash. A solitary tear formed, while a few smaller wounds marred his complexion, though these were of a much lesser extent.

[charred parcel]

can you go to the sharing button and share the whole chat? post the link to the chat?

[rugged olive]

How do you do that?

[charred parcel]

there is a sharing button in the top right hand corner, when you click that you will get a link to post

[rugged olive]

When I click it. The Share Link to Chat says.

This shared link has been disabled by moderation

[charred parcel]

ok, then it's likely that you are bumping up against the content standards, you'll need to start a new chat and use a prompt that is less likely to bump up against the content standards. Anything that would not be considered G Rated is probably not allowed

[rugged olive]

...This is just silly. For something that helps and has been just to now start makes no blasted sense. Guess I will either need to find something else or make it rated E with flowers and unicorns so it doesn't offended anything.

[charred parcel]

OpenAI have firmly set standards, anything outside those standards is going to keep triggering content warnings.

[rugged olive]

Either way this is silly and just a waste of my time. I thank you for attempting to help but it wasnt much to help sadly. Chat GPT is just simply to picky.

[steady cedar]

How would I gain access to GPT-4? Seems a little greedy to make Plus subscribers spend money on the API just to gain access to GPT-4...

[novel comet]

Hello, I am struggling with the Human verification puzzles. The images are super blurry but most importantly I am not exactly 100% sure I know what the puzzle is. Is it match the number of "objects" that you see or is it Solve the equation on the left and find the answer on the right?

[steady cedar]

You sure youre not an AI? lol

[novel comet]

have you done them?

[rugged wing]

No one cares about Dalle 3

We want improved:

• Code Interpreter

• Increases token limit in general, but specifically for PDF pages.

• Improve OCR capabilities that recognise mathematical symbols and expressions, as good as or better than MathPix.

[steady cedar]

For anybody that knows, in the Playground on OpenAI's website, what is the "System" box for? Can I use that to store constant data that I want the system to remember even when I reload the website? Or does it serve another purpose?

[west grove]

anyone knows why i have 'account is not active. Please check your billing details?" I have purchased pre paid credit, havent used any of them. I updated the billing details, using two cards to try to see if there was anything amiss. I also think there's a problem with the Usage tab it just goes blank, might be related

[trail cairn]

that may mean that api access has been revoked from that account

[trail cairn]

nearly the same as custom instruction on the regular website

[rugged wing]

Hi guys, I hope someone can help me. Since yesterday I get the following error message: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

Unfortunately speak very poor English and have to use Deepl

[trail cairn]

if you have nordvpn then disable its web protection, else try firefox

[rugged wing]

Thanks for your quick reply. I have indeed recently North Plan. However, I currently have disconnected

[trail cairn]

disconnected it not enough, it also installs web protection service that is always on

[rugged wing]

i understand I will try firefox

[west grove]

i have just created the account...

[rugged wing]

test

[rugged wing]

thank you very much for your help. I was able to turn off the threat protection and now it works again. Without your tip I probably would not have found this.

[trail cairn]

[rugged wing]

I accessed "https://platform.openai.com/account/usage", but nothing came up and I couldn't see anything.

Is this my network error?

[dapper osprey]

How do I switch a chat in chat.openai.com back to ChatGPT 4 model?

[charred parcel]

once it has switched to GPT-3.5 you cannot go back to GPT-4

[dapper osprey]

Why?

[charred parcel]

that's just the way it works

[dapper osprey]

That's bad UX.

[charred parcel]

feel free to add to #1070006151938314300

[dapper osprey]

Someone already suggested it, I added a few details
https://discord.com/channels/974519864045756446/1092669344829427823

This is really just a few lines of JS in the browser, so why is it not deployed yet?

[trail cairn]

try in private mode

[dark perch]

oh nevermind, its my ISP, i tried it with a different ISP of mine and it worked

thanks ❤️

[smoky pilot]

@trail cairn, expert.

Hi. How's your weekend?

I'll use ada-002. But, what's this?

?

@trail cairn

Hello.

[rustic marlin]

those models are very deprecated, it worths more to use the 3.5-instruct models

[smoky pilot]

OK. So, for embedding, what model should I use?

[rustic marlin]

ohh, nvm

[smoky pilot]

What you mean?

text-embedding-ada-002 is not good?

@trail cairn, what model shall I use?

text-embedding-ada-002?

[regal lynx]

Please don't consistently ping members. If they haven't answered after the first one, they might not be online

[smoky pilot]

Thanks for you advice.

[sudden delta]

wait do they give free credits for playground now or is this just a tiny test that i never found before, i never bought credits but i have gptplus

this usually never works

[wintry crystal]

give me a title

[trail cairn]

yes that's the only one you can use with embeddings

[inland birch]

Guys, i have question

How to fix this?

I want to cancle billing

[charred parcel]

can you try to do it from Firefox on a computer?

[warm zealot]

dont I get 5$ free trial?

[trail cairn]

you only get free credits for the first account associated with your phone number

[smoky pilot]

Thanks, Radium.

@trail cairn,So, can you tell me the limit of it?

[warm zealot]

i thought you can only have one account per phone number

[trail cairn]

what limit

[uneven hare]

I am in a rather unique and uncomfortable position with my account - The Google SSO used to create the account is one I need to close. However, I've found no way to disassociate the Google SSO from the account and convert to a regular password based login. Anyone have a suggestion?

[trail cairn]

2 accounts

[smoky pilot]

The limit of pages. And the limit of tokens.
For text-embedding-ada-002

[trail cairn]

there is no page limit, only token limit 8192

[smoky pilot]

If 8192, it's too small.
How can I embed 10G?

[trail cairn]

in pieces

[smoky pilot]

Can you explain it in more details?

I mean, in that case, I can add the embedding data several times?

[trail cairn]

the examples show how to do it

[smoky pilot]

OK. So, it's perfectly possible to embed 10G?

[trail cairn]

yes it's possible to create hundreds of thousands of embeddings out of that data

[jolly moth]

Does anyone know the limit of using Bing AI? I want to know the limit of requests we can send

[smoky pilot]

OK. Thanks very much.

[smoky pilot]

And, the database is being changed everyday, and I have to embed the differences everyday. Even in that case, if you do embedding every day for differences, is it OK for accuracy?

[trail cairn]

its something you need to test

[smoky pilot]

OK. But the acurracy used to be OK?

In your experience.

[trail cairn]

its for our use cases

[smoky pilot]

What's the most difficult thing in embedding? Can you tell me from your experience?

What was your most serious problem up to now? In terms of embedding.

[trail cairn]

only challenging things have been the token limits for both embeddings and chatgpt models

[smoky pilot]

But you told me there's no limit for embedding.

[trail cairn]

no limit how many you can create

and small, good quality embeddings work well

[smoky pilot]

So, you mean, for large, you gotta be careful?

[trail cairn]

you need to split them well for them to be good and for the sematic search to work

[smoky pilot]

So, "Question answering using embeddings-based search", is it a good reference?

[trail cairn]

its one of them, one of the simplest

[smoky pilot]

Can you recommend for large?

[trail cairn]

you need to start with the simplest to first understand how they work

[smoky pilot]

I already read it. What I mean is reference for huge amount of data.

[inland birch]

Didn't work, any more solution?

[trail cairn]

did not read them all, many of your questions already had answers in the examples

[charred parcel]

we can't help with account specific issues from here, best thing to do is seek support at https://help.openai.com

[smoky pilot]

Anyway, thanks very much. You're an expert that mastered AI. You will change the world.

[regal lynx]

What if.... Radium is the AI

[inland birch]

I already visit it, there're no way to solve my problem. Any advice before i give up?

[charred parcel]

you can seek support by emailing [e-mail]

[trail cairn]

👀

[inland birch]

Oh, thanks. I will try that. How long it's take to get reply?

[charred parcel]

it could take some time, be as detailed as possible in your email so the concern can be addressed easily

[inland birch]

Thank you.

[lyric willow]

im so glad OpenAI won't make powershell scripts that download things, saving the world

[rugged wing]

Hello la team, I use the gpt 4 chat and therefore I pay the subscription but the problem I notice is that the data remains in 2021 and not in 2023 but when surfing the net it is clearly said that an update update was carried out.. How to access the 2023 data?

[trail cairn]

there is no 2023 data

however there are plugins which can search some latest info from web

[rugged wing]

ok thank you

Do you have a plugin to recommend to me?

[trail cairn]

webpilot

[rugged wing]

thank you very much Radium;)

[amber oriole]

I am having an issue get my free credits..

[regal lynx]

what exactly is the issue?

[faint temple]

Why does it say on ChatGPT iOS that my ChatGPT says it using an older version of custom instructions?

[olive nebula]

have you updated the app?

[faint temple]

I’ll check

It updated to the latest

Is this because I’m not on iOS 17?

[olive nebula]

if you have an older iPhone (for example, X, 8, or earlier, you may not get the update (you cannot have access to future major iPhone updates))

[faint temple]

I’ve got iPhone 14 pro

[stone galleon]

How do you do these captchas for the verify you are human? I keep getting it and I can't seem to complete it, it keeps asking me to try again?

[onyx girder]

Hi I have an issue for the past 2 days I can't make ChatGPT Plus work anymore. I have the following message: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[faint temple]

Is version 1.2023.258 the latest?

[faint temple]

https://community.openai.com/t/error-something-went-wrong-if-this-issue-persists-please-contact-us-through-our-help-center-at-help-openai-com/166153

This should help

[olive nebula]

yes (1.2023.256 in android devices)

[faint temple]

But I’m on iPhone

Wait

Oh sorry I misread it

Ok so I don’t know why it’s saying it’s using older custom instructions

[olive nebula]

i guess you'll need to report the issue

[faint temple]

Ok

Where do I do that?

[astral quest]

Is there a way to make custom instruction profiles? Or do I need to save both texts in a text file every time?

[olive nebula]

https://help.openai.com/en/

[faint temple]

Ok

[faint temple]

Is it happening for u?

[olive nebula]

i've never used custom instructions before

[faint temple]

Ok

[olive nebula]

• i am an android + windows user

[astral quest]

Apple only ❤️

lol

[astral quest]

Well thats no bug, thats a feature

[faint temple]

How?

[onyx girder]

It doesn't really help, it just list a load of people having the same issue but not really anything to try to resolve? Am I looking in the right place?

[faint temple]

Maybe try #1070006915414900886

[astral quest]

You can't change the context of an existing chat if I'm not mistaken

[faint temple]

Ok

[lofty zinc]

why do I always keep getting this? I stg it's been there for weeks now, all questions results to this

and the help center, well walks on eggshells trying to explain this

[grave shell]

I don't understand who invented this captcha?
Am I supposed to get epilepsy from trying to solve it? I'm half an hour I can't access the GPT, solving "wrongly" everytime the "MINIMUM" amount of captchas (which is 5 x 10 images)
I need a GPT to solve your captch to use gpt-4? is this the new joke around?
seriously, can anyone raise this issue?
I never in my life saw a more difficult captch than that, even when going on Thor browsing of paypal they don't give you such hardcore
Serioualy ease the challenge, what is this joke

[runic quest]

How can i add images to gtp 4?

[plain tusk]

has anyone used the offline whisper version?

it works but for some reason it refuses to use my gpu and uses my cpu instead

UserWarning: FP16 is not supported on CPU; using FP32 instead
warnings.warn("FP16 is not supported on CPU; using FP32 instead")

any idea how to reconfigure it?

[rugged wing]

Can you all access the openai "usage" tab?

I can't access. I think this is an error.

[junior yew]

Hi. I have Dell E2 for image generator with credits but this image generator is not generating quality images

Anyone has any idea how to play around better?

[hoary seal]

wait a few weeks for dalle 3?

[waxen drum]

Can someone explain to me Chat GPT-4 sessions? if I close the browser, the session is ended, and a new session begins in the context of the same conversation. According to Chat GPT-4, it has no memory of prior sessions, but it has context... whatever the hell that means.

So, my question is, is it able to remember stuff in previous sessions or use the context to tailor its replies accordingly, or not?

The second question is, can idleness close the session even if I haven't closed the browser, if so, after how much time?

[zealous harbor]

I really could use help. I keep getting hit with the "Verify you're a human" tests. But hte tests are IMPOSSIBLE! I have tried many times. I pay $20 for a service I can't use because I can't pass these damn tests!

[winged ginkgo]

exactly. that is so f'd up. feels like they are using us as lab rats to gain info for their AI models

[trail cairn]

its not normal to get them at all and usually the reason is on the user end, browsers, extensions, settings, etc.

[rustic marlin]

the capchas are indeed very hard

blame the people abusing the system, not OAI

it is because of this capcha that the system isn't overwhelmed with scrape bots causing it to be "at capacity" all the time

[rugged wing]

Had the same issue with a diff number.

[trail cairn]

its also possible that they terminated the trial program on some regions or completely when they introduced the credit purchase feature

[rugged wing]

aw, shame..

[trail cairn]

even with the $5 trial credit people kept profiteering by selling the $5 trial api keys for $10 to non supported region users

[lucid ravine]

Is chatGPT down today? my plus subscription seems to not be working...

All I get from chat history is "Unable to load history".

[smoky pilot]

Is there anyone who knows about portrait generating?

@trail cairn, do you know about talking face generation?

[untold current]

I'm a lawyer and I have a lot of text/digital data of all sorts of cases (civil and criminal), laws, acts, statutes and various things related to law. I want to train gpt on my data to get more accurate and precise results. What shall be my approach? Shall I finetune or use embeddings? When shall you suggest either? I saw a few videos on youtube but wasn't be make sense out of it. I have searched online and found that some open source embeddings models are better than openai's. If I use those models for embedding my data, will gpt api able to query over that data correctly? I'm also little confused about embeddings as to what it is exactly. For example I type in some question related to coding or law, it is possible that not every type of situation or code for each situation was used for training the model. The model makes automatically makes sense out of the datasets. My question is, whether embedding the data will allow the gpt do the same? Like if I fed a case to to gpt using embedding, will be able to connect the dots. Please help. Thank you

[untold current]

I'd be grateful if someone could help. I have been pondering over this since a week but wasn't sure what to do.

[smoky pilot]

Use embedding.

[rustic marlin]

keep this in mind: https://openai.com/policies/usage-policies

Disallowed usage of our models

• Engaging in the unauthorized practice of law, or offering tailored legal advice without a qualified person reviewing the information

• OpenAI’s models are not fine-tuned to provide legal advice. You should not rely on our models as a sole source of legal advice.

• High risk government decision-making, including

• Law enforcement and criminal justice

[untold current]

Thanks for clarifying @rustic marlin, but I'm practicing lawyer with more than 8 years of experience. I'm not asking this question to provide or get AI generated advice to someone. I want to do this to help myself in studying a particular which I get from my client and to be honest it truly boring to pick a book from shelf everytime I want to study a case. This is to organize my knowledge and expertise to make my work easy

[rustic marlin]

as long as you are willing to have a professional using the AI as an assistant, always reviewing the outputs of the AI, it should be ok with the usage policies.
for your question about integration, it can be more tricky than that

it will require some good amount of coding experience to build a system that could trace the correlations between cases

[untold current]

I will never use give my client any advice that was generated by AI even if it reaches AGI (until the judge is AI too 😉)

[rustic marlin]

the main goal is to first, have a system that could execute the search on your dataset, then with that, you can feed only the relevant content to the AI so it can generate the text about it

[untold current]

Yes I want to know first hand, what I need to do (something like a road map)

[rustic marlin]

nah, that would be actually fine, the point is to just never not have a human in the loop

[untold current]

I understand @rustic marlin

[rustic marlin]

Your use case might not be something that is easily solvable with a language model though

[untold current]

So first I will need to create embeddings, correct?

Does embeddings need to be in a particular format (just like fine tuning) or it just needs to be fed to database using embedding model?

[untold current]

Will this involve any AI or just simple python script?

[untold current]

Will gpt api won't do correlation for me? This is one of important question in my original post

[rustic marlin]

im not sure, but I think you are over estimating the capabilities of the AI in this case

the AI can do correlations, but you have to first provide it the data, which is possible with small datasets that can fit within its maximum token capacity

a system like you described would involve a few steps
the first one would be to have a reliable way to execute saerches on your dataset, that could be done with embeddings, but also with just searching keywords across text documents

Then, having the search done to narrow down only on the important required text, which must be an amount of text small enough to fit on the GPT token limit, it will be able to hallucinate a correlation between the facts you presented to it. You then would have to fact check that output

[untold current]

Yes @rustic marlin, as you told me, providing data is through embeddings, correct? Or through fine tuning?

[untold current]

Wow now I have an idea of how it should be

[rustic marlin]

there are many different ways to do what you want, there is no single answer and picking one requires some planning

to explain in short, what GPT does is re-write text you give to it

fine tuning is for making it do this rewriting in a different way than the default without you needing to give it context every time

embeddings are a way to create a "semantic map" of data, allowing you perform searches

This was a very succinct and poor explanation, you really might want to look further into it

[untold current]

Wow @rustic marlin . I have understood everything. I will try to implement these using github repos and some of my knowledge of programming

[rustic marlin]

nice, happy to help =)

[untold current]

Thanks alot @rustic marlin for helping me and giving me your precious time. I found a github repo Somewhat similar to what we discussed. The name of repo is quivr by stangirad

[crude musk]

chat gpt down?

nothing is working

[vale stratus]

Status.openai.com says its fully functioning…

[uneven hare]

This "phone verification" restriction, combined with the blocked ability to update login authentication (https://help.openai.com/en/articles/4936824-how-do-i-switch-my-authentication-method) is giving me serious AI anxiety. I need to kill my Google Workspace account - but I originally signed up to ChatGPT with it. I'm looking up buying a burner phone, as that's preferable to dropping a phone number I've had for decades, but wow has this account policy trying to prevent abuse created an abusive moment for an honest user.

[trail cairn]

you can create 2 accounts with one number

[uneven hare]

unfortunately it tells me my number is already registered twice. I don't have another account and am positive I never tried to before. So, I don't know why it's showing two. But, from what I'm reading, the policy is the policy and every request to have an exception seems to be denied.

My biggest fear is buying a burner only to find out that number is already blocked or registered

[trail cairn]

🤷‍♂️

[uneven hare]

ha... I know...

[trail cairn]

prepaids are pretty cheap so its not a big loss if that happens

[uneven hare]

I know - a year ago I wouldn't have blinked. But 2023 hasn't been a good year so I'm pinching a lot of pennies.

I found a SOLUTION!!! I realized that while I can't close my Cloud Identity and keep my ChatGPT access, I was able to kill all billable services on Google (and keep the login identity.... Maybe I just needed to vent on it one last time

[trail cairn]

[smoky verge]

do we have a date for the realase of Dall-E 3 in Chatgpt for plus users ?

[trail cairn]

october

[smoky verge]

We should expect end of october no ?

[trail cairn]

"DALL·E 3 will be available to ChatGPT Plus and Enterprise customers in early October."

[smoky verge]

ooh i didn't see that

[wet plinth]

anyone having issues with logging into chatgpt? Just a blank screen? I have tried diff servers and cleared cache and cookies, any other advice?

[waxen drum]

how do i get free credits for Dall-e as a subscriber to chat GPT-4? I read I get 50 credits?

[quiet robin]

Hello, I've generated an AI logo that I really like, but I'm unsure how to edit the text to read 'Dynamic Edge Consulting.' I would greatly appreciate any assistance in creating a logo pack suitable for business cards. Thank you in advance!"

[elder palm]

legally acquire adobe illustator 😉
vectorize left side
find similar font to right side. reminds me of hind madurai personally
type it out
boom logo

or

Wait for dalle 3 in october and get working text

[rustic marlin]

that seems pretty easy to trace over on an image edditing software

and have a propper vector

you know... get rid of the AI artifacts like this

it might be better to use the AI as an inspiration rather than actually using it as the logo

[hardy anchor]

Hey, I'm trying to get a solid output in the same JSON format all the time. Any way to do this by settings?

I use prompt engineering for it, but maybe there's a more solid way?

[tribal talon]

custom instructions

WHich is in settings

[hardy anchor]

Thanks

@tribal talon How's that possible in the API?

[tribal talon]

If you're using API with chat completions, put expectations/instructions for the bot in the System message

[hardy anchor]

Okay, so no setting there?

[tribal talon]

idk what you mean with API and settings. System message is essentially the settings for your bot.

    {"role": "system", "content": "You are DiscordGPT. You sit in a discord channel answering questions. Make all replies short and use emoji often."}

Something like this will tell your bot how to reply. Put your prompt engineering for your output there so it knows what to do with each reply

[hardy anchor]

Yea that what I am using right now :/ Would be a good feature

[smoky pilot]

Is there anyone who knows about GANs?

@trail cairn, Do you know about GAN?

[trail cairn]

what about it? not really openai related

[smoky pilot]

Video generation, like synthesia.

@trail cairn

[robust sun]

Is the voice and images feature in the chat gpt app only for plus subscribers?

[rigid musk]

Yes

[trail cairn]

"Plus and Enterprise users will get to experience voice and images in the next two weeks."

[rigid musk]

You didn't read the #announcements

Like radium said, click on the link in announcement. All detail .

[storm summit]

Omg omg omg omg open AI you guys are giving me a heart attack

Let's go OpenAI ❤️

I need to breathe ugh

[late cedar]

is there going to be some sort of API for the new features?

[rigid musk]

Later

[late cedar]

thanks!

This might be a stretch to ask, but is there a general timeline for when that's going to come? Like soon later or later later?

[rigid musk]

No eta

[late cedar]

alright!

[bold turret]

buttonOpenAI is just not working for me today. There's a new home screen with the list of prompts and a login button at the bottom. When I click the login buton nothing happens. If I manage to log in through another part of the website, OpenAI is just a black page.

Same problem when I use another computer/browser

[smoky pilot]

@trail cairn, for embedding, do you use K-Means Clustering?

[trail cairn]

no

[dusky kraken]

Hi team. I’ve been a PLUS subscriber for many months now and never seem to get the beta features or be part of any releases. Is there a way to skip this jail somehow ?

[trail cairn]

like what in the past or now?

[smoky pilot]

I'll give you the whole requirement. Can you give me advice?

[trail cairn]

only if the answer can be determined in a minute, i wont analyze anything large

[smoky pilot]

OK. Can just a minute.

For security, can I send you DM?

[trail cairn]

no

you should not share anything like that type in dm anyways

[smoky pilot]

OK.

So, the requirement is like this. To create a chatbot for car database.

The car database has the selling car's list. And each car has its own url(where did I buy it) and has its own stock numbers.

The system should quickly identify product URLs provided by customers and retrieve all associated product information from the database.

[smoky pilot]

And each car has its own description. And the chat bot should find correct one once they receive the requirement.

Database size is 10G

[trail cairn]

that's just a simple search from embeddings, or possible hybrid search depending of the ui, etc.

[smoky pilot]

How come it's simple search?

[trail cairn]

well if the user is telling what he is looking for then its already a good base for sematic search

even better with selection bubbles

[smoky pilot]

Selection bubbles?

[trail cairn]

those that many chats use or even chatgpt has

[smoky pilot]

OK. Thanks. Can you tell me the whole flow of it?

[rugged wing]

That voice chat will work in English only?

[trail cairn]

there is nothing more to than do the search and use the best scoring embeddings

[smoky pilot]

I mean, if a customer sends a requirement to chatbot, then what shall I do?

[trail cairn]

that just narrows the search as more info is gathered

[smoky pilot]

So, what references should I read?

Selection bubbles, and ...

[trail cairn]

gpt inferring, embeddings hybrid search

[smoky pilot]

Can you give me some source code really similar to this?

[storm summit]

is the voice feature

[trail cairn]

i dont have any but there are #1037561385070112779 and github

[storm summit]

working ?

[trail cairn]

"Rolling out over next two weeks"

[storm summit]

......

is microsoft copilot using the multimodal feature

the one they showed where

it figured out a maths equations from a grapth

[smoky pilot]

@trail cairn, really thanks.

[next ruin]

Is the voice actor of Juniper the same voice actor that voiced Samantha in the movie Her?

[trail cairn]

scarlett johansson? did not sound like her that much, imo

[next ruin]

Was it Scarlett Johansson? I was just listening to the Juniper sample and she sounds a lot like her

[trail cairn]

yes, its a long time since i saw that movie but seen others where she is

[tired flame]

probably one people are thinking but now chatgpt can recognise voice can chatgpt4 be used to make an alexa replacement?

[smoky pilot]

@trail cairn, for inserting database to openai, what shall I use?

Can you explain in details?

[storm summit]

Sounds like her

[trail cairn]

insert your db? you call the api or some open source thing to create embeddings and store them yourself

[smoky pilot]

I mean, for 10G, for example, I want the chatbot that if I request the product url of a car that stock number is 00200, then the chatbot should reply '.....', like this. In order to make the program like this, how can I make it?

[trail cairn]

you havent even read or tried the first embeddings example if you are still asking that 🤔

[smoky pilot]

I've read. But can you explain about it? Maybe one or two keyword is enough, I think.

[trail cairn]

no, read that example as its been specifically made to explain the basics and includes code

[smoky pilot]

What you mean by the first example?

I'm not sure what you mean. Really. There are so many examples. But what do you mean exactly?

[trail cairn]

you even asked about it one point, the qa one

[smoky pilot]

I'm sorry. But you never answered in details. Maybe, it must be faster than you give only url.

[trail cairn]

#community-help message

you are not listening

[smoky pilot]

OK. Thank you.

Lexical-based search
Graph-based search
Embedding-based search
So, what shall I use?

[trail cairn]

read further

[smoky pilot]

So, in openai, the number of token means the number of alphabets?

[trail cairn]

no

[smoky pilot]

Can you give me the answer?

1000 tokens, it means?

[trail cairn]

you need to go back and read the openai docs as everything is explained there multiple times

[smoky pilot]

OK. Let's study QA first.

[trail cairn]

if you dont even know what a token is it makes no sense to skip to embeddings

[ancient kiln]

Anyone missing the “new features” button on their app and the website? It’s a plus account.

[rigid musk]

#announcements

[ancient kiln]

Yes I’ve watched that. Never read anything was missing.

[rigid musk]

We’re rolling out voice and images in ChatGPT to Plus and Enterprise users over the next two weeks. Voice is coming on iOS and Android (opt-in in your settings) and images will be available on all platforms.

[smoky pilot]

@trail cairn, OK.

[ancient kiln]

I have read that. Just confused since the new features button used to be visible. Thanks

[flint ore]

everytime i try to access site from computer, it only shows this

[ancient kiln]

Do you have plus?

[flint ore]

i dont have anything

[rigid musk]

Try firerox as test

[trail cairn]

check your browser error console for cert errors

[flint ore]

this stuff?

[trail cairn]

click one of those

seems something on your end is blocking/disconnecting

try what neko said

[flint ore]

Ok works

A bit strange how that just happened

[sharp wasp]

is whisper out?

[trail cairn]

yes long time ago

[sharp wasp]

is there any tutorials for it in java

[trail cairn]

https://platform.openai.com/docs/api-reference/audio/createTranscription?lang=node

[sharp wasp]

emm

thats not java

unless im mistaken

[trail cairn]

if you mean pure java and not javascript then you probably wont find any, could still try your luck with github or google

[sharp wasp]

rip - i just need some sort of way to turn my voice into a string 😭

[trail cairn]

java is not really good or common choice for such things nowadays

[sharp wasp]

so true 😭 i made this and wanted to remove all text and make it work entirely from my voice

[trail cairn]

if its not for a school project then why not try and learn some other language?

[sharp wasp]

ehhh

i dont have a gpt api key 😦

[trail cairn]

github has open source version of whisper

[sharp wasp]

i mean

i wouldnt know how to get ai on python without chatgpt library

i only know how to do it on java 🤷‍♂️

[trail cairn]

why without?

[sharp wasp]

bc i dont have a api key

[trail cairn]

anyone with openai account can create one as long as you pay for the usage

[sharp wasp]

yeah but i want it to be free

[trail cairn]

then your only option is to use the open source projects and models which some require beefy computer

[summer bay]

am i at the right place to find support for OpenAI Gym

[trail cairn]

nope, its not maintained anymore

[summer bay]

you just can't use it anymore?

[trail cairn]

most people moved to the other project that replaced it but its not by openai

[summer bay]

may i ask which?

[trail cairn]

it was listed in the repo or somewhere, they had their own discord too

[summer bay]

I'll figure it out, thanks a lot

[rapid ermine]

Anyone else having issues with their usage page not loading?

[empty ingot]

Just to confirm, for Plus users that do not have access to new functionality yet, the "New Features" button in iOS in Settings is now gone, right?

[tired hamlet]

Guys, I bought a subscription, but I can't connect the plugins! How can i?

[storm summit]

Did it get removed ?

[timber shadow]

Go to your account settings in ChatGPT, then Beta features, and enable the Plugins model.

[primal rock]

This is true for me

[storm summit]

?

[tired hamlet]

I see this only

screenshot from website

[west panther]

Open the GPT-4 plugins mode and go to plugin store

Not the chatGPT settings

why the sc so dark af

[tired hamlet]

blm you know

[west panther]

where is that toggle ui lol

[tired hamlet]

idk

im start crying in 3..2..1..

[timber shadow]

You can't currently use Beta models in the app, only on the website. It'll be in the GPT-4 dropdown menu on the site though.

[storm summit]

Is the voice available or not

[rustic marlin]

this dark mode is so dark that im surprised you can see anything at all

[tired hamlet]

the problem is, its a website

[rustic marlin]

just for good measure, turn off the extensions messing with the page

[tired hamlet]

nah its just more comfortable for eyes

[timber shadow]

It's wild that it affects your screenshots like that though.

[west panther]

Maybe try logging out and in?

[storm summit]

Why does she have it?

[tired hamlet]

ill try

[rustic marlin]

yea, I mean, jsut turn them off to see if the buttons that should be on the settings show up

I too use this sort of extension

[tired hamlet]

i turn off all plugins that i have in firefox

but it still

[tired hamlet]

oh

I love you Lugui

ill name my sun lugui

i see it

[rustic marlin]

happy to help =)

[tired hamlet]

i love living this life

[tired hamlet]

[west panther]

Hey, you know what? In this picture the settings did had the toggle button

[tired hamlet]

OOH i see now

Sorry guys

[empty ingot]

Plugins has always been on since I've had plugins. We have a couple of accounts and this is true on both of them (I am referring to the web interface since Plugins does not show up on the iOS mobile app).

[empty ingot]

It is true that this is gone or it is true that this is present? Sorry, your response is ambiguous to me.

[west panther]

It's gone for me too, but it only had the disabled browsing with Bing option so Idc

[primal rock]

My apologies. The new features category in settings is gone.

[empty ingot]

Got it, yeah, so now it's a sad waiting game. Lol, I need to diagnose some plant diseases.

It seems like this new functionality might be a huge time saver for some tasks that are normally enormously challenging as far as categorizing and searching goes.

[tame tiger]

Anyone familiar with this error? I have been getting it for days!!!! "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[rigid musk]

Test with firefox

[dim pond]

Anybody has good coding Custom Instructions perpherbly JavaScript?

[storm summit]

OpenAi I just want to say I love you guys so much 🙂 💖😘💖

[spiral fossil]

glazing hard

[thorn ore]

I've tried and tried over the last several days to use Chat GPT with no luck. I had previously used it for about a month without any issues or problems. I have tried everything, and even upgraded to Chat GPT 4. I have cleared my browsing history, tried using another browser, and even tried creating another account. This is the message I keep receiving: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." Can anyone help?

[thorn ore]

me too, I'm hoping to find answers here. 🙂

[thorn ore]

do you know why it would vary by browser. Thank you for your help! 🙂

[rigid musk]

you might have somethign block with chromium

[thorn ore]

Thank you. It did work!!

[loud apex]

Does anyone else have this message when accessing the chat?

message:
Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts. ChatGPT September 25 Version

[somber grotto]

yessssss

are u able to access?

this is my screen when I click on chat gpt

@loud apex where u able to access?

[loud apex]

my screen is the same

I had been without internet for a while. When I started using it again, it got this bug

[remote terrace]

I get this error on everything i search on DALLE can anyone help me

[inner belfry]

I'm stuck at this page.

[arctic tundra]

Is there any update on Musenet status 🥲😔

[sinful quarry]

wth is even that?

[wild robin]

May I ask questions about Dall-e here?

[onyx kraken]

I think so! #images-discussions might be the perfect place to ask!

[wild robin]

Hi, I am a newbie and I have some questions about Dall-e
I paied $15 on Dall-e to use its AI art generation function. The images generated were very basic and poor quality.
I then discovered that MS Bing is powered by Dall-e
The quality there is much better.
Do other people have similar experience?

[regal lynx]

In my experience, Dall-E 2.0 (labs website) is more difficult to get "eye pleasing" images out of compared to Bing's image generator because it's an older model. With that being said, I think Dall-E 2.0 is a lot more fun to work with 🙂

[wild robin]

Thank you, Dinasaur Oatmeal! Someone said in the other channel that MS Bing is using Dall-3 now! Wow, I thought Dall-3 won't be out until October! Maybe this explains the difference!

Seems I paied $15 in vain!

[brittle summit]

Does anyone know if Dall E or ChatGPT does logo recognition?

[bronze charm]

Should be able to when we get access to image input

[brittle summit]

Thanks.

[shadow dock]

Id like to open a support ticket

[bronze charm]

Contact [e-mail]

[shadow dock]

via email?

[bronze charm]

Yes

[shadow dock]

Well I have a pretty straightforward question

[regal lynx]

you can ask here and see if someone can help

[bronze charm]

^

[shadow dock]

I am paying for GPT4. I am using the UI on chat.openai.com. Am I going to be charged more than 20 dollars a month based on my usage?

Its to my understanding you are charged based on your usage for certains tiers of it

[regal lynx]

if you have a ChatGPT Plus account, you won't be charged more than $20/month for how often you use GPT-4 through https://chat.openai.com/ . However, the API is billed separately, and you'll have to pay to use it

[shadow dock]

how do I know which Im using?

[regal lynx]

are you using the website I just linked?

[shadow dock]

I access GPT4 through chat.openai.com

[regal lynx]

you won't be charged extra for that

[shadow dock]

And I am only seeing 20 dollar charges to my credit card

Okay perfect. Just wanted to make sure

[regal lynx]

that is correct. That's the price for ChatGPT Plus

[shadow dock]

Why is there 2 names for ChatGPT Plus? (1. GPT4 2. ChatGPT Plus)

[shadow dock]

thanks

[regal lynx]

ChatGPT Plus is the subscription service name. GPT-4 is the model that ChatGPT Plus allows you to access. You can access the model outside of the subscription service (via the API on the playground website)

[modern umbra]

Hi, is it true https://chat.openai.com/ is down today?

I got this page instead:

Free Research Preview. ChatGPT may produce inaccurate information about people, places, or facts. ChatGPT September 25 Version

That's all, without the usual chatbot functionalities.

[regal lynx]

ChatGPT is working for me without an issue

[modern umbra]

Is it possible I experience these problems, because they are currently rolling the multimodal version of ChatGPT?

[regal lynx]

I can't confirm nor deny that as I don't know what the backend of OpenAI looks like (I don't work for or represent the company)

[onyx kraken]

ChatGPT+ is like a subscription.

It gives you a few neat things

1. More access to ChatGPT 3.5 than you got for free (more tokens in the same time period, plus faster output shows up, so you wait less)

2. Some access to ChatGPT 4.

3. Some access to beta test or enjoy earlier access to new features.

4. Whatever else, those are the 3 I focus on 😛

GPT 4 is a model of AI, and it may have more than one version itself. It's the model that ChatGPT-4 uses (not sure what version) and you might run into it through a few other services, like Bing is said to use a version of GPT 4.

However, there's also the playground, with its API access. And that's not the same as your chatGPT web access, where you have plus. But it's another, totally different way to access some of the AI models, and you pay per use. I've never used it myself. But at least one version of GPT 4 is there.

[jolly moth]

Because in summary, those are different models of ChatGPT

[shadow dock]

I see

Very helpful. Thanks for all the insight

[regal lynx]

To add to this very neat explanation, the API lets you alter a few settings like temperature of the model, output length, etc.

[onyx kraken]

I wonder, if we're very, very creative and persuasive, if we can get similar effects just through our prompt-tastic powers of AI whispering!

[regal lynx]

similar effects as ChatGPT? I assume so because you can input a system message to the API

[onyx kraken]

As if changing the temp and output length? Hehe!

[regal lynx]

that's what the GPT-4 API Chat mode looks like. It's a pretty neat UI with easy-to-use sliders

[onyx kraken]

That is nice. And they look easy to use! But since some of us just use the web interface... I still wonder if we can persuade the model to just act like we shifted those sliders and amuse/please us 😛

[regal lynx]

I know some people have tried to get ChatGPT to output a specific number of words before and had minimal success. That may be changed with this brand new update, though!

[onyx kraken]

I wonder. Does it cut the AI off mid-word if it tries to use too many?

I wonder if Advanced data analysis could be set to do the same thing to the model's output 😛

[regal lynx]

I'm not sure to be perfectly honest. I rarely use ChatGPT 😅

It is rather ironic considering my role on the server, but I'm a Dall-E fan at heart

[grizzled gazelle]

Hi all, does anyone know when the new vision GPT models will be available for API use? And at what price point?

I have an app that uses some other 3rd party vision api, and I would love to try GPT's instead.

[regal lynx]

we don't have any details outside of what the announcement said

[steady cedar]

How can I get access to the GPT-4 API?

[regal lynx]

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

[cursive mountain]

I keep ending up with this page when I attempt to login to chatgpt. How can I login to chatgpt like I used to before?

[onyx kraken]

Some people said when they reset their modem it was okay for them again.

[visual jasper]

Hello, new to the server here (but not that new to OpenAi), and I have a question.

I was wondering if it is possible to have a standalone/offline version or model of ChatGPT to use, without having API calls.
For example, to use in a Unity game?

I have some ideas that I want to explore for small games, and hopefully release them, but I don't want the game to cost me money every time a prompt is requested. So I was wondering if it is possible to integrate a version/model of ChatGPT into the game itself?

Example of a game would be (but not one that I would actually make):

You ask a question (type it) to the magic 8 ball, shake your phone and it gives a vague answer.

[bronze charm]

OpenAI's LLMs are too large to be ran locally (if they gave access in the first place)
I would look at some open source models on HuggingFace
You could also ask ChatGPT for alternative courses of action regarding this situation 👀

[visual jasper]

I feared that size might be a problem, I'll look into alternatives then 🙂

[late snow]

hm question on the upcoming rollout where chat gpt can speak , does it apply only to the mobile app version?

[onyx kraken]

Appears so, announcement says ios and android for that portion. But we only know what the announcement says

[cursive mountain]

Thank you!

[lethal panther]

GPT-4 is rumoured to have about 1.8T Params. It doesn’t use them all at inference thanks to mixture of experts architecture, but goodluck inferencing several hundred billion parameters 😂

[primal robin]

Hi guys sorry about this newbie question but does the gpt-3.5-turbo-16k-0613 support function calling or that’s just the **non 16k **version ???

[cyan jewel]

I tried, it doesnt work.

[untold current]

Does oepnai provide support on discord? I'm not able to select Google play balance for subscription in chatgpt android app even sufficient balance is present. Has anyone faced similar issue?

Whenever I click on Google play balance, nothing changes and the previous selected methods is unchanged.

[rigid musk]

Not on discord. Also I don't think they troubleshoot Google payment. Anyways if you want to discuss it: [e-mail]

[untold current]

I have contacted Google support and they told me to Contact openai.

And to be honest I dont think openai support will reply to my query before two to three months

So thought it would be better to ask here on discord if someone faced similar issue

[rigid musk]

You can always post in #1070006915414900886 .

[untold current]

Will do for now. Thanks Mate @rigid musk

[rigid musk]

Have a good day

[untold current]

Have a good day partner

[white osprey]

good day

remember y'all to touch some grass

[fierce haven]

what's grass?

[white osprey]

sorry mods that's offtopic

I know

[fierce haven]

never seen before

[smoky pilot]

Is there anyone who can explain about accuracy of embedding in openai?

@trail cairn?

Who can tell me the accuracy?

[trail cairn]

be more specific

[wooden cave]

I had this. chatgpt links to oaistatic, which was blocked by my custom dns, resulting in chatgpt not loading properly. that is probably why and you can check this by checking your browser development network logs. either use a different dns server, or unblock that domain on your dns.

[plain tusk]

Anyone having the same problem of openai rejecting your card? I'm using a visa card which works with pretty much everything (amazon, google, elevenlabs, etc) but not with openai api

[rigid musk]

It is a real credit card or a virtual one

[plain tusk]

real

[rigid musk]

It is your credit card or someone else

[plain tusk]

mine

[rigid musk]

Which country ?

[plain tusk]

japan

my current country of residence

[rigid musk]

Here what you can try

1. contact your bank make sure they doesn't block the transactions
2. try with a different computer or cellphone see if that's works
3. if none work you can contact [e-mail]

4. you can always post in #1070006915414900886

[plain tusk]

hmm alright, thanks!

[inner belfry]

Any idea on when ChatGPT would be readily available again?

I still can't access it.

[jovial surge]

@inner belfry you have secure DNS enabled through OpenDNS. They have issues

[rigid musk]

?

[jovial surge]

Do not use Secure DNS with OpenDNS, use cloudflare

[rustic marlin]

The ChatGPT page is a huge target for scrapers to get free AI. So, because of that, the service will be very picky with anything that modifies the normal flux of web requests. So, overly intrusive filters, extensions, antivirus VPNs, and any sort of thing that acts as a middleman between you and ChatGPT will very likely trigger the suspicious activity detection, causing you to get more captchas to solve or straight up be blocked from it.

[twin palm]

Any reason why this happens alot? I use Opera's nav bar feature which has chat gpt on it, ChatGPT works fine on that but I cant log in whenever I open chatgpt on a new tab.

[tacit garden]

why is it not running??

nvm I figure it out

thank you

[smoky pilot]

OK. My database has more than 10M products. And each product has about 40 description in details. For example, id, production date, style, factory, ... And if I use embedding, the accuracy could be fine?

[arctic holly]

Do you all know when the option to turn on chatgpt voice will turn on for iOS ?

[dreamy plaza]

that's what I've been having since yesterday, I've sent in a message with what I've been seeing on the bot so hopefully they'll fix it soon, thankfully the app still works

[smoky pilot]

@trail cairn!

Is there anyone who can help me?

[peak schooner]

How did you figure it out?

[arctic holly]

For ChatGPT Plus users on iOS when will the feature for ChatGPT to speak to us in a voice be available to turn on? https://openai.com/blog/chatgpt-can-now-see-hear-and-speak

[idle socket]

Does anyone have the new multimodal feature yet?

[arctic holly]

What does this feature do?

[tacit garden]

it is the ip that is associated with, I just change it.

on my end

[timber shadow]

All we know is within the next 13 days.

[steady cedar]

When will GPT-4 32k be available? Was it made purely for advertisement purposes?

[regal lynx]

I don't work for or represent OpenAI and don't know when it will be publicly available. I do know you can access it with an Enterprise account, though