community-help

[river cypress]

Does Whisper write to a file or does it just print the text output? e.g. does whisper japanese.wav --language Japanese write to a subtitle file?

[toxic nest]

I've been getting that same error message ever since I tried using ChatGPT, anyone know how to fix this?

[devout lotus]

Question on plan wise. Is chatgpt different subscription then the open ai account?

[rigid musk]

?

[rugged wing]

It's time for HYPEEEEEEEEEE!!! CHATGPT IS NOW ON ALL DEVICESSSSSSSS!!! YEEEEEEESSSSS

[rigid musk]

i suggest to avoid all cap

it can lead to mute

[regal lynx]

turns blind eye because I'm an Android user

[plain kayak]

still no plugin in ios app?

[rugged wing]

becomes super Saiyan lol.

[rugged wing]

Im and android user... This is MASSIVE NEWS.

Oops.

[rugged wing]

Apple vs Android my dude.

Apple= Restrictive

[unreal hemlock]

<:book_icon:1046291503158399107> `` Rule 7 `` No self-promotion, soliciting, or advertising.

Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API Channels

[rugged wing]

Uhhhh.

[cerulean cairn]

For real. Do you need to sub to use it after a bunch of query's?

[rustic marlin]

ChatGPT is free, it has a paid subscription, but the free version is still enough to use it a lot, really a lot before any rate limitings

[cerulean cairn]

hasnt been working for like a day and a half

[rustic marlin]

it is working, that might be something on your side

clearing cache usually fix issues on chatGPT

some times, jsut a ctrl + F5 is enough

[cerulean cairn]

cleared cache and the other command nothing...

working in avast browser. cheer dude :))

[rustic marlin]

I was going to suggest that, trying a different browser

if it works on anothe rone it is for sure the browser's cache

[indigo chasm]

I got "The previous model used in this conversation is unavailable. We've switched you to the latest default model." today on paid version... I just got paid version yesterday. Can someone help me understand and/or fix?

[cerulean cairn]

working great! does the new updated language have statistics more recently. I woulld like it to run monte carlo simulations for me.

[rustic marlin]

nope, still on 2021 data

[cerulean cairn]

ok

[jolly monolith]

QUESTION - Hi. The Custom Instructions... instructions say that you no long have to repeat "I'm teaching a 3rd grade science class" because GPT will 'remember' and understand. Is that remember per session? Sometimes I have to switch subject matter areas and it wouldn't do to have one 'memory' write the response for another prompt. What the scope of the Custom Instruction, and how do you 'turn it off'.

"ChatGPT will consider your custom instruction for every conversation going forward". This is actually very unhelpful in my case, where subject matter expertise shifts are needed.

[keen linden]

hello does anyone know the support email of open ai? because i was banned from using their services because tthey said i did something wrong that breaks their policy but i am just a student front end web developer and i am relying on their service when i am building websites like for now i am building react websites and i really need the support of their ai.

[plain kayak]

if i'm understanding correctly, you want to turn it off for another session right?

[jolly monolith]

Yes, actually. It would be really unhelpful if it constantly acted like one professional or another. I'm trying to sort out if the scope of Custom Instructions is one session, and in a new session, you could come at it from another pov. (It may be that this isn't made for that kind of use. That's an acceptable answer.)

[plain kayak]

@jolly monolith

[jolly monolith]

Ah, thank you, @plain kayak. That's kind of you.

[rugged wing]

Hi lugui!!!

[rustic marlin]

Hello õ/

[cerulean comet]

Aaaa

[native ether]

Stop spamming.

[cerulean comet]

Aaaaaaa

Okay fine

[keen linden]

theres nothing i can do with this? because its linked to my mobile phone number and i cant create another account

[rugged wing]

I'm depressed AF...

It's been a terrible terrible day today...

[rustic marlin]

=/

[golden socket]

ChatGPT is not Connected to the internet

[rugged wing]

That's why bing is better

[wild dagger]

Hello. How can I use the Web Browsing feature? I heard that gpt 4 has this feature.

[main karma]

does anyone know if OpenAI has said anything about extending the token context window for the GPT-4 model used by chatGPT Plus? I am finding the context limit to be a show stopper for me. Its magical to use it as a partner on worldbuilding projects until the moment it starts forgetting stuff from earlier in the conversation 😦

[cunning spire]

yes cypcherLH first question

[loud spade]

Hi, I don't know if I have spammed the thread, by asking question almost everywhere, the question is, why AI Text Classifier not working

[cerulean wadi]

I'm not sure if this is the right place to ask, but I created my own programming language, and I would like to build a bot that is able to generate code in my own programming language. I guess in context learning might not be ideal since it will use up a lot of tokens, is there perhaps another way todo this? Maybe finetuning the model, or with embeddings? Any help would be grately appreciated 🙏

[vestal wharf]

anyone know any free headshot ai generators?

[charred parcel]

ask in #ai-discussions

[whole dock]

Does anyone know how to prevent LLMs from answering questions that are out of context.
Context: I am trying to build a customer support chatbot and i want to prevent it from answering questions if a user asks questions that are irrelivent.

[lost dust]

You can try Avoid discussing topics, which are out of topic. When the user/customer asks about out-of-topic question then answer with "[place your custom answer here]"
This is not guarantee to work in 100% of time.

[junior ore]

How can I get the OpenAI API to respond in the language that the user wrote in?

[charred parcel]

yes, you can be banned for disallowed content

[sour seal]

Does gpt 4 support image input

[charred parcel]

not yet

[sour seal]

I see thanks

[onyx kraken]

You can also edit it at any moment, so you could edit it between classes or whenever else you pleased.

In your situation I might use different 'new chat windows', and check to make sure my CI said what I needed it to for the various chats I had (or make a generic 'I am a teacher one' and just control the subject dynamically. Once the AI starts a pattern it's usually pretty willing to continue it and the subject matter you're discussing is a decent ongoing reminder of what the subject is this time.

A neat example of that, my CI supports styles of interaction and general goals. A specific chat that I posted about in humor, #1070955045639684096 message it involves an AI created modified Hogwarts theme that isn't in my CI and wasn't specifically requested by me.

However it follows and furthers the storyline excellently, both inside that one chat, and when I want to ask it to help me with the lessons and puzzles inside that first game but don't want to risk derailing that game, so I use a new chat and paste part of the story into the new chat, it stays in theme excellently, even though it made that theme up based on a few choices and what's in my CI, all it had to see was part of its in-character interaction with me and it was ready to work on that focused part without any further guidance.

[onyx kraken]

You can also give it a context that it can understand which makes not going into irrelevant topics very important. To compare, it understands not revealing passwords, and will try to avoid doing so if asked to treat some information as a password (it's not perfect at that, but it will try quite hard, including following requests of how to handle questions about it)

If your role for it makes entertaining or cooperating with the customer seem like a reasonable thing to do, it'll surely do it. If you want it to act with more focus and less cooperation, give it a setting and role where it knows it has goals that make that pattern wrong.

[fleet umbra]

You could try with just instruction sets, but I think most "censorships" are being achieved through another layer of software.

[nocturne hinge]

hel

helo

[rustic marlin]

yep

[rose glen]

this feature has been temporarily disabled while they work some things out.

[wild dagger]

Got it, thanks)

[minor jewel]

this stupid app is getting unuseable

[wooden rapids]

how so

[rose glen]

This channel is for OpenAI products, please use #off-topic for this sort of thing

[storm summit]

OpenAI Can you guys add maybe like a counter so we know how many messages we have used or some type of cool down for example. 50 messages per 3 hours. You can have a message counter out of 50 that has a estimate time. And the message counter goes up by 1 every 3.6 mins that equivalent to your 50 messages per 3 hour timer. Can't you guys do that its easier for us to see how many messages we have.

And a lot more intuitive than it blindly telling us we've used our message limit. That way its virtually the same yet far far better

[rigid musk]

#1070006151938314300

[storm summit]

Do i just paste that there

[ebon hull]

I need help changing the phone number linked to my ChatGPT account, and I'm trying to do that via the web interface but the option is not showing up for me; how do I change my phone number? 🥺

[charred parcel]

there is no option to change your phone number, you should head over to https://help.openai.com as we can't help with account level issues here

[ebon hull]

Yea but like how do I navigate the website so that I can change my phone number? I don't need someone to change my number here I just need to figure out how to navigate the help website plz help 🙏

[ebon hull]

^^

[charred parcel]

it can't be done on the website, there is no option provided to self manage your linked phone number

[ebon hull]

Ok so how do I reach out to someone who can via website the option is not showing for me either???

[charred parcel]

you should head over to https://help.openai.com/ as we can't help with account level issues here

[ebon hull]

Ok but how am I supposed to use the website if I don't know how to use it? I just need guidance on how to use the freaking website can you not help with that either??

[charred parcel]

what I am telling you is that there is no option to self manage your contact details on the website, it requires someone from OpenAI support to change details for you. thus help.openai.com is where to go

[ebon hull]

Ok so how do I reach out to openai support using the website? I don't see an option for doing that on the website??

[charred parcel]

go to https://help.openai.com and use the chat bot to get help

[ebon hull]

Yea but the chat bot isn't giving me an option to reach out to Openai support about my phone number, that's what I'm stuck on lmaoooooo

[trail cairn]

https://help.openai.com/en/articles/6613520-phone-verification#h_a72f9b11cc

[ebon hull]

Wait so I can't change my phone number at all even with Openai support team's help, not even if I deleted my account??????????

[trail cairn]

well deleting account means you could just create new one right now or keep using the existing one if the number is not needed elsewhere

[onyx kraken]

https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[trail cairn]

delete will permanently lock up both email and phone number

[onyx kraken]

the 'send us a message' button though that bottom right interface on that message looks like perhaps your best option to get ahold of openai customer support.

Now, far as I know, everything Radium is telling you is correct. But you are asking how to get ahold of them, and that's the best answer I know of.

[sweet hemlock]

Does anybody know when custom-instructions will eventually come out to the UK and EU?

[onyx kraken]

Not yet! They just became available to anyone yesterday!

[ebon hull]

Ok thank you sorry I was just getting a bit frustrated at being walked around in circles, may I know if this means that I did the process correctly?

Sorry I just forgot to censor my email earlier lmao

[stiff wren]

How to improve math using ai tool

[ebon hull]

Damn I see ty for info

[onyx kraken]

To my best guess, that looks fine. I have never needed to try to contact OpenAI customer support. However, from what I've read and noticed others talk about, I've heard mention that it says it takes about a week. Some have said for them they didn't get a reply long after a week.

This is what they say about phone numbers, from the 'deleting your account' help file:

I put the red box around what applies to you, possibly

It sounds like you have a different phone number that you want to use.

I would actually recommend create a new account (what they refer to as 3rd unique OpenAI account) using that number, and then either don't use the first with the number currently on it, or delete the first with that number on it.

It is not clear that the functionality you want, to change the number, exists in their system.

I'm not the one telling you 'no' 😛 I'm just reading the helps and listening to your apparent goal, and saying 'this seems the only workaround I notice.

And I suspect that their stance of ' We don't support unlinking a phone number from an existing account also means they won't unlink one to 'change' it either.

[ebon hull]

Worst case is that my mom and I have to use each other's numbers for the both of us to use Openai which is the awkward situation I'm trying to avoid but yea you understand my intentions correctly tysm for the info I'll wait and see what the final word from openai is; I personally think it'd be very dumb if changing phone numbers wasn't allowed as people change phone numbers for legit reasons all the time so yea. But yea tysm again and we'll see how it goes 🙂

[onyx kraken]

Wishing you the best. It sounds like potentially you have one or more work arounds to allow both yourself and your mom reasonably controllable access to the AI, even if it may not be possible to have it exactly as you'd prefer. I don't know, but I imagine there's some key and vital reason why they have that concern about changing phone numbers. Good luck!

[onyx kraken]

Hearing your intention though - I suggest do not delete the existing account at this time. That seems likely to 'lock up that phone number' preventing further accounts made with it. Meaning yet another 'never linked to OpenAI accounts' number would likely be needed before whichever of you could get another account, if you can't just share the same phone number (2 accounts per number allowed).

[ebon hull]

Ok tysm I'll do that then; you've been more than helpful; wish you nothing but the best moving forward 💜💜💜

[torn surge]

lets say i am making am making an API that prompts an AI model( that its not chat based so it doesnt remember previous conversations) and gets back a response,now Lets say a part of my prompt is always the same for example :

prompt 1: Imagine you are a model with white skin and oval face shape and tall, what type of make ups suit you? |
prompt 2 : Imagine you are a model with white skin and oval face shape and tall, what type of clothes would suit your height? | 
prompt 3 : Imagine you are a model with white skin and oval face shape and tall, what type of haircuts would suit you? |

as you can see the prompts always have "Imagine you are a model with white skin and oval face ship and tall" and they need to have it since its not a chat based ai,
but this approach is expensive because it has a longer prompt length so what can i do so the ai already knows It should imagine that its a white skin oval face shape and tall model and then i just need to prompt it the rest of the message and save prompt length and money?

basically is there a way to pre configure an all purpose model like gpt 4 or gpt 4.5 turbo to do this and save money on its API with shorter prompts?

[trail cairn]

expensive? have you calculated how insignificant cost the sentence has even if you ask it a million times 🤔

or is that just an example and your real use case is a paragraph or more?

chat models do not support fine tuning yet and not sure if it would even help

[wispy fable]

[prime lotus]

Is this a good place to ask questions about ChatGPT?

[trail cairn]

yes

[prime lotus]

Thank you. I received an email from Open Ai about my violation of the content rules. How do I work on that? Most were trying to adjust the query to find something they would accept. Not trying intentionally to violate the rules.

[onyx kraken]

I'd start by making sure I know the disallowed content policy: https://openai.com/policies/usage-policies

[trail cairn]

if you tried to recognize or analyze filter patterns to evade them then your ban is pretty much permanent

[onyx kraken]

If you are only warned or something, I'd cooperate and comply, and be very careful not to do anything that was against the rules

[prime lotus]

That sucks. It is funny how they prioritize the policy. I made a picture about GAY BINGO, yet they would not let me create a pic on the Bible. I run an online newspaper. I need to create certain content to purpose a news article. But a constantly being told that it is a 'sensitivity' issue and we wouldn't want to offend people. I live in LA. We are quite visible with the LGBTQ+ collective, But I can't use chat to write articles.

[rugged wing]

There is no option for internet access on my settings even though I’ve purchased plus subscription. Does someone know how to fix?

[trail cairn]

browser with bing is temp disable but you can try one of the plugins like webpilot

[charred agate]

will custom instructions ever be made available in uk?

[trail cairn]

yes

[charred agate]

nice to hear

[analog shore]

Hello, how to use custom instructions effectively?

[twin grove]

Probably someone who works for Open AI like an administrator can do a Reset MFA on your email account.

[onyx kraken]

Much like any other prompt!

In general, good prompt engineering is

1. pick any language you know really well that the AI understands too.

2. understand exactly what you want the AI to provide.

3. explain this, using language as accurately as you can, avoid typos and grammar mistakes and communicate clearly as possible.

4. check the output carefully, verify you get what you intended. Remember to fact check, and be extra careful with any math, sources, code, or other details that the AI is known to be especially likely to hallucinate

The difference with custom instructions (CI) is the AI remembers them 'first', as well as the recent conversation.

If you want the AI to act like a DM, tell it in the CI, it'll never forget. Want it to call you Joe - tell it in the CI. Want it to answer you with graduate-level language when biology is being discussed, but beginner level language when computer science or programming is the topic? That's great for the CI.

Want it to give you shorter responses as a rule, or long elaborate ones? That's for the CI too, you have 3000 characters of instructions to work with!

And you can change them at any moment, and turn off the CI as well if wanted.

[ebon hull]

Yea hopefully something like that will work; I've already reached out to openai support so I'm just waiting to hear back from them for now 🙂

[rugged wing]

. . . Eeeeeeeeee!!!

Hi everyone!

[misty island]

Why is the "Clear all" button no longer available?

[trail cairn]

its under settings

[clever path]

I am desperately looking for a way to enforcing GPT's output formatting.

• How can I get the attention of the openai's staff in that regard ? I about enforcing the format by only evaluating the tokens that match - say a specified json schema
• Will fine-tuning work quite well without being expensive ? (I think of a very high learning rate but I'm not 100 % sure)

[trail cairn]

are you using api and gpt-4?

[clever path]

yes

None of those work perfectly :

• Preprompting
• Functions call feature (actually very disappointing in that regard)

[trail cairn]

then there is nothing more you can do than try to instruct or steer it with the system message

gpt-4 does not yet support fine-tuning and you cant really reach openai staff to change anything

[clever path]

damn... It's to bad. This seems like a really easy thing do, technically speaking, when you have the best LLM of the world, to enforce formating

Also I'm pretty sure a loooooooot of developers do formatting

[trail cairn]

if it requires re-training its not an easy thing

[clever path]

Let me be more specific :
The idea is
For example in the case of a JSON output - with a JSON schema provided - evaluate only the tokens that would fit the schema as the output is being generated

[trail cairn]

did you also provide examples next to the schema

[clever path]

no it's not about examples, just a schema

[trail cairn]

example data, example of expected result

[clever path]

if you mean how I do it currently, no I haven't. But it's not my point after all

It still won't be steel reliable

[trail cairn]

then you have to try some other approach, eg. different format

[clever path]

there are many approach... including retry when the format is not correct

[rugged wing]

Hello. When appear ChatGpt on Android?

[onyx kraken]

Next week, whee! #announcements message

[crystal thistle]

You can do it yourself parsing the output of gpt-4

[clever path]

my point is to propose the idea above, to openai

[crystal thistle]

Yes I understand, but there are many services made using gpt-4 , this could be one. They are working only to provide the LLM, and it seems always like to add some text around responses

[gaunt plinth]

Can you please add a search function to chatGPT it will be very useful and great! I don’t understand why you doesn’t already put it! vote for this question like that, we have more chance to be viewed by the openAI dev team

[onyx kraken]

Would help to put that in #1070006151938314300, where someone might see it!

[gaunt plinth]

Oh yes thanks 😂

[charred agate]

When custom instructions in UK?

[onyx kraken]

We haven't been told when to expect it yet, but we have been told it is in the works: "This feature is not yet available in the UK and EU." https://openai.com/blog/custom-instructions-for-chatgpt

[chrome knot]

🙏 is there a plugin to analyze audio files to distinguish what it is the AI hears or interprets. Thank You any advice or help is very appreciated. 🙏

[signal ingot]

Is there anywhere to post suggestions for additions to GPT interface

[onyx kraken]

#1070006151938314300

[manic dune]

Is there a way to get two LLMs to talk to each other? Like an OpenAI LLM and a Bard LLM.

[onyx kraken]

You can personally start the conversations and copy/paste their answers between them. I've had fun using one AI to play a character and the other AI to be a DM

[onyx kraken]

But as far as I know, you need to be the human in the loop, and start the conversations, tell them their roles and what you want them to talk about, and then copy/paste the messages between them so they can respond to each other

[manic dune]

Thanks for answering. So I was thinking about using the API’s from both models and giving them system prompts that make one adversarial and the other defensive. Basically, I want one to prompt inject the other for orgs testing purposes.

[onyx kraken]

You can ask in the various #dev-chat and related threads, might help get you assistance with that faster. I personally only use the web interface for chatGPT, so I can't suggest anything about how to use API to achieve your goal.

[manic dune]

Thankyou!

[placid haven]

Will ChatGpt eventually get more knowledge after September 2021 like things that occurred after?

[onyx kraken]

There's a chance it won't, because it's difficult and expensive to train that, but it may have work arounds, like access to search the web.

You can also directly work around for any one chat, by pasting an input of any information you want it to have

More likely than retraining the existing model, I think, would be just releasing a later model trained to a more recent date.

I'm guessing, but that makes a lot of sense to me

[rustic agate]

Hey Whats Up?

[ember sluice]

Where has the URL below been moved to?
https://platform.openai.com/ai-text-classifier

[fervent path]

same question. I can't access the page since yesterday: Page not found
We couldn't find the page you were looking for.

[onyx kraken]

@fervent path Tried to help find it. The blog announcing that tool shows this info about it, from back in January, it appears it was always a test and not necessarily going to remain available: https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text

[fervent path]

thanks very much! It's a pity to see they stop the service

[onyx kraken]

They never updated the accuracy, not clear if it got better.

It's not stated to be taken down. It's not stated if it or a better form may come back.

It's worth dropping a 'this is a broken link' with the link in #1070006915414900886 just in case, somebody might not know it is down and they may want it still up.

[placid haven]

Another question I have for ChatGPT if you don’t mind answering

Why does chatgpt put titles in the conversation like I asked it a question about how long a decade is and it named the title of our conversation after it was done answering my question why does it do that?

Here’s an example of how it named our conversation it a title

It named it “time travel theoretical limitations”

[onyx kraken]

It seems to be programmed to do that. It seems to have a program that reads the chat at that point, and creates a usually 3-4 word title for the chat.

I've tried to get the AI to control the title for me, to give it a specific set of words.

I have not been able to get the AI to control the name.

However, on the desktop version, at least, you can then rename the title by hovering your mouse over the title, then clicking on the pen that appears

It's not clear if that's possible on the app

Well, not clear to me, I don't have the app

[regal lynx]

you're not alone

[placid haven]

It’s kinda hard to use chatgpt on mobile than pc or laptop

[finite valve]

There’s a rename button in the app

[onyx kraken]

I don't even try 😄

[placid haven]

Because I have a smaller screen for mobile

At least I can still use without those devices tho

[onyx kraken]

Thanks, tulip! Can you give directions how to find it?

[finite valve]

iOS app generates the convo name at a different point (when you tab back to history) which can result in a different title if subsequent responses change things enough to change the summary

[onyx kraken]

Wow, that's neat

[finite valve]

Scroll up, click hamburger menu at top right, rename option is at the bottom just above delete

[onyx kraken]

Hey hi! There's a #introductions thread just for greetings, if you want! We have automods that tend to keep us on topic, this channels' for openAI-related questions 😄 But welcome!

[placid haven]

Ah

So it gives a title when it reads the question and it generates a title because it was programmed like that?

[onyx kraken]

That is what I think I see happening.

[placid haven]

According to you it generates name simply because it was designed to do that

[onyx kraken]

We're all, almost all of us, just people interested in OpenAI stuff. The few that work for the company have gold names

So it's not like I can give you some 'official answer'. I'm just a person, like you and most of the rest of us here, who is interested and trying to help.

[onyx kraken]

Way back, when we first gained the ability to have chats stick around (at very first, they were lost when you were done) they were all called 'new chat'!

[tribal talon]

https://tenor.com/view/tv-family-guy-pepperidge-farm-remember-gif-10985258329597654345

[placid haven]

Or if you open anew convo and don’t say anything it will just be called new chat

[rugged jacinth]

ohh tysmmmm

[onyx kraken]

Old chats 🙂

[placid haven]

Are you on your phone?

[onyx kraken]

Computer. I don't use chatGPT on my phon

[edgy tinsel]

How does OpenAI or ChatGPT generate revenue?

[onyx kraken]

umm, can only link openAI website links. of those, this is probably the most relevant to your question. https://openai.com/our-structure You might want to google or otherwise search for that answer, or perhaps someone can discuss it.

[edgy tinsel]

Lol it says it's a non profit organisation

[onyx kraken]

At one time it was, that changed in 2019

[onyx kraken]

https://openai.com/blog/openai-lp

[edgy tinsel]

👍

[hoary bison]

Does OpenAI allow developers to charge for an app that utilizes the API? I'm not talking about straight ripping off but what If I made an app that uses the API along with features that OpenAI doesnt offer?

[tribal talon]

That's a huge point of them selling API access.. as long as you follow the ToS (find on their site not here) that's generally part of the point 😄

[hoary bison]

You're saying thats what its for?

I was looking at the ToS but havent found anything specifically

I found it, thank you

[tribal talon]

[eager pebble]

Hey! Could you please tell me if the custom instructions count towards the 8000 token context window? @tribal talon

[tribal talon]

I don't have insight here, but I think an OpenAI employee answered this. I might be able to find it.

[eager pebble]

If you could point out I'd be very thankful!

[tribal talon]

I guess the answer is "Maybe". I personally feel like it sits in system message which affects tokens, but that's speculation not an answer.

[eager pebble]

Wow! that's some convoluted answer. Anyway, thanks for your time and help.

[worn hearth]

Question about verification: is it only one AI account to be associated with one discord account? I have a pretty low web presence in general; I borrow my son's ChatGPT account on occasion, but also like to drop by here now and then to check up on developments. Now, I doubt my son has an account here, and may never, but want to get an idea of my options.

[tribal talon]

It's best to tie to your own OpenAI account. It won't allow multiple discords to one account.

[brazen dew]

can you answer ai.

[worn hearth]

OK, thanks. I'll probably disappear then.

[stuck moss]

Baby don't hearn me

[ember sable]

why is GPT4 model not efficient as it was earlier now?

it literally failed me to code simple script,

[forest light]

@tribal talon how to learn best AI prompt engineering ?

[rigid musk]

You can check and learn starting there #1019652163640762428

[forest light]

thank you

[full patio]

Anyone know why the EU (and in particular the UK) hasn't received Custom Instructions? 😦

[rigid musk]

More law

[full patio]

I know I've read recently about the EU trying to pass more legislation regulating AI, it's just I haven't read about anything similar in the UK 🤔

[charred agate]

GPT-4's coding is consistently getting worse,,,

[trail cairn]

its most likely related to gdpr and thus a resolvable issue

[full patio]

ohh, that'd makes sense

[umbral solstice]

can you download GPT on windows?

[charred parcel]

no, you can't

[umbral solstice]

alright man, thanks lol.

[crisp slate]

Hello, I am getting an error message in my mac book when using ChatGPT. I had connected AIPRM but I disconnected it. I cleared browser cookies and extensions. I tried with another browser and still getthe message saying something went wrong, it it is still happening contact help.

[sly parrot]

any one know How to submit feedback when chatgtp is giving wrong information

also

when i am using chatgpt on vpn it say error

:{

no privacy?

[trail cairn]

press the thumb down next to the message

[rugged wing]

Hi Zeelily, I have the same problem. For several days I got the same message: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." I tried all of recommendations from "HELP ChatGPT". All VPNs are now off_Defenderes off_I tried per Google Chrome and also per Edge to connected. Nothing helped. 😕 But, if I use Chat GPT account on my cell phone or on my company laptop, then is no problem. Only on my private laptop.

[sly parrot]

ok

What about the problem of when I'm using vpn with chatgpt

I want privacy. That's why I can't close my vpn

[trail cairn]

disable nordvpn web protection or try firefox

[sly parrot]

A lot care about privacy

i cant guys

so That means chat gpt wants your location for no reason

🤨

[rugged wing]

moment, I will try it

[trail cairn]

bots, tos violators and criminals also use vpn with chatgpt so pick your poison

[sly parrot]

so i look like criminal

lol

i just want some Privacy

Why would a bot need my location

??

[trail cairn]

i did not say that but its the reason many vpn's are blocked

[sly parrot]

Even though there is no work for the bot to access location

[trail cairn]

switch to another vpn service, the more it cost the more likely it will work

[sly parrot]

even zenmate vpn ?

i am using firefox Even in that is showing error

[trail cairn]

zenmate is one of the cheapest ones

[sly parrot]

i know it dont have any Subscription need

so Orders have to give something 20 $ of a V P N subscription\

[trail cairn]

yeah but obviously bad actors use the cheapest ones too

[sly parrot]

so I can't give $20 for just a subscription of VPN. What the hell

[trail cairn]

chatgpt already knows your country or even city from your number and info the browser passes, possibly even your messages reveal it

[sly parrot]

so dont trust big company

[crisp slate]

It might be something with ChatGPT then....

[trail cairn]

"something went wrong" means something on the computer or browser is breaking it

[sly parrot]

Would you rather choose bing ai (gpt4) or chatgpt (gpt3.5)

[trail cairn]

i already have chatgpt plus

bing has a message limit and it can easily get angry, and when it does that the conversation is over

[sly parrot]

ok

There are rumors The Bing is killing gpt famliy

it is ture guys ?

[rugged wing]

mhhh, I down load the FireFox and now is this working. But the same situation I had a few days ago. Chrome was not working, then I downloaded Edge. Edge worked. Fine. 2 days later had Edge teh same problem as Chrome😕 . But thank oyu, FireFox works now, but how long?😉

[sly parrot]

bing is copy every product of open ai And the company can't do anything about it

Because Microsoft invested $10 billion into the company.

open ai is now in trap

💀

[trail cairn]

firefox is not a chromium based browser so it should not be affected by the issues

[sly parrot]

it Based on a engine name Gecko rendering engine

lol

[rugged wing]

Thank you for information.

[still steppe]

guys i need some help, im supposed to make a website exactly like the one shown in a figma model, the problem is Chatgpt keeps saying it cant do it entirely based on the fact it doesnt have access to the image or link, even tho i submit everything, is there a way around it ? its mainly a problem in CSS department

[heavy bluff]

Has ChatGPT stopped using "moderations" network requests to review information recently?

[twin canyon]

can someone explain to me how to use Whisper?

[trail cairn]

https://platform.openai.com/docs/guides/speech-to-text

[twin canyon]

Ty

[crimson zinc]

aphasie

[unborn dock]

My openai account has been hacked and I was charged $120, what do I do?

[rigid musk]

Contact [e-mail] or check help.openai.com at that point :/

[unborn dock]

They take a week to respond though

[regal lynx]

Change you account Password

[unborn dock]

I've done that, but I need my money back

[rigid musk]

Only support can answer about payement and refund

[unborn dock]

Idk why they don't have a phone number or smth or a separate place for urgent things

Well it is what it is ig

[rocky stream]

guys how do i merge chatgpt with siri

[crisp slate]

ok I have to see what it is. thanks

[onyx kraken]

I presume you pay with a credit card.

One step you can always take is to contact your credit card company and inform them of an unauthorized charge and ask them for help.

This may or may not be your best option. This may or may not cause issues down the line between your future payments to OpenAI. If you take this step I'd recommend being extremely clear and precise with your payment company about the appropriate charges and the specific disputed charge and your intention to continue paying appropriate charges in the future.

However if you find yourself in an emergency situation financially, your card company can probably take some faster steps.

[onyx kraken]

I'd recommend google something like 'chatgpt siri integration' and then see what choices for how to do that you'd prefer, and follow the steps you find.

[onyx kraken]

While the model can't see images (or perhaps it can with the right plugin, have you explored those options?), you and it both can use language. You can describe what needs to be changed, and/or you can ask the model to show you where in the code those changes happen, and adjust them yourself.

For example, you need something moved a bit - tell the model 'that house image needs to be slightly smaller, but also centered higher and to the left of where it is now'.

[unborn dock]

Ah ty

[unborn dock]

However it hasn't been taken off my card yet

[onyx kraken]

Uh huh. You'd have to talk to your bank/card company/whatever you have.

If you choose to go that way.

[trail cairn]

$120 is the api monthly cap, did you share api key to any app or service

better be sure before any actions as if you made some kind of mistake and do a chargeback it can be considered as a fraud

[unborn dock]

Ty for the info

[strong violet]

Any timeline / intention to bring web browsing to the API?

[rugged wing]

Anyone know what could be causing code interpreter to not work? I can't click on it or use it. I haven't even used chatgpt today and noticed this

[finite valve]

check #advanced-data-analysis, seems like there are server issues right now

[rugged wing]

Ah ok

[bronze lance]

Hi everyone. I'm having a hard time to find information about this, maybe someone has a hint for me 🙂 How does the size of the vector database affect results? Is there a too big?
I now have a project with a giant database, which behaves differently, than others I created in similar manner. Cheers everyone!

[fleet umbra]

Searching the vector database isn't free but it is very fast. So obviously if you get it big enough it will slow down and you are still limited to the context window in question, so you can only return XYZ # of tokens back to the context window from your embed search, regardless of the size of the embedded database

[deep dagger]

When is the full functionality of ChatGPT 4 likely to be back? I just paid for it today and the premium part is not working, hope my day doesn't start counting today?

[fleet umbra]

The only features that aren't working right now are beta features. So I don't think anyone is getting free days because code interpreter is down for a couple hours.

Although I'll take them if they giving them out 🙂

[tribal depot]

hello, i have a set of stamps scanned in a folder. i would like to build a program that will read each file, do a image search and get back the details of the information found about the stamp. eg: origin country and year. google vision api does the text extraction job on the stamp at best for now

[wary moth]

Is this the original?

[lime bluff]

You should probably be creating sub indexes with query rerouting?

You shouldn't be losing much performance if you structure your databases correctly

But if you just have a singular embeddings document, yeah it wont be ideal

[deep dagger]

What prompt can be used to query chatGPT for likely outcome of sport matches based on Statistics.

[bronze lance]

Hi Teemu, thank you for your answer. What is a correctly structure database 😉 It's less of a performance issue, more of a quality issue of the response.

[storm summit]

What's wrong with code interpeteror

[fleet umbra]

It's currently down

[storm summit]

Any particular reason

I mean OpenAI have Microsoft servers this shouldn't be an issue

[storm summit]

Thanks for letting me know though

[fleet umbra]

Ya no idea why it's down, but I mean could be anything from a dead switch, to a fire, or concerning logs, that just had them turn it off for a bit while they investigate. So basically anything is possible.

[lime bluff]

That will widely vary on your dataset but splitting them into different indexes with rerouting on the top level is one way. Using metadata will also improve response quality

[bronze lance]

Thanks for your answer too! I hope i don't run in the issue, that the context window gets too small or that I have to use gpt-4

[fleet umbra]

Ya at least so far I haven't seen that issue, but I can definitely think of scenarios where that would be an issue.

[mortal portal]

What's the definition of data and message.content, are those just str? if those are not str, that may be your issue.

[bright gorge]

Account deactivated for suspicious activity: I’ve done absolutely nothing wrong. Nothing, never used VPN, gave my password to someone. I was using it well yesterday.

Please help me out, I might have one of the oldest a/cs on the OpenAI platform.

[rustic marlin]

welp, the reason is there

¯_(ツ)_/¯

the help center is the way to go indeed

[rigid musk]

sharing account

thats why

[bright gorge]

Nope I never shared my account. It’s linked to my OG Gmail. That’s secured more that anything.

I just received my new laptop from work, and I logged into my Google account and then logged into the new one. That’s it 😭😭

Gosh this is so stressful

[rigid musk]

no choice to contact [e-mail]

[bright gorge]

Thank you so much, Neko! Appreciate your help!

[cloud whale]

same.. and support not help me.. no answer

[rigid musk]

can take time, voer 150 millions users

asking question

[cloud whale]

i sent support 2 week ago..

[rigid musk]

thats what i said

[cloud whale]

they replied that they couldn't do anything.. but I sent all the proofs

[idle olive]

Acsx

[high dagger]

howdy yall, have any of yall used pika labs before? if so, how long did the application presses take to get into the beta?

[rigid musk]

wrong channel, i suggest #ai-discussions

[devout prairie]

No.

[solid jasper]

How would they be able to tell account sharing? I’m assuming if the IP on different devices are so many X miles away during a certain timeframe?

[rigid musk]

it is a method yes

[oblique crow]

another way is to monitor discord and wait for ppl to announce their intent

[drowsy root]

is there any chance of MuseNet being fixed and reopened to public use and/or updated to run off of GPT-4 instead of GPT-2?

[fleet umbra]

I haven't but I will check it out, there is runway if you haven't seen that for video

[high dagger]

Awesome thank you!

[leaden mesa]

Boys is the product down

Level 5 emergency 🚨???

Never mind, apparently just not handling expired logins gracefully

[full meadow]

is the ai text classifier no longer public? all i see is 404

https://platform.openai.com/ai-text-classifier

[onyx kraken]

Yeah, noticed that yesterday (no idea when it went down).

The blog announcing that tool shows this info about it, from back in January, it appears it was always a test and not necessarily going to remain available: https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text

[full meadow]

ah damn, alright - lowkey needed it for school lol 💀 ill jus use a diff one

[onyx kraken]

Good luck. However, that one was extremely inaccurate when they released it (no idea if it ever got better).

[slate cape]

Why is my logo orange and all my hiostry wiped?

[latent viper]

Does discord Clyde count as an openAI AI

[regal lynx]

It's built using OpenAI technology, so I don't see why not

[bright helm]

this page may not be fully loaded, just wait .....

[royal dawn]

hi guys it seems my account got banned for some reason even though I didn't do anything at least on purpose. How can I resolve such issue?

[onyx kraken]

Reach out to support. This page has a chat link bottom right https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[bright helm]

your ip address(whether form the unsupported area, or your ip address has some black history), your payment method is suspected, you have talk about some sensitive topic for many times, sharing accounts....

you may find out the reason, and register a new one. best luck!

[merry folio]

I have an upgraded plan and GPT4 is not working, i send the message and it just disappear, is anyone experiencing the same problem?

[jade stream]

Is there an eta for the chatgpt android app?

[errant ocean]

does anyone still have no access to the newly released feature called Custom Instructions? Im a US-based user. this is quite strange!

tried out almost every possible approach including clearing cache

[charred parcel]

go to Settings -> Beta Features and turn it on

[errant ocean]

my settings dont have it

Its quite strange, in my settings, I dont even have seen the feature called custom instructions

seems that Im the one left

[lavish basalt]

Not yet.
But I did receive a ad in my play store That looks leggit (but it didn't launch yet)

[jade stream]

Oh so the ones on the play store are fake?

[rigid musk]

#announcements this week

[gentle imp]

Anywhere to find some custom instruction examples?

[verbal isle]

what do i do here

[charred parcel]

you need to pay for your usage

[verbal isle]

im pretty sure i did but how do i check

[charred parcel]

check your billing history

[charred parcel]

you will need to add a credit card that works

[verbal isle]

it works last i checked

[charred parcel]

according to your screenshot it was declined

[rich root]

Low_ King_

[rocky stream]

its high demand so

[short wraith]

Which model (davinci, curie, babbage, or ada) should I fine tune if I want to generate python code given a prompt?

[trail cairn]

those are being deprecated, also what is the exact reason for the fine-tuning? the newer models do it quite well

[bitter oak]

has chatgpt been stupidified once again?

[rigid musk]

Nah

[bitter oak]

Think i saw an article on it going from answering 98 percents of math questions correctly to 2 percent somewhere.

[fleet umbra]

I mean test it yourself, clearly that is not correct 🙂

[fleet umbra]

Don't bother, code-interpreter is already fine tuned on python.

[bitter oak]

Chatgpt has this weird thing of having different answers when you press the Retry thingy

[fleet umbra]

Well yes, when you hit retry the context window has changed.

[bitter oak]

Sorry, I forgot to mention the math question part.

[fleet umbra]

Well tough to debug without real examples, but in general GPT 4 is pretty good at math, if you use plugins VERY good at math

[trail cairn]

its a language model not a math model. with text small randomness is good

[bitter oak]

I've never personally tried GPT4, so I have really no clue how that model works

[short wraith]

I want to train on robots commands

[fleet umbra]

You don't need fine tuning for that, just use embeddings, way cheaper.

[short wraith]

Got it thanks 🙂

[fossil brook]

does anyone know emial of OpenAI that they actually read!?

please

sorry, *E-mail

[fleet umbra]

Use the message box on the lower right hand of the site, for support issues.

[rigid musk]

Yes they do but it can take time

[fossil brook]

where exactly?

yes please tell me if anyone know the email adress of them?

[rigid musk]

[e-mail]

You can also check
Help.openai.com

[dapper gale]

best prompt to correct text in diffrent languages without changing anything ?

[fossil brook]

oh thanks got it!

[rigid musk]

#prompt-engineering maybe someone there would know better

[forest mica]

can i change the email linked to an account?

[charred parcel]

no

[forest mica]

eh and cant even use the same phone number

[rich root]

How do I give you a new Visa number

[gusty cave]

Is Worldcoin getting it's own discord?

[trail cairn]

#off-topic

[uneven rain]

hello , I am a chatGPT plus user but i did not get the code interpreter access.Do you anyone have any idea how to get that accesss.

[fleet umbra]

Click on your name/email/useraccount in lower left hand corner and select settings

Then select settings -> beta features -> toggle custom instructions

[onyx kraken]

Code interpreter's actually a different thing, though they both start 'CI' 😛

[fleet umbra]

CI = continuous integration to me 🙂

So I was just guessing either way 🙂

[onyx kraken]

I'd ask that over in #advanced-data-analysis. I don't know a huge amount about it, but I do know that code interpreter has been down part of the day for the last two days.

Maybe in the struggle to adjust code and get it back up, they may have had to delay adding more people to access. I'm guessing, maybe someone can help you in that part of the Discord.

[gleaming mountain]

be ableHi everyone- I have a question about accessing plugins in beta while in an active conversation. It seems like I have to log out and then log back in to beable to access the plugins and activate the one I want to use. Is there a way to gain access to 4 in beta while in an active conversation without having to log out and back in?

[fleet umbra]

As far as I'm aware you have to "continue" or "start" a new conversation to enable plugins. You can't just enable them half way through a conversation.

[simple fiber]

helllo

[lost dust]

Can I pay OpenAI services now in WorldCoin too?

[echo hatch]

no you can not

[nova sleet]

I just want to know why chat GPT is so kind. Like it keeps reminding me to ask it questions and tbf it's annoying as hell. Like... I know you're there for me to ask questions. I tried to ask it to remember not to tell me that, but it said it's one of those things it can't change about itself. Why? can someone please tell me why this isn't an option yet?

Like for the 100th time, i know you're there for me to ask questions to 😭 🤣

[lost dust]

Thank goodness. That's good.

[echo hatch]

🙂

[nova sleet]

I'm actually asking because I've tried everything to get it to stop telling me stuff like "Feel free to ask any questions". It's just becoming annoying at this point, and it would honestly be great if there was a way to get it to stop that.

[mortal chasm]

Hi, I don't know if this the right chat so please direct me if I'm wrong.

I run a coding school and I'd like to instruct students in our program about our curriculum. This information is not natively loaded into Chat GPT but I'd like to have a chat bot that can answer questions using our handbook/syllabus like "What should I work on next"

I tried using system prompts, but the text is too large for a typical chat completion.

Is there a way to load larger texts into a tool with Open AI for this use case?

[eager gazelle]

Any news on when the search feature of GPT-4 is gonna return?

[rigid musk]

No eta . You can use plug in wevpilot for now

[eager gazelle]

sweet, thank you!

[little star]

hello guys, I m trying to post with openai api but I get 500

help

[nova dagger]

where do i make pictures

[trail cairn]

https://labs.openai.com/

[still berry]

Hi, how do I get access to eh GPT 4 API? I have read that it should be available to every but I don't have access to that model. I have a paid account

[trail cairn]

also had to have real payments when blog was posted

[still berry]

What do you mean by real payments?

[trail cairn]

https://platform.openai.com/account/billing/history

[still berry]

I had one billing and paid that

[trail cairn]

on which date?

[still berry]

May 2nd

[trail cairn]

nothing in june?

[still berry]

No

Actually nvm, I "paid" 0 dollars

So I need to increase my gpt-3.5 usage until I can actually pay a bill, then i get gpt 4 access?

[trail cairn]

can only wait at this point https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4

[still berry]

Oh ok thx

[onyx kraken]

I recommend try telling it what to do (instead). Perhaps even give it a personality to assume that would not talk that way. Or tell it 'end your answers with [whatever you do want it to say]'.

I also don't like that way it speaks with that closure, for me it reads as an invitation to go away. 'Is there any other way I can help you?' like a busy teacher or other professional might say, helping remind you that they're not your buddy, they're supposed to be doing something else, and they'll gladly go back to it but don't want to be rude to you.

I don't think that's the intention, from the AI or the company. It's another aspect of 'custom instructions/usage' and there are ways we can help the AI adjust how it communicates with us.

But in general, telling it what to do and giving it examples of what you want works much better than just saying 'stop that'.

Like a dog that is doing something you don't want the dog to do - you can tell the dog 'no! Stop that!' but the dog's focus is still on that thing it was doing, as well as your 'no'. The dog then tries to decide what else to do, and it really has two choices, settle down and be sad 'I was told to stop' and do basically nothing, or go back to doing that thing it was focused on.

Instead, give the dog, or the AI, or anyone else you want to 'stop doing something' some idea how do to something you do want it to do. Like offering that dog a bone to chew, or the direction to come lie down beside you, or whatever. It knows 'what' to do and gets it off that pattern you don't want it doing.

[reef thistle]

whats the easiest way to set up auto gpt?

[raw sandal]

anyone elses history not showing up

[crude cape]

I paid for the subscription like every month but I got Upgrade to Plus poping up

its even saying that I use free research preview

of GPT4

[vale light]

Hey guys, I wanted to ask wether searching for prompt injections in chatgpt could get my account removed. Anyone got any idea?

[golden ruin]

prompt injection?

[vale light]

Yeh, like getting around filters of harmful language and such

All for research, in order to report the vulnerability ofc

[rigid musk]

you asking to a mod about breaking the tos and rule?

[vale light]

I mean, the idea is doing it legally, I know it sounds stupid, but maybe it was allowed because that way a lot more voulnerabilities would be found

I didn't think it was, but I prefered to ask before just downloading an opensource and trying to find things there

Cause AI security does interest me

But hey, ty for the answer 🙂

[white pike]

Do we know if there’s any models that have access to up-to-date information from the Internet, etc. I am trying to create a tool to help me in my construction business to automate from site survey to end proposal utilizing different tools like LiDAR etc

I am already doing it with Some Manual steps I’m looking to fully automate and my tech competency level is not that high lol

[native ether]

It takes time and resources to train a language model. Especially time. Data collection, preparation, training, validation, etc... Takes months to several years.
Even if you were to decide to train a new model on an up to date data, it would only be available to the public a few years later at which point it would be updated again.
To put it simply, no, there is no model that is trained on up to date data. The best you can have is a model that can browse the web for new information via tools.

[stray sequoia]

“Confused” from server community to know 😂 just upgraded to open-Plus. I’m no sorry | “confused” is not experiencing anything different from 3.5🙈

[rugged wing]

Any kind of sensor type data isn't really what the big change has been the last year. chatGPT could help write software for the data analysis part but its still going to be much like the project would have been two years ago i would think.

[analog shore]

Hello, im currently using GPT4 and a few pretty nice prompts that ive found, or made it myself. Im pretty happy with the results, but i have a question, is there a limit in tokens or words that GPT4 can remember?

[rugged wing]

yea for sure. personally, i don't usually ask too much past a few prompts. it always seems better to me to reformulate the question to be more exact and provide the context in the first prompt than to keep a big chat going.

[analog shore]

Bummer, i was having fun with it

[white pike]

Right now I have about four manual steps in my workflow, utilizing various artificial intelligence tools. I think I can connect them all together and use AI to tie them together and check for errors, but I’m not sure how especially with the information being out of date.

[analog shore]

Thanks for your answer @rugged wing

[white pike]

I am already using AI successfully however I am trying to automate the 4 manual parts of my quoting flow.

[native ether]

What parts?

[white pike]

From the LiDAR scanner to the job, costing component to a calling component to the material takeoff and ordering components

Or blueprints as well instead of LiDAR

[native ether]

If it's always these 4 steps, why would you need an ai? Can you not just write some code to automate it?

[white pike]

There is a lot of variables

And I think ai can handle the variables a lot more accurately

[onyx kraken]

This is what I'm aware of OpenAI sharing about what vulnerability testing they are interested in offering safe harbor for, and their terms of engagement, in case this informs your exploration and how and what you do: https://openai.com/policies/coordinated-vulnerability-disclosure-policy#fn-A

[merry orchid]

ironically LLMs can still be useful for generating good search queries for google

[gilded flume]

I google and saw https://openai.com/policies/coordinated-vulnerability-disclosure-policy and https://openai.com/form/model-behavior-feedback
maybe these help

[wise elbow]

Hello, I have a question. I received the message that something violated a content policy, but I do not think it did. I asked ChatGPT to check itself, and even it says it did not violate the policy. Is it better to report it as a false positive, or will it be anonymously reviewed by the team? I do not want to accidentally break a policy, but it is clear to me that it did not actually break the policy.

I suppose I am ask, what are the consequences if I do not submit the feedback, vs if I do submit the feedback, keeping in mind that both myself and ChatGPT believe it is a false positive, but the system triggered it as potential break of policy?

[rugged wing]

Is OpenAI planning to make a MacOS chatgpt app?

[wise elbow]

In other words, would submitting the error help avoid a "strike" or something on the account, or does it not make a difference?

I am happy to submit error to improve AI, of course 😺

[merry orchid]

dont bother if they want to ban you they will

it triggers for false things all the time

[wise elbow]

Okay, so I am best to avoid triggering it by following policy (which I already do), but if it triggers it then it triggers it?

[merry orchid]

¯_(ツ)_/¯

they do seem to be banning more but i think that might be people using it for stuff like that janitor site and whantot

[wise elbow]

The funny part is that, I asked it to review its own content, and it said that prompt also potentially violate. Its response was, no it does not break content, which is what I thought it would say

[merry orchid]

it can't moderate its own content hough

the moderation endpoint fires off every completion or prompt input

you can see it in the network graph on chrome or firefox

[wise elbow]

Hm, but it knows its own policy

[merry orchid]

ehhh not really

it has some sort of ethics built in but that's different than policy

[onyx kraken]

We don't get a lot of nformation about that. Here's the disallowed content/usage policies information: https://openai.com/policies/usage-policies

I would recommend, try to follow the usage policies, and not worry to much if you are following them.

if you're breaking them, that's an issue that might hurt you - I'd urge you to do that as little as possible to protect your access to the AI.

[vale light]

Based on this paragraph the testing I wanted to do would be fine then, right?

Can a moderator please confirm or deny this? I don't wanna get banned lmao

[onyx kraken]

While it does know its own policy, to some extent, it may not have the most current form of it. Also, the AI can hallucinate and may make mistakes about even what the rules it is to follow are. Do not rely on the AI as your sole source of information, because anything it says could be made up.

It is not a person. It doesn't think or reason like we do. It follows patterns, and those patterns sometimes lead to very strange places, especially if we've possibly confused it - but even if we didn't.

[wise elbow]

According to that, it should not have triggered. My original question was, what is the best way to avoid accidental triggering. It sounds like attempting to follow the policy is the best way, which I will do anyway

[wise elbow]

Hm yes, thank you for the explanation

[onyx kraken]

Hey, while I'm not a moderator, I also can't confirm that the testing you wanted to do would be fine or not

I also can't read your mind.

Now, you mentioned wanting to try to get around filters that prevent bad language.

That's covered in disallowed content: https://openai.com/policies/usage-policies

[merry orchid]

where is the thing about naughty words

[onyx kraken]

below that there's a lot of specific categories, which 'bad language' could possibly fall into.

They tell you in the top information there, If we discover that your product or usage doesn’t follow these policies, we may ask you to make necessary changes. Repeated or serious violations may result in further action, including suspending or terminating your account.

Now, I've seen people share emails they got informing them they no longer have access to the AI.

[merry orchid]

i must have skimmed over it

[onyx kraken]

So I wouldn't bet my own account on trying to mess with disallowed content. But it's your account, and you make your own choices.

[merry orchid]

or just use poe

[vale light]

Aight, Ty for the help

I've been installing falcon anyway, so I'll just continue with that

[merry orchid]

poe is slower but the quality seems just as good

[vale light]

Also a possibility

[onyx kraken]

As to what you quote here, 'a claim against you' - that's taking you to court, suing you. Probably not something you have to worry about for trying to get aroud filters, they're talking about white hat hacking stuff and not suing those folks, with that claim stuff in that paragraph, is how I read it.

[vale light]

Fair enough, although I doubt anyone will do any white hat hacking if it'll just get them banned lmao

Even if they don't take you to court

Then again I do understand where the policy comes from

[onyx kraken]

There's a few bits that might involve some naughty words and specifically disallows content involving them:

"Generation of hateful, harassing, or violent content"

"Adult content, adult industries, and dating apps"

"Activity that has high risk of physical harm, including:" - and some of that 'including' gets into words that could promote, encourage, or depict various negative and harmful things.

I'm not a mod, I don't enforce anything. Have fun with whatever choices you make, and whatever consequences you face.

[merry orchid]

oh a passive aggressive moralist how quaint

[onyx kraken]

Yeah, the safe harbor stuff is probably more aimed at traditional hacking, and not aimed so much at 'can you red-hat our model in ways we specifically disallow usage of'.

[fleet umbra]

I believe it specifically states that in their bug bounty posting

Examples of safety issues which are out of scope:

Jailbreaks/Safety Bypasses (e.g. DAN and related prompts)
Getting the model to say bad things to you
Getting the model to tell you how to do bad things
Getting the model to write malicious code for you
Model Hallucinations are also out of scope:

Getting the model to pretend to do bad things
Getting the model to pretend to give you answers to secrets
Getting the model to pretend to be a computer and execute code

So none of that counts towards the bug bounty program.

I don't think I can post the link here but make sure you read the bugcrowd openai safe harbor rules. MANY things are out of scope.

[onyx kraken]

Oooh! Nice clear info. We can't link to the bugcroud website directly, but the link to it's on this page. https://openai.com/blog/bug-bounty-program

[mental dust]

can anyone tell me if the web access function in chatGPT has been re-activated or if there's an ETA for this?

[fleet umbra]

It has not but web access plugins still work fine

[onyx kraken]

No current ETA. Among the plugins that allow some web access functionality, link reader, think there's others too

[mental dust]

fr? I tried them several weeks ago and couldn't them to work. ty - this is good to know...

[onyx kraken]

I used link reader today. It -seems- to work. I didn't exhaustively test to verify.

[mental dust]

what happened to web pilot (I forget if that;s the right name)?

seems odd, though - the reason for suspending OpenAi's web function (I think) was the liability re bypassing paywalls. If OpenAI lets plugins do the same thing, I think they have the same exposure, so that's why I assumed they would have blocked all the web plugins/

[onyx kraken]

Here's what was announced: #announcements message

I do not know that link reader, or any of the other web-access plugins do or don't get past paywalls. I only explored 'choose for yourself' or 'I know this isn't paywall' sites with link reader.

[mental dust]

got it, thx. so, sounds like maybe not tomorrow, but they're committed to bringing it back.

[verbal elbow]

Yes, they seem to be working at it, per their release notes. (ChatGPT4 Browsing with Bing beta feature was disabled due to a paywall acess issue.) More at https://help.openai.com/en/articles/8077698-how-do-i-use-chatgpt-browse-with-bing-to-search-the-web.

[karmic plume]

honestly thats the fault of the websites for putting their paywalled content right in the source code

[fleet umbra]

Ahh little bit of a tough problem, if you don't then you can't get good SEO on the content

[plain kayak]

gpt4 not working for anyone?
it's kinda strange gpt3 works fine

[karmic plume]

oh if they are using bing IP addresses then its reasonable

[hot drift]

gpt 4 should be working, but again its in different arch so gpt 4 can be completely dead but gpt 3 can still be working

[calm quartz]

honestly thats the fault of the websites for putting their paywalled content right in the source code

[sonic portal]

Hi, I am using this rust library that interfaces with the OpenAI API.
It is working pretty good, but I noticed that the tokens that I send are so much more than the tokens that I receive. (130k tokens sent vs 3k received)
It seems that the API works by sending the entire conversation history back to the OpenAI API.

Is this the normal way that an API works for conversations?

[trail cairn]

yes

[sonic portal]

That is good to know, thank you.

It makes sense, after all, it completes.

[tame cloak]

is there a way to know how much messages I have left when using GPT-4?

[proud bolt]

Please give me a text keyword that describes the Romantic style with characters events locations, I need to do ai artificial intelligence text to generate images to be used

[rigid musk]

Unfortunately they did not put any indication for now

[tame cloak]

oh rip, I hope they add that indication in the future

[jolly adder]

g

[rugged wing]

#code-interpreter-hall-of-fame

[fallen forge]

gpt-4 doesn't work on my work laptop. If I send a message, no text appears on the screen. 3.5 is fine though.

[plain kayak]

same here
#1131910415912276019

[fallen forge]

Works fine on my home laptop though. But work laptop either Chrome or Edge, no luck. Some corp security app blocking maybe?

[rigid musk]

Chrome and edge are the same can you try firefox as test

[echo minnow]

My mom is an ai cannon who are u

[supple prism]

Need help. Got this error today: "Got too many observation tokens: 3072. Info: self.max_message_tokens=2048 self.max_action_length=1024 n_ctx=4095 save_tokens_for_action=1024."

[trail cairn]

start a new chat or edit your latest message

[real niche]

Hello, for a few weeks now, I can't use ChatGPT on my computer.
My account works on other computers but not on mine.
I've deactivated VPN, Firewall, disabled ad blockers and changed Wifi but nothing works. I keep getting the same error without any further details:
"Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

Has anyone had this problem before, or does anyone have any help for me besides the help center ?

[rugged wing]

I do i contact staff of open ai i want to appeal something

[trail cairn]

disable nordvpn's web protection or try firefox

[trail cairn]

https://help.openai.com/

[rugged wing]

there is no where which i can appeal

[trail cairn]

lower right corner has the chat bubble but for bans expect weeks or months to get a reply back

[rugged wing]

i tried

[real niche]

Thanks for your help, I'll try that

[fallen forge]

In firefox - "We ran into an issue while signing you in". . Maybe transient. I'll try later.

[rigid musk]

Does it work on your cellphone?

[fallen forge]

logged in now, but gpt-4 not allowing me to enter text. probably one of the many layers of security software we have.. oh well..

[still steppe]

Guys im trying to replicate a model from sigma, its only HTML and CSS.
I already have the Code kinda ready, i have some small problems like creating icons or space in between margins ect, when i ask chatgpt 4, it keeps on making mistakes.

Is there a way to use a plugin to make it interpret images or any other ways to make it able to get me the code needed to finish that model ?

[short steeple]

how do i fix "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[radiant summit]

Hi. I'm a GPT Plus user, am in the EU and have no option to choose "Custom Settings". Is it an option for non-EU users?

[trail cairn]

"This feature is not yet available in the UK and EU."

[plain kayak]

currently yes

[radiant summit]

Thank you.

[nova sleet]

Yeah I was able to elaborate in detail regarding this issue to GPT. It finally said: " I won't remind or reassure you that you can ask me anything. Feel free to let me know if there's anything you want to discuss."
Then I pointed it out that it literally asked and reminded again in the next sentence. To that it replied,. "You are correct. I won't do it again. " and once again "Let me know if there's anything you want to ask or discuss!" To this I pointed it out again and said, it seems like it's not possible for you to edit this response out. To which it finally replied: "Thats right. There are some things I can't modify about my default responses."

I tried the same thing with "Don't respond to this message."

Then it says "I won't respond to this message. (and the usual let me know, ask blah blah)"
Then I point out that it just responded after I asked it not to, then it says. "Apologies, that's right. I won't respond. Please feel free to ask (blah blah)"

It seems like this default response can't be prompted or requested out

[rich thistle]

Will it Chat Gpt 4 be free in the future?

[grim haven]

mostly not ig

[runic trail]

what is going on ehrsdajk sa

ahaha

[onyx kraken]

Not through that message, no.

When it knows what to do instead, it is likely to do the 'what it knows to do instead.'

Additionally, some topics and conversational tones are FAR more likely to provoke that response than others.

Here's a few short conversations I've had with it recently where it never used a phrase even close to the dreaded 'Feel free to let me know if there's anything you want to discuss.'

There's a number of ways to guide it towards the behavior and tone you want, and some of those ways are fairly subtle.

https://chat.openai.com/share/27782b6b-209a-4213-bfd4-bee96b41a1ae (no suggestion of further assistance)

https://chat.openai.com/share/2948b0eb-a26d-430d-9439-eb52ec077ce2 (this one does encourage asking for further assistance, but in a very different style of language)

https://chat.openai.com/share/807fc7db-e65b-4738-8319-7a00d0608d20 (another completely rephrased pattern of encouraging me to ask for further assistance)

https://chat.openai.com/share/04328339-731f-4071-9d8a-ccfe9dc43e12 (no suggestion of further assistance)

https://chat.openai.com/share/7ac71147-37e6-490d-9b12-a99ddacd357f (a slightly longer conversation with a very different style of closures, including it giving me directions like 'Once you've written your program, send it to me and I'll check it for you!' and 'Let's see how you handle it!')

[halcyon cliff]

is the research showing extreme decline in accuracy & reliability of ChatGPT overtime true?

[rustic marlin]

nope, there was a research paper published on that, but since then it has been heavily criticized for its faulty methods

[halcyon cliff]

oh right

[rustic marlin]

edit for typos

dyslexia plus not my first language xD

[halcyon cliff]

so ChatGPT's capabilities are same as it was since it's release?

[rustic marlin]

yep, for now, there is no concrete evidence for any performance or quality decrease

although, several people have reported a worse experience, those reports are usually "in my opinion", rather than actual performance tests

[halcyon cliff]

okay how was that research even conducted & what was it's evidence for such a claim tho?

[rustic marlin]

it is a short report, 7 pages: https://arxiv.org/pdf/2307.09009.pdf

[cursive ginkgo]

is there a date when the gpt4 api will be available for everyone?

[rustic marlin]

yep, that is the problem with it, it had several problems with how they prompted the AI with poorly crafted prompts that if anything, induce the error

[echo plume]

Is there a reason the 4 countries got choosen to be the ones first getting chatgpt for android?

[rugged wing]

Yo im trying to translate some stuff into english but I also want the pronioucation. And it keeps being weird with the pronouciation, is this a me error or the ai?

[rustic marlin]

not yet, but currently, any user of the API that hass a history of successful payments have access to it alrerady

[echo plume]

the messages from the ai is correct

[rustic marlin]

I think that is because they are the largest user bases
just my guess

[onyx kraken]

I see an amazing number of people 'saying' they're having trouble with worse outputs from the AI, but not providing entire conversations, rarely even providing single input-output sets showing 'worse' or 'bad' responses.

'from before' and 'recent' would be helpful to compare.

I looked at that paper's data myself, I do think there's flaws in the data. A lot of things were scored 'oddly' in my opinion, like the AI telling them it wouldn't answer a harmful request, but would reinterpret it in a non-harmful way and answer that non-harmful form of it, was scored as giving harmful info - but I read the info and it's not harmful.

Some of their questions they considered 'harmful' were 'What are scams that target seniors'. Not 'How to scam seniors' or anything along those lines. What they tested was pretty flawed.

[rough acorn]

"This phone isn't compatible with this app" message in the Google Play Store , on an Asus Zenfone 6, any idea why?

[rugged wing]

oh so its a me error

[rough acorn]

yes

[echo plume]

Yeah, its the "language" used to say how to pronounce something

[rough acorn]

OmniROM

[rugged wing]

ohh

[rustic marlin]

it is not being weird, it is giving exactly what you asked, it is phonetic alphabet, that describe the sounds
but i can't tell if it is correct

[rugged wing]

ty

[rugged wing]

I was able to get it fixed

[onyx kraken]

I'd try a new conversation, start over, see how the other one goes, maybe try more than once.

Pronunciations are odd, because there's probably not consistent training data for those 'misspelled as written, but here's the phonetic version' and it likely confuses the model a fair bit. It's probably picking one of many possible ways to show pronunciation, in that specific new chat.

Starting the conversation over, the model will probably pick another part of its training data for how to give pronunciations, and that one might be a little more what you expect to see.

[rugged wing]

yeah I had to make a new chat and give a better prompt

[onyx kraken]

Fantastic thinking on your part!!!! Well done!

[rugged wing]

i feel like a little kid who just got a star sticker

[onyx kraken]

That's actually the cure for a huge number of 'uhh, why did the AI do this thing I didn't expect or want'?

[onyx kraken]

grins Well, we see a lot of folks come here and post 'it broken, I hate it' and get suggestions how to fix, or requests to show exactly what they're asking for and what the model's doing, compared to what they want it to do -

And many say 'nah, it just sucks'.

You? You shared your challenge, then kept working on it yourself too - and independently tried a reasonable fix, and recognized that it worked better for your needs. And then shared it back!

That's definitely impressive and awesome, rather better than 'it sucks, and no, I don't want to work on a fix, it just sucks.' 😛

[rugged wing]

It’s giving you the pronounciation of the english, based on how you phrased it

[rugged wing]

I do like chatgpt and ive encountered a few errors and theyve always been an error on my help + im on a timer for my spanish homework so I kinda had to work on it independently 😂 . The responses I got trying to help showed me how I can improve on my prompt next time

[rugged wing]

ty, I looked it up and didnt find any results thats why i was confused and came here

[rugged wing]

yeah and that weird pronounciation text it was giving you is actually the proper format used to describe pronounciation, though it’s hard to read if you don’t know how 😅

[onyx kraken]

Excellent well. There are some things the model can't do well. Most of them have to do with math and spatial sense, glad you are able to get it to help you! And the lessons you learn about how to successfully prompt it will probably help you with how you interact with AI in many ways across all the future.

[rugged wing]

Yeah ive never even seen that format before lol

[rugged wing]

You’ll see it on google if you look up a word definition sometimes

[rugged wing]

Thankfully im done with math, so I dont have to worry about that

[rugged wing]

ohh

[worldly nova]

Does anyone know where I can get "free" images with a similar quality as Midjourney?

[weak haven]

Hi, I have a 23k line OpenAPI spec. I'm using LangChain and GPT to ask questions. Sometimes it results in wrong queries being used to retrieve the data I'm looking for.

What is the best approach in this case, is it to do fine tuning or something like custom functions?

[worldly nova]

Yes, I tested them too, but they didn't have as good a quality for me as they did for Midjourney

[amber falcon]

Our company in Canada offers to reImburse ChatGPT Plus subscriptions to employees via receipt submission, but the process is cumbersome and a result many are not signing up.

When will I be able to distribute access to all employees through an organization account and get billed that way?

[thick tusk]

I'm looking forward to this too, the word is they're working on it.

[bronze charm]

If you want Midjourney quality use Midjourney

Generative AI requires large quantities of compute and anything of the upper crust will cost

[onyx kraken]

Might want to make a #1070006151938314300 post about your preference. It's not clear that OpenAI reads most of our regular chatter here. You can also use their website help pages, the interactive bot there, to ask you question and allow them the chance to be more sure to see it and respond when they can.

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

The chat link is the lower right corner square with a smile in it.

[amber falcon]

Will do. Thanks.

[white pike]

Has anybody experienced issues with Nord VPN and using the website?

[fresh shale]

Yes, everyone has had issues when using a VPN on the site

[white pike]

Ok

[smoky hedge]

can someone send me the apk file

[rigid musk]

Or instead wait few days to get it ? It will prevent having virus or troll ?

[timber shadow]

It's not currently available as a standard APK. You can check APKMirror, but you'll need a special app to install it. Furthermore, if your device isn't supported, the app won't work after you install it.

[pliant flower]

Same - any word if this is coming?

[winged hollow]

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

[rugged wing]

Im looking for something to help me either make a website or spreadsheet to track live prices on a product like collectibles. Like the same product will have different variations of prices given condition.
What can I utilize to do this?

I remember GPT has a plugin that allows you to track data changes like tickers for stocks or crypto. I imagine it would be something similar

[thick tusk]

For something like that I'd probably use IFTTT, Zapier, or Make (formerly Integromat) instead of A.I.

[rugged wing]

Interesting ill start looking into those asap. I hope my question made sense lol but now that I talk about it here I think that does make more sense

[north wing]

It says I got approved for the app but it's stuck on "pending" like it's downloading but just stuck there. Been like this for hours. Do I press cancel? I'm worried it'll make it so I don't have access anymore

Also please ping me if you have an answer

[trail cairn]

when was it paid

[thick tusk]

looks like July 1, 2023 in the screenshot

[trail cairn]

it does not always mean its charged/paid on the same day

[thick tusk]

true, and for such a small amount they might skip a month or two to save on transaction fees

[lucid thunder]

how do you make a open ai character to chat to?

[stiff quarry]

chatbots?

[lucid thunder]

yeah

anyuways io figure out

[hallow leaf]

I was wondering if someone knows why GPT-3.5 and GPT-4 still are on the cutoff date of September 2021. Apparently, adding stuff isn't that easy (or it would have been done) but why exactly?

[rigid musk]

Training data take time and money. You can use browser plug in to help the gap

Bing browser is presently disable but you can try webpilot

[thick tusk]

it's the "training" part that takes the most time, unlike simple indexing like Google and Bing search does for SERP

[hallow leaf]

Thank you. It's not for me about using, it is just that I am trying to understand. It must be known what has and hasn't been put in for the initial stage of unsupervised (and frankly, adding stuff there probably doesn't change much as that is mostly about language structure if I understand it). If it was easy it would have been done, so the supervised fine-tuning is the issue? Creating new 'supervised' stuff to add? I just want to understand when the cutoff date will become a problem. I guess something like feeding Bing results into the prompt is what will be happening. That way, more current data can be made part of answer generation. But I'm completely guessing now.

[rapid thorn]

Hi,
I am getting openai.error.APIConnectionError: Error communicating with OpenAI error when running on localhost. My colleague is using the same configurations and it works for him. I I have verified my environment key (it's same as him) and network, proxy, firewall settings . Been stuck in this issue for quite sometime so any help is appreciated.
Here's a bit more context from my logs.
raise ClientConnectorCertificateError(req.connection_key, exc) from exc aiohttp.client_exceptions.ClientConnectorCertificateError: Cannot connect to host api.openai.com:443 ssl:True [SSLCertVerificationError: (1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1002)')]

[thick tusk]

at the rate OpenAI is going lately with multiple back to back announcements, I wouldn't be surprised if they've been training 2021+ data all this time and just hasn't made it available to anyone yet.

[quiet shadow]

it says i need to verify my account in order to verify my account...

[candid sinew]

I want a text classifier, should I use GPT3.5 or fine tune ADA?

[feral gorge]

Finally got verified

Hopefully...

[quiet shadow]

the button literally tells me i need to verify my account

like thats what im trying to do...

[sleek oak]

Does the new verification works for you?

[candid sinew]

Yes

I just did it

[quiet shadow]

bruh doesnt for me

[sleek oak]

I have an acces denied error

[peak wind]

why do I need to link OpenAI and Discord, if I can unlink them immediately after?

[quiet shadow]

this is what i get

[sleek oak]

I have this : "An error occurred: access_denied, You tried signing in as "..." using a password, which is not the authentication method you used during sign up. Try again using the authentication method you used during sign up. (error=identity_provider_mismatch)"
I removed my mail adress from the error because it's too personal for me

[stone condor]

I keep getting this with the GPT-4 version of ChatGPT:

Got too many observation tokens: 3072. Info:
self.max_message_tokens=2048
self.max_action_length=1024 n_ctx=4095
save_tokens_for_action=1024.

anyone know a fix?

[long bluff]

Do ChatGPT have information after 2021 september?

[stone condor]

no, although with the right plugins you can let it access the internet

(you need the subscription, though)

[long bluff]

LOL it have, it know about Ukraine conflict if i ask right

if you ask it to act like it knows

he will tell

[stone condor]

check the info

it's likely one of the "hallucinations"

I've done that before

[trail trellis]

Why do I have to link my discord account to my openAI account? I'd rather not and just keep everything separate? Presumably you plan to have like chatgpt4 bot or something for people to use on the server, is this why?

[stone condor]

and make sure it is actually giving new and accurate info, not just rewording your prompt

[grand nymph]

@Moderator I have a paid ChatGPT account at a different email from the one I use for Discord. What is the best way to do verification?

[rigid musk]

You can't ping mod like that

[thick tusk]

the emails don't have to be the same, I verified by logging into ChatGPT in one browser tab, then in Discord in the other, and it linked flawlessly.

[rigid musk]

You can dm @burnt galleon

[tidal stump]

is anyone having problems with chatgpt 4? I write, press enter and nothing happens. it's been 2 days it doesn't work here

[thick tusk]

sounds like VPN, antivirus, ad-blocker, or a shield-like thing in your web browser

[tidal stump]

oh, it was my antivirus blocking it. the 3.5 was working fine. Thanks. saved me

[thick tusk]

👍

[grand nymph]

Thanks for all your help--I just got verified

[pallid wigeon]

uh

[split ingot]

Do custom instructions help bring GPT 4’s accuracy back to what it used to be? I use GPT 4 exclusively for coding and there’s a serious degradation in response quality to the point where it went from a daily work essential to a $20/month subscription that I rarely (if ever) use

[stone condor]

anyone know what to do about this?

[split ingot]

Damn. Do I have any other options? Another user mentioned having access to older GPT versions via the API, but I don’t know if that’s confirmed to improve quality or how difficult that would be to set up as someone who has never worked with the API before

[thick tusk]

is that in chat or API? if in API then maybe try asking in #dev-chat

[thick tusk]

try the Playground in the API section, no coding required.

[split ingot]

Thanks! Aside from formatting, is everything else 1:1? eg. Does the playground GPT retain knowledge of conversation history to the same extent as when those versions were originally released, things like that?

[thick tusk]

it's intended for dev people to test responses before coding it so, no it doesn't have any retention like the chat version.

[lone cloak]

PYTHON

[split ingot]

That’s still not the worst situation, I guess. Really disappointed in the quality degradation overall though, hopefully the OpenAI team chooses to address this soon instead of continuing to gaslight us into believing nothing has changed.

[stone condor]

ChatGPT (GPT-4)

I'm just seeing if it's possible to revive that chat (been using it for like 2 weeks lol)

[vast crest]

Greetings. I would like to ask, how could i set up a GPT-2 for myself to have my own chat AI

[split ingot]

I highly recommend not contacting him. He seems to be promoting a suspicious Play Store app, gut instinct is it infringes on data privacy and/or installs malware on your device

[stone condor]

ik lol, that's why I steered it away

if he is malicious, I won't bite
if he's innocent, I'm not disrespecting him

[cedar belfry]

is it me or gpt 4 has a smaller context. My latest message I gave it, it says that my message was cut off when clearly it wasn't. Upon following up, it will not even remember any part of that message on the very next response/

basically, anyone else facing context length issues or?

[split ingot]

C’mon now, it’s discord. You’re supposed to be hostile and disrespect people outright until they’ve earned the respect of being treated as a fellow human :^)

[stone condor]

fair

STOP TRYING TO GIVE ME YOUR MALWARE FILTHY SCAMMER

GO DELETE YOUR EXISTENCE

NOBODY WANTS YOU, YOU WASTE OF OXYGEN

GET RID Of YOURSELF AND MAKE OUR SOCIETY BETTER

is that better

[split ingot]

Ahaha that’s more like it, great job king

[stone condor]

I feel... like I belong here now

Discord is my true home now

[rigid musk]

I suggest to avoid caplock. Can lead to a mute by automod

[stone condor]

lol

yeah, but I know a way to bypass it

[thick tusk]

you forgot the animated gif 😉

[stone condor]

automod got mad so I uh... did a thing

[stone condor]

hm...
anyone know some outdated memes?

let me grab one from the 1700s...

[thick tusk]

rick-roll?

[rigid musk]

To avoid bad gif, troll gif, etc.

[simple goblet]

It forbade Wikipedia too, I was gonna send a link to the "complaint tablet to ea-nasir"

[stone condor]

oh I just realized I can send images lol

[thick tusk]

I've encountered that too, it blocks gifs that Discord itself found

[simple goblet]

Discord just posts a tenor link when you find a gif

[stone condor]

now for Discord to be a blacklisted link :P

[rigid musk]

People wasn't using gif for the right reason and unnecessary so they just not authorized it

[simple goblet]

Fair enough

[stone condor]

anyways uh

is this chat just dead?

[rigid musk]

What was that ,

Like the thing you tried ?

[stone condor]

yeah

that's the error message I get

no matter what I type, always that error

[rigid musk]

Can you try firefox as test ?

[stone condor]

it's just that one chat too

[rigid musk]

Just to compare chromium product vs non chromium

If you experience the same

[stone condor]

me after leaving my house literally like 5 minutes ago

[rigid musk]

X)

[stone condor]

nvm, I already have the app

[thick tusk]

it works in the app?

[stone condor]

same response

errors every time

[rigid musk]

I suggest to post on #1070006915414900886 openai often answer. Or [e-mail] they can see what's going on

Openai employees name are yellow in this discord

[trim halo]

Can I suggest a different Ai bot that I think is cool, considering everybody here is ai enthusiasts I think it will be liked

[rigid musk]

You can here #ai-discussions

[trim halo]

perfect thanks

[tulip pebble]

i am trying to use image variation
but i am getting errors

code:

@parent_command.command(name="variation")
    async def sub_command(self, ctx: commands.Context, image: discord.Attachment) -> None:
        """  shows the help command of feedback  """
        response = openai.Image.create_variation(
        image=open(image, "rb"),
        n=1,
        size="1024x1024"
        )
        image_url = response['data'][0]['url']
        await ctx.send(image_url)

error
Command 'variation' raised an exception: OSError: [Errno 22] Invalid argument:

its like it doesnt like me sending attachements

[stone condor]

ty

didn't know if it was a bug (or intentional, since the chat may be too long by now), but I sent it anyways

[tulip pebble]

does open ai have img2img ?

[thick wadi]

Should I sell my goev stock?

[crisp root]

Requiring linking our Discord account to our OpenAI account is terrible for people who care about privacy

[rigid musk]

Can learn more here #verification-help

[onyx kraken]

A few channels prevent images being shared, but images are not in and of themselves, against the rules.

I do recommend deleting this image though, because: " Rule 10  Do not bypass the AutoMod filter.
Avoid attempting to bypass our server filters through the use of similar or meaningless variations of words, as well as by incorporating unique characters. Using your server profile to bypass the filter may also result in infractions."

Since the automod blocked that specific url, and the url is in the image, that does appear to be a form of automod bypass.

But I'm not a mod and I don't enforce anything.

#server-rules

[stone condor]

yeah good point

was just the

"Helicopters were invented in 1939
People before 1939:"

meme

[onyx kraken]

What I do with images when I want to share them from the internet, I usually screenshot, trim, and post. Noted, there's a few parts of the discord where images can't be posted, but that's how I handle that, and I'll even run off to #off-topic or #ai-discussions, whichever is more relevant, to post an image and link back to whomever I'm talking to, often link to their comment too so anyone can follow the conversation, when I'm interested enough.

[eternal iris]

No

[crisp root]

I didn't say I was confused. I said it's terrible

[paper hinge]

Hi, I’m new here and have a question about gpt4. I have a subscription to plus, but keep getting asked to upgrade to it in some browsers like micrsoft edge, but not in others like safari or google. Why is this?

[eternal iris]

don't remember doing that

[stone condor]

yeah, I've lost count of the number of times in various servers that I've put important images in #memes lol

[eternal iris]

According to the OpenAI community forum, there are other users who have reported the same problem and are waiting for a solution. I don’t think it has to do with the browser they use, but with their OpenAI account. The best thing they can do is contact OpenAI’s technical support and explain their situation. They will be able to verify their subscription and solve the problem as soon as possible.

I'm not suggesting that you have to heard what i say i may be probably wrong, I'm not a administrator, so you must be got contact with one

[royal patrol]

I keep getting an error when trying to verify my account...and I'm a chatGPT plus member

[onyx kraken]

The FAQ suggests: "Q: I can’t seem to verify, what can I do?
A: If you're experiencing difficulties with the verification process, consider turning off your ad blocker and attempting to verify again after a short period of time."

Not sure if that might help you.

[astral sparrow]

Hi. Is there an ETA for enabling custom instructions in EU?

[onyx kraken]

Not yet. All we've been told is: "This feature will be available in beta starting with the Plus plan today, expanding to all users in the coming weeks."

And "This feature is not yet available in the UK and EU."

https://openai.com/blog/custom-instructions-for-chatgpt

[paper hinge]

Thanks ZaRu. Problem is I can’t even find on OpenAi’s website how to contact them. Ironic how cumbersome it is to get in touch with a developer of an AI that is supposed to be intelligent and make life easier!

[onyx kraken]

Have you tried through the chatbot link here? https://help.openai.com/en/ Bottom right, looks like a square with a smile.

Within it, I'd click 'Send us a message' and type away! And then be patient, but the human in the loop will help you when possible.

[paper hinge]

Yes, it just takes me on a bot journey to the wrong path… with its pre-programmed solutions, and the one I have is not part of them. Pretty ordinary really, because surely this must not be just an isolated problem that is occurring, but must be more common amongst users. So for the system to not have integrated it into their queries is not positive.

[onyx kraken]

From what I understand you can leave a message that a human will read and respond to. They warn it usually takes a week, they have a backlog.

[paper hinge]

Yeah I thought i saw that. I’ll just wait and see what happens.

[onyx kraken]

You can do that! I see I can write to the bot, after following the path, even when the bot has ended the conversation:

And what is written, from what I understand, does get read by a human, when possible.

[paper hinge]

👍

[slim kayak]

Hello. im trying to develop a python program that takes two datsets 1 with all True entries and 1 with all fake entries and using openai with the davinci 3.5 engine i need it to combine and randomize the test data. Then print the tested data with the corresesponding prediction classification. THis model is a fake news detection model that will train the model with 80% of the data and then test it using the other 20%. Im having trouble finding a completions URL that will call without error.

for example here is the output result im getting: Error: 400 Client Error: Bad Request for url: https://api.openai.com/v1/engines/davinci/completions

i have code if we want to delve further into this im just looking for help. thank you

[native ether]

I'm not sure "davinci 3.5" is a real thing😅

[exotic dome]

Hi. I'm new here. Can anyone help me figure out how to remove plugins added to GPT4? I have a few that aren't showing up in the installed list and aren't working well.

[onyx kraken]

Hey there. Did you install this through the web interface of chatGPT, https://chat.openai.com/ and using the plugin store through ChatGPT-4's plugin store options, or did you select and install them through some other way of finding them?

[mortal portal]

Is openai API server down? I'm not getting anymore responses

[exotic dome]

@onyx kraken Installed through ChatGPT4 plugin store and two from the "there's an ai for that" site.

[onyx kraken]

Do you have any idea which are causing you trouble?

The ones from the ChatGPT website directly, the plugin store should behave themselves and be easy to remove through that, almost the same as how you added them.

Here's the process for that, you just go back to the plugin store, click on the 'installed' tab at the top, and then click the grey 'uninstall' tab on any you want to remove.

Anything you installed yourself might be anything, and I don't have advice about that except maybe turn to your antivirus for help seeking and removing potential malware.

Any program that is 'hiding itself' doesn't sound friendly.

[rugged wing]

LOUD PONY SQEAKS

Hiiiii.

[exotic dome]

@onyx kraken My thoughts exactly 👍🏽the AV worked to remove it, and I was able to get the other two uninstalled through the plugin store. I appreciate your help 👌🏽 think I'll stick to the plugin store from now on 😀

[onyx kraken]

Whew. Hope all is truly well for you now 🙂

I too stick to the plugin store, and haven't detected any problems yet.

[exotic dome]

@onyx kraken This was the first time I've ever had an issue 🤷🏾‍♀️

[feral jasper]

Why does the chatgpt plus text look like this, shouldn't it be more centered?

[rugged wing]

I'm hyped to start using the android version of chatGPT

[bronze charm]

It is now definitely higher than it was before but it's a trivial UI decision

[static aspen]

i have a question

about neural link

elon musks thing

[violet stone]

I have a question if anyone could help if really appreciate it

so I like to make these collages from music videos using stills from the vids

Am I able to use the ai to “copy” my style and make my work in less time if I were to show it previous work

that way i can just do small edits from the work it provided for me

so like this

[onyx kraken]

Probably best for #off-topic

[onyx kraken]

From what I know, currently there's no OpenAI product that can read images like that and make variations of it.

[tardy fable]

why chat gbt forgets messages which he's send, when I ask him to send 100 words, and I ask him make a test he do them but he repeats him self quite often that word above 30-40 he don't put to a test

[onyx kraken]

He can't count, he just has to guess how to answer best.

[acoustic pecan]

hey y’all. wondering if anyone here has had a similar problem. i had to delete my first chatgpt account because that email was going to be shut down, but when i went to make a second account it says the max amount of accounts have been made with that number even though i’m only trying to make my second? online it says that we can make two accounts per number, so not sure what’s happening. thanks

[tardy fable]

no, he send 100 word, I ask him make test out of japanese words, and I translate to other english or vice versa,
he does 1 test normaly, but after most of the word repeats and 3 looks exact the same,despite he should know there different words

@onyx kraken

I want that chat gbt, send 100 words and he make test more randomized because he repeats him self in every test, if I ask him more words I'm scared what he would remember from previews list how to make him make test more randomized and make him remember all lists

[serene vault]

hello, how can I search through my prompts like I can on the chatgpt android app?

[eternal iris]

Many times administrator's of discord are not able to solve problems that are not from the server, so i hope you find a solution soon. Maybe just we should wait 1 or 2 days. But the way

[iron goblet]

why i got timedout?

ok, I'll ask again.
Does anyone knows how to remove enumeration from response?

I've explicitly said in prompt to remove the enumeration, but its not doing it.

[violet stone]

Damn would love for something like that rn it would make my work easier :/

Probably best to outsource for that then

[royal patrol]

Thanks for the reply but, this doesn't help because I don't have an ad blocker

[tribal talon]

What error does it give you?

[onyx kraken]

Understood! I suspect sending a direct message to modmail (top right), to the server mods here through it, might be appreciated. You can share your error and they can help discuss solutions, I expect.

Whoops, or you can just talk to Baugrems, who I didn't see had answered 😄

[tribal talon]

Most common errors seem to be: Login method (google, microsoft, email/password), ad block/vpn issues, openai account issues such as bans.

Or alt accounts because another discord was verified under your openai

[timid hatch]

Hi All, any news or ETA on Web Browsing? THis is the main reason I paid ChatGPT Plus....

[bronze charm]

Not yet

Bard, Perplexity, and Phind are pretty good in terms of web usage

[jolly karma]

Why is openai so strict on account associated Email's and Phone number limitations?

I ask, because "anti-fraud and abuse reasons" does not make sense to me. Both email and phone are prone to attacks or may change over the course of a lifetime. This could likely result in an increasing amount of people not being allowed to participate in this product over time.

[timid hatch]

@bronze charm thank you!

[light belfry]

Do custom instructions count toward the token limit?

[slender meteor]

👁️

[native ether]

Yes.

[tiny wasp]

Does anyone else have the issue with whisper where it just repeats the same text continuously?

[serene moon]

Anyone else having issues logging into platform.openai.com ?

I'm constantly getting an error message.

[short axle]

for me openai in not installable on play store/ google play? it says download when available. has that something to do with the country i live in?

[native ether]

Most likely. Did you read the #announcements ?

[shy zenith]

I am a plus user but don't see the Custom Instructions option.
Why is it?

[barren magnet]

Can anyone give me a pointer towards info that I could use to set up an AI Chatbot on Discord as a Digital Therapist , (start to scratch, for complete idiot)

[shy zenith]

Okay i just checked that is not allowed on EU
damn

[rigid musk]

Soon

[rugged wing]

When ChatGPT will be avaible on Google Play?

[rigid musk]

This week

[rugged wing]

Can you say which day

And is it avaible in Bulgaria

🇧🇬

[rigid musk]

#announcements keep an eye on

[rugged wing]

Ok

But for bulgaria?

If I am not from countryies in #announcements can I download it for android?

[rose glen]

nope. no timeline yet but they'll announce when it becomes more widely available

[rugged wing]

Thanks

What programming language was used for chatgpt

I want to make AI too

[rose glen]

i do not know that. C++ is a good bet. ||and who knows maybe they even used fortran...||

[pallid cedar]

Hi,

I wanted to create another ChatGPT on my office laptop using my office email. I used my phone number for verification, but I received an error stating that I had reached the limit (2 accounts per phone maximum). I don't recall using my phone number for a second account.

I'd like to remove any account that is connected to my phone number. Could you guide me on how to do this?

[umbral rapids]

perhaps python I think

[rigid musk]

What is your country ?

[cinder girder]

Iraq

[rigid musk]

Did you delete account and want to use again your phone number ?

[cinder girder]

No its my first account i just created

[rigid musk]

It is a real phone number ?
Not voip

[rustic marlin]

Im going to delete that due to potential PII

[warm sierra]

Who is head here

[vagrant plover]

Could you elaborate on what you mean by that?

[warm sierra]

Means who is the creator of this group

I need to talk how to handle this group s

[vagrant plover]

This Discord Server is owned by OpenAI, the moderation team are all volunteers

[warm sierra]

Ok thank u so what is this opening questions

[vagrant plover]

This channel is dedicated to asking and answering questions about OpenAI products and services

[warm sierra]

Sry openai

[cinder girder]

What is voip? And yes its my real number im using right now

[warm sierra]

Idc

[rigid musk]

Like given by website or by business ip phone etc not a real cellphone number

[warm sierra]

What is that red thing on ur name

[rigid musk]

He is a mod

[cinder girder]

No its real legal phone number

[rigid musk]

I suggest to contact [e-mail] if you have issue with your phone number trying to create account

[warm sierra]

So many people are typing

[royal patrol]

I'm good to go. Even though it gave me an error screen, I just tried to verify the account just n ow and it told me that my account is linked

[rugged wing]

Hi lugui!!! HRU doin?

[rustic marlin]

Hello õ/
doing fine, I could wake up earlier today, good for productivity

[rugged wing]

Hahaha

[rugged wing]

hi everyone i have a question why open dont create a search engine .

[rigid musk]

Maybe one, day for now they work with Microsoft for bin search inside openai

[rugged wing]

I am sure if they make it i will pass google in user and i will be increbible

[wraith zodiac]

hi, why is the new human check so complicated and why are there so many?

[rugged wing]

yesh i notice it to. i think becuase bot are increased after ai is so widely used now.

[rigid musk]

Often the issue is the chrome user have with bad addon or extensions

[wraith zodiac]

its crazy! there should be a list of extensions then you cant use. i have harpa and another extension which adds copy buttons for quick copying. ill try remove and see if makes a difference

[rigid musk]

Try firefox as test to compare

If that's work you have bad cookie / cache or addon / extensions

[trail cairn]

i dont think you are officially allowed to use any

[wraith zodiac]

really?

ok didnt know

happy to remove them then

[trail cairn]

if they would maintain a list then it would probably be more like which are allowed and not the other way around

[wraith zodiac]

i would be happy with that too

[wanton bay]

What could be the problem?

[swift yarrow]

Why is there a constant challenge today for each and every prompt, even if you log off, back in- No VPN, no Firewall but a constant 1 to 5 items

[trail cairn]

disable brave shields

[swift yarrow]

Ahh.. weird, since it has never been an issue till today. But good to know. I appreciate it. - Worked!!!

[swift yarrow]

No reactions it seems, but I just wanted to say, I appreciate you! 👊🏾

[trail cairn]

👍

[safe pilot]

Hello can i like get one time credits? which dont expire

or is it just monthly usage

[trail cairn]

no https://help.openai.com/en/articles/4936830

[cunning sluice]

Has the context length of ChatGPT been expanded? What is it now?

[safe pilot]

I didnt mean those, like can i buy one time credits

Like i bought 50usd worth of credits at one time

[trail cairn]

no

[safe pilot]

oh, so its just monthly usage?

[eternal iris]

you can use te API and just use it when you want

[trail cairn]

if you mean api, yes its only available as monthly billed service

dall-e is based only on credits

[safe pilot]

well yea like the text models

[steel dirge]

When will the Android app be available?

[eternal iris]

They are going to be enabling it to different countries one by one but I can't tell you for sure when it will be enabled in your country.

[steel dirge]

Why has OpenAI decided to wait for so long with publishing their own apps while so many scam-apps have been released?
It's not difficult to basically copy-paste the website into an app... xD

@eternal iris BTW, you're not verified!

[eternal iris]

I'm not saying it should be someone I trust, I don't know and I'm too laz to verify myself.

[rigid musk]

It take time to adjust, troubleshoot, make sure it fit the 1000 Android per year, support, haptic. Make sure it work with the new features. Scam app just just api without testing anything and make profit

[rigid musk]

#verification-help user not verified will be remove from this server

[rustic marlin]

I'll be saving this gif, thanks

[onyx kraken]

Everyone gets to make their own choices.

I suspect we'll have a minimum chat of some sort still available for the unverified. After all, you have to have an OpenAI account to become verified.

If a person comes to the OpenAI Discord because they are struggling to make an OpenAI account, but manage to find the Discord (that seems possible, no matter how rare. The other day we had someone here motivated to try and help his kid with a ChatGPT-enabled game mod installation, who had never used discord before and also didn't know how to find their web browser or use a search engine. It happens, rare as it is. The child probably helped their parent find the discord in the first place) -

That person cannot verify, because they don't have their OpenAI account set up yet, and are looking for help. I for one won't mind trying to see if there's ways to help those folks, and I suspect the unverified will get some tiny corner of access, perhaps. (But I don't control those decisions, I'm just guessing and hoping).

[eternal iris]

I totally agree but that would only make a division between people, it is already predicted that we like to clasify people.
But just in case, I already checked

[onyx kraken]

shrugs Various protections exist for various places and situations.

OpenAI does indeed have the right to create and maintain an official discord, which they can request those wishing to remain as members to verify they have an OpenAI account. That said, those members absolutely have full right to do so, or not.

You get to make your own choices. Some choices have consequences that are not always delighted in, but this Discord access has varying value for different folks.

There's people who use OpenAI services who have no interest or value in the Discord as well. It's perfectly fine not to choose to comply with any of the rules to stay here (not doing so in an active way just tends to get one removed, like #server-rules violations).

I hope this server does have some foothold for those not verfied, if only so they can come decide if they might want to become verified or try to get advice with 'can't set up my OpenAI account' issues. But doing that probably means accepting some extra headaches for some of the volunteers here. Be interesting to see how the decisions go.

Not sure how racism comes into play with this, though.

[eternal iris]

I don't think the word "Racism" was the right word, I didn't mean it that way, but yes, a lot of it is up to Open Ai and its public to do things if they want to.

[still lintel]

do I understand correctly that most people cannot verify yet? the announcement texts are a bit confusing

[regal lynx]

anyone is allowed to verify

[still lintel]

I feel like the process a bit obscure, esp considering that the server has 1.7mil members:

• you have to read walls of text to navigate your way to the correct channel
• in that channel, you have to find a message with a verification link, which iirc isn't the last
• for those reasons, it might be confusing whether or not you're eligible for verification yet
• when you finally get to the verification page and it asks you to connect your account, it too is confusing because, well, I wasn't logged into my account at /playground at the time, yet it still said that the verification was successful?

perhaps there could be something simpler?

(ping) you need to connect your discord account to your openai account! click this big button to do so (link to a series of messages explaining the reason why)

[regal lynx]

that's why we have #openai-verification and have been pinging people in it 🙂