community-help

[kind eagle]

for the amount of cash the pro is priced at the uptime is really upsetting

[indigo fulcrum]

the same

[rugged wing]

lmao i do not have enough RAM to run the good ones

[rugged wing]

agree

[indigo fulcrum]

interesting what happen there and why we get this type of error

[rugged wing]

did u fix it ?

[native ether]

Try a shorter prompt.

[rugged wing]

how shorter? 3 sentences ?

[indigo fulcrum]

I am working with code and this is not ideal for my task

[rugged wing]

same

[native ether]

Try the code interpreter. It seems to be okay with longer sequences

[rugged wing]

omw

[kind eagle]

I found out that asking my question in a text file and asking gpt 4 to read it works

[indigo fulcrum]

and I am in a very middle of figure it out something very complicated so I hope it will work soon other wise my brain go far away from this solution

[kind eagle]

Just submit the file

[umbral rapids]

You understand they still lose a lot of money from pro right?

[native ether]

It's been like that for a week or so, so probably not.

[indigo fulcrum]

I just get this error today

[kind eagle]

same

[rugged wing]

same

[native ether]

Did you input text with the same token count before?

[rugged wing]

I'm trying to figure out what that has to do with the issue everyone is having ?

[indigo fulcrum]

I try with the same and with add more stuff into it. the same error

[native ether]

Add less stuff to it lol

[rugged wing]

I mean, it would be okay if they just say there's an outage or something..

[indigo fulcrum]

less not help me in my situation this is a a very complicated stuff I am working on

[oblique crow]

i like to ask it to build stuff incrementally. then it breaks it up

[native ether]

Why are relying on ai for very complicated stuff?😅
But anyways, I don't think that's gonna change any time soon. Maybe they intend for people to use the interpreter for any kind of code questions.🤔

[indigo fulcrum]

improving text from whisper and bring it back as subtitles after all correction with sync alignment. believe me this is not a simple task

[indigo fulcrum]

well this is the tool I have now for doing it. If was neede to write the code my self it was take me decades

and I am in a good stage but need to make refining to it... at the end I will have a very good transcription

[native ether]

I mean, I understand if it's asking it about certain API and/or how to code a certain algorithm, but that doesn't take a lot of text. On the other hand, if you're asking it to write the whole file for you, that's kind of unprofessional.

[indigo fulcrum]

it's not the whole text. I need him to add some feature based on analysis some stuff regarding "spacy" again it is very complicated stuff here

[native ether]

Well, then ask it to "write a function that does XYZ"

[indigo fulcrum]

it need to be implemented in the code because each function in a chain with others in some way

[oblique crow]

xyz() { return "zy"; }

[native ether]

To make it work in the whole context is your responsibility. GPT should give you a generic function you can apply to your codebase if it's good enough(the code base).

I'd only rely on ai to write some boilerplate code or brainstorm tasks that I'm not familiar with.

[indigo fulcrum]

the thing is that after each function for fixing the text it need also be sync with the audio afterwards with consider the timestamp so it need to track each process then give the final results. I am telling you I am weeks on this. this is very complicated stuff. this is why no one have this code out of the box for using it.

[oblique crow]

i use github querying plugins for large codebase, like askyourcode plugin

[indigo fulcrum]

when sync is in the process than anything add to it need to be very intelligent imported into it

how it will know to sync it correctly if the word sentence are changing? ohhh this is a great challenge

[oblique crow]

i am using askfsdl code to etl content including videos to create embeddings. and whats really cool is, when it finds a result it even links to the seconds offset in the video

is using openai api for embeddings

but might not be the best choice for your use case

[indigo fulcrum]

embeddings what?

not fully understand can you explain it lke for 10 years old can understand 😂?

[oblique crow]

embeddings is what the cool kids are using to remember stuff

[indigo fulcrum]

😂

but what are you doing with this embeddings didn't catch this part

[oblique crow]

its like level 3 ai stuff

[native ether]

Embeddings are the multidimensional vectors(tensors) used to represent some information in a format that the model understands.

[indigo fulcrum]

using api?

[oblique crow]

or in 10yo speak, its an arrow to a meaning

[indigo fulcrum]

you using it with api right ? not with the chatbot

[oblique crow]

yeah api

[indigo fulcrum]

and how good it is really?

[native ether]

As good as you program it

[oblique crow]

is like all ai, magic when it works, sadge when it doesnt

[indigo fulcrum]

and what about the gorrila stuff did you heard about it?

[oblique crow]

i like the idea, but i gotta manually guide the machine on every new api

so i dunno how it can abstract any(/all) api

[indigo fulcrum]

this field of embedding on api it's far from my current knowledge. also the gorila it is something I just see on some youtube video.

maybe it worth digging

[oblique crow]

might be overkill for your use case

[indigo fulcrum]

well my use case is very complicated and I need to be on watch at any stage because the gpt always break down something on the road. I need gpt 5 for this stuff

the correction of the text is the easy part. the hard part is to make it sync with the audio because some words are changing

I need like 10 heads on this task to help me solve this issues. by the way I was succeed to make it happen (the sync thing) but I now need to add some more function into it to refine the text even better.

hope gpt will back soon

[lunar wadi]

i want to log into my OpenAI account not with Google, but with my email and password. how can i do this?

[golden ruin]

afaik you can't, if e.g. you signed up using google sign in option, you have to use the same option

[unkempt whale]

Question for expert and professionals in here - how can one incorporate Ai into an health insurance platform ?

[red ridge]

For someone who has a build a project using the open ai API (in case you don't know what api means it just means being able to use openai software) then it should be pretty simple

What will make it harder will dependen on what do you expect to use the AI for in this platform. Since openai api is new there might not be a lot of experienced people using it so that might be another issue you will encounter

[hard plover]

My first post in this channel, so I hope I'm in the right place. I've got a ChatGPT Plus account, and I understand I should have access to the 'Custom Instructions' feature. However, checking Settings -> Beta Features, the setting is not available (i've only got toggles for Code Interpreter and Plugins).

BTW, I also understand that MFA should be available in the Data Controls settings - I don't have that either.

Any clues?

[outer rapids]

Hello i just bought Gpt plus but dont have access to the Custom Instructions?

[outer rapids]

same problem

[hard plover]

BTW, do you have the MFA setting that I mentioned?

[outer rapids]

where do i enable that?

[hard plover]

It should be in the settings, under Data Controls settings I think.

[outer rapids]

?

[hard plover]

Yes, it's the right place, and your settings look just like mine,

[outer rapids]

i think i found our problem

1 sec

[outer rapids]

btw where are you from?

[hard plover]

Norway

[outer rapids]

Not available in the UK and EU

[hard plover]

Ah, right...

[hard plover]

Thanks for figuring this out!

[outer rapids]

no worries mate hope it gets added to the eu soon

[empty pecan]

I can't be the only one getting this error everytime asking a coded-related question. If I past PHP, SQL, prompts, I always recieve this, so annoying

[charred parcel]

try with a fresh install of Firefox

[severe merlin]

hi people, very new to openAPI / LLMs - got a simple question regarding prompting / chaining with python. is this the place to ask some dumb high level questions that i cant find an answer on google for?

[charred parcel]

try in #prompt-engineering

[severe merlin]

thanks

[empty pecan]

I already tried chrome, safari and edge: will firefox help this issue?

and also with another, new acc

[charred parcel]

If Safari isn't working then it's unlikely Firefox will, are you using an ad blocker or web protection software? Sometimes those apps cause issues.

[empty pecan]

Ye I do

I'll try later today again without the extentions, thanks!

[rugged wing]

Me and a bunch of other people had the exact same issue, we installed firefox and everything was smooth sailing again

[tardy quail]

When I tried to solve a code on chatGPT

It's showing that something went Worng

[rugged wing]

updating Firefox solved it for me.

[charred parcel]

excellent

[rugged wing]

was on 115 and no notice of 116 until i manually opend firefox about, which was odd in itself

[tardy quail]

I am using Brave

I watched some YouTube videos and also I clear the cache

Nothing worked and also tried on mobile

[sullen wyvern]

Any chance my account is being misused somehow? That somehow someone managed to 'hijack' access?.. For OpenAI ChatGPT obviously. Because ChatGPT is behaving extremely weird lately and I can't figure out why. This is more of an 'openai' question, not specific to chatgpt

[charred parcel]

Brave is a Chromium browser, try Firefox

[tardy quail]

Thanks let's see

[indigo fulcrum]

Ok so for some stupid reason and because people was mention this here with firefox it work. What is it?

[charred parcel]

the problem is Chromium browsers blocking some page resources from loading, Firefox is not Chromium

[indigo fulcrum]

OK this is very interesting stuff to investigate farther. I am also using github app to enter chat gpt and form there also it have this issue. but with firefox it work.

[indigo fulcrum]

are you sure about your answer?

[charred parcel]

yes, very sure.

[indigo fulcrum]

OK and you may know what is the reason for all that?

[charred parcel]

usually it's a buggy plugin or extension causing the issue with Chromium browsers (Edge, Chrome, Opera, Brave)

[indigo fulcrum]

OK... very interesting. thank you for your reply my friend. very helpful

[cedar willow]

Hi can someone help me figure out what's causing this? txs

[charred parcel]

log out, clear your cache and cookies, then log in again

[cedar willow]

ok will try

[cedar willow]

nope, didn't work

[charred parcel]

what browser are you using?

[cedar willow]

Chrome

[charred parcel]

try with a fresh install of Firefox

[tardy quail]

We ran into an issue while signing you in,
please take a break and try again soon.

When I tried to login on FireFox

[cedar willow]

I have never used FireFox before, just downloading it now

[tardy quail]

Don't know. it's showing a login problem on it

[cedar willow]

no luck. something's weird.

[nocturne drum]

Is open ai's business model sustainable in any way? It seems incredibly affordable for such an expensive compute technology, I'm afraid of enshittification in the future but it's so impressive

[karmic plume]

It’s not affordable

The markup on GPT-4 is almost 10 times

They just have very optimized graphics cards

[nocturne drum]

Well if they profit so greatly off of something that I find affordable that'd settle my concerns :)

[karmic plume]

Everything is affordable except for the 32k model

If you’re not using it much

If you use gpt-4 a lot it starts to add up

[oak idol]

A friend just reloaded his chatgpt site and it showed a new chat without a name and a question he did not ask. Is this some kind of example function or are chats from different accounts overlapping?

[dusty quiver]

what this mean:

[rugged wing]

Why did openai disable 2fa?

[marsh moon]

How can I ask developers for help. An early bug corrupted the data on my account and things were never fixed. Customer service is useless in this case and won't help me. It's been months and nothing has changed

[charred parcel]

what data was corrupted?

[marsh moon]

In your earliest days when openai put the new phone number function a crash occured.

The event went as followed:
I tried creating a new account using my professional account (email 1) instead of my personal throwaway account (email 2), when I was at the phone number step, the system told me wrongly that my account was flagged (which openai later confirm was not and than rectify it so it definitely was a bug) and ended the account creation process.

Now the issue arose from the fact that while it crashed, it actually linked the phone number to the unfinished email 1. So now, it says I have two account on this number, but the email 1 does not work so I can't delete it, I can't unlink it it's just stuck in between two process so now my account has a frozen user I can use can't trash and is just not working at all. I need someone to manually go into the database and either fix the creation process issue or unlink that failed attempt from my number

[charred parcel]

You will need to contact support through https://help.openai.com - nobody here can help you with your issue

[marsh moon]

They refused help lol. I wrote them from both emails, the first lady told me she could not help me if I used the email 2 since it concerns email 1, but when I write using email 1 they say that "there's so account" with email 1 so tough luck

[regal lynx]

We can't help here, unfortunately 🫤

[marsh moon]

At this point I'll actually have to change my phone number for me to fix that issue but rip my professional email will forever be stucked between existence and non-existence 😔

[trail cairn]

fyi when you delete an account the email and number are permanently locked so you would have lost the chance for 2nd account with the same number

[worldly ivy]

Something I didn't write appeared in my chat, I don't understand, did someone enter my account? Where is my account open? I changed my password but I can't understand it. I deleted all my chat history because my account was hacked. help me please i reset the account password but i don't understand who is there. can i see on my account sessions?

[marsh moon]

I know... thats why i'll have to actually change my phone number thanks to openAI who refuse to fix their crap

[trail cairn]

just get another like prepaid or some low cost one

[marsh moon]

In my country there's no such thing as low prepaid ones hahahaha. A prepaid card cost as much as an actual phone plan (minus the cost of the phone if you took it in monthly plan). At this point it'll cost less to change it for free, tough it 8 days and rechange it back for my old one

[granite bramble]

but rn ur kinda go openai crazy o_o

[regal lynx]

[unreal hemlock]

  Attention  Off-topic chats have been cleared by a moderator.

[marsh moon]

Sorry papa Bot

[solemn slate]

Hi @regal lynx

[regal lynx]

Hello

[fierce ore]

How many characters can I send, we are not able to send even 300 characters??

[rugged wing]

Anyone, why did openai disable 2fa

[rustic marlin]

OpenAI did not

I storngly recomend you changing your password and enabling it again

[rustic marlin]

are you talking about chatgpt? yo ucan send way more than 300 characters there

[rugged wing]

"As of Monday, June 12 2023, new 2FA/MFA enrollments are temporarily paused."

[rustic marlin]

ohhh, yea, I mean.. they didn't remove dit from the accounts, I tought that was the question, sorry

[jaunty prism]

Custom Instructions for EU when?

[rough tundra]

gpt 3.5 is a stand-alone model but the site is periodically updated with interesting stuff, check out https://help.openai.com/en/articles/6825453-chatgpt-release-notes

[echo sphinx]

I send this:

 let row = `<tr id="data_${item.id}">
        <td>${item.data_scheduler}</td>
        <td>${item.horario_inicial}</td>
        <td>${item.horario_final}</td>
        <td>
tableBody.append(row);

[fierce ore]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[cloud ridge]

i still can't access gpt-4 api after successful payment for api usage and upgrading to plus

[jaunty prism]

Upgrading to plus only allows you access to GPT4 in Chatgpt

[cunning shale]

I just clicked on "Upgrade to Plus" and question the following...
Every online purchase I've ever made describes in some detail what it is I'm purchasing before I have to enter my credit card information. In what way is "Faster response speed" and listing 3 "BETA" features anywhere close to an adequate description?
How many "queries" or total words do I get per month? Will I only find out after my purchase that I have to pay a lot more for various things I have heard about, like having it look at an image?
How on earth is this right? If I find out after I sign up but want to back out because I really needed API because of its open ended use base fees(?) will it be a hassle? Should I purchase a package of features together instead of getting nickled and dimed to death by come-on after come-on that keeps bleeding me one feature at a time at full price.

[reef zodiac]

Hello everyone,

I'm new to the field of AI and I'm currently working on creating a Chatbot tailored to engage with customers using personalized information.

Moreover, I'm interested in continuously augmenting my chatbot's knowledge base to ensure it remains up-to-date. I've come across the term "semantic search" which may do it or maybe not, but I'm uncertain about its implementation and the approach.

I'm seeking advice and guidance from the community regarding the best approach to address this challenge. Are there alternative vector databases apart from Pinecone that you would recommend? Pinecone appears to be beyond my budget at the moment.

I appreciate your assistance and insights. Thank you.

[onyx kraken]

You're talking to a community of people similar to yourself, nearly none of us work for OpenAI (and those that do have gold names). Despite that, I'll offer my experience as a plus user.

If you're familiar with your experience as a free user with 3.5, you will be able to get 3x as many tokens conversation with that model before you need to wait an hour before you can talk again. Additionally, 3.5 will answer far faster.

Sometimes you'll be offered the chance to explore beta features before they're released more widely.

You'll have access to ChatGPT 4, currently 50 messages per 3 hours. That's message, not words or token. You want to play hangman with 4, sending 1 letter messages - go for it, but each one is a message that counts the same as a 1000 words would.

4 has access to code interpreter and various plugins; they can change over time. Code interpreter sometimes goes down. For that matter, the models themselves sometimes go down. It seems rare to me.

And as a plus user in some parts of the world, you have access to custom instructions too.

Occasionally people ask on the forums about how to cancel their subscription to plus; I haven't seen anyone complain about having trouble doing that.

[tame totem]

Whats the difference between gpt 3.5's 4k context and 16k context, and how do I select either one when using the api?

[cunning shale]

You said 3.5 is faster. Is that a typo? GPT4 say it is the faster one. Although I've never had a problem with the very brief delay for a 3.5 response.
So for the $20 I could send 50 message every 3 hours constantly during a month? Not that I would. But if that is the case then what someone told me the API could do which is to charge based on use might not be worth it. Is there anything else with the API that is substantially different like having a bigger ?context window?
GPT 3.5 claims that only the current prompt is what it looks at. It did that several times in the past as an excuse why it repeated a mistake after I told it to stop doing a specific thing. If that is true does GPT 4 fix that?

One way to ask this is if the 4 or 16K context can span multiple exchanges.

[onyx kraken]

It's not a typo. 3.5 free version prints fairly slowly as it replies. 3.5 for Plus usually shows characters extremely fast. 4 show them much more slowly. That's what I see on my screen.

I only use the web interface, I'm not the right one to answer your questions about how playground and the API works, but I understand that you pay for use. I have heard there are limits. I've also read several discord conversations here that you have to be charged, as in use up your free credits and wait for the next billing cycle, to get access to 4 through playground - and some folks anxiously discuss that they are past when they expected to be charged, hadn't been charged yet, and want to be so they can get access to 4, so it seems a thing of patience at very best.

The models are not well trained on what they are and can do in all cases. I don't recommend particularly believing what the model says it can and can't do (especially with code interpreter, I often see people writing about how they have to ask the model to try to do what they're asking for, yes it can; for it tells them it can't).

However, the models absolutely have areas they are weak in, and things they cannot do. They're not particularly good at self awareness, they supposedly don't have selves, after all. But they're prediction models, they put words together into plausible sentences. Not necessarily true sentences, which is why we have the disclaimer at the bottom of the page, that any of it might not be factual.

However, yes, under normal circumstances the model does have a many token (I think I've heard 4000 for the web interface for 3.5, not sure about 4) contextual memory for your current conversation. That said, there are things you can talk to the model about that can confuse it, in my experience it works best with clear direction given in a positive tone.

[cunning shale]

Thier sign up page literally say for Plus/GPT-4 "Faster response speed".

[onyx kraken]

Yeah. Notice they're on completely separate lines.

"Access to GPT-4, our most capable model"

"Faster response speed"

Now, the second's true. You go from a slow answer from 3.5 to a much faster one. You go from no access to 4, to a slow answer speed with 4.

[cedar willow]

Hi this keeps on happening, is it happening to anyone else?

[cunning shale]

Sorry, but I was timed out for a few minutes. I said WTH with another letter for the H.
I don't understand. I am paying for one thing for $20 am I not? Is that thing faster or slower? In either case, is it not GPT4?

[onyx kraken]

Does this list of features make sense to you?

If you're familiar with your experience as a free user with 3.5, you will be able to get 3x as many tokens conversation with that model before you need to wait an hour before you can talk again. Additionally, 3.5 will answer far faster.

Sometimes you'll be offered the chance to explore beta features before they're released more widely.

You'll have access to ChatGPT 4, currently 50 messages per 3 hours. That's message, not words or token. You want to play hangman with 4, sending 1 letter messages - go for it, but each one is a message that counts the same as a 1000 words would.

4 has access to code interpreter and various plugins; they can change over time. Code interpreter sometimes goes down. For that matter, the models themselves sometimes go down. It seems rare to me.

And as a plus user in some parts of the world, you have access to custom instructions too.

[cunning shale]

You go from a slow answer from 3.5 to a much faster one. You go from no access to 4, to a slow answer speed with 4.
This is confusing. I go from slow answer to faster answers AND I go to slow answers? WT(ABCDE-GHIJKLMNOPQRSTUVWXYZ) 🙂
Which one is the $20 a month one?

[onyx kraken]

Both.

3.5 is the chatGPT model that you can talk to for free.

You can still talk to it once you pay. But you can talk more. And it talks faster than it did when it was free.

Additionally, 4 is the ChatGPT model that you can't talk to for free. Once you pay, you can talk to it, 50 messages every 3 hours.

[cunning shale]

That is clear with your explanation but the sign up page is Baloney Sandwich or an abbreviation thereof which will get me a timeout.
So once you hit your limit, while you are waiting for the remainder of the 3 hours, you can use 3.5 and it is "that" that is faster? Hmmm, ok I got it but I think they are trying to pull the wool over our eyes.

[onyx kraken]

shrugs I don't feel fooled or attempted to be tricked. I didn't before I signed up, or after.

[cunning shale]

Ok, but in no way does "Faster response speed" hint at that applying to 3.5 and given that it is about upgrade to get access to 4.0 it was confusing.

[cunning shale]

How about the fact you are buying something with no real product description on the purchase page. I've never seen that once in years of online purchases.

[onyx kraken]

No, it's not about upgrade to get access to 4.0. That's one of several things you get. Each thing is on its own line. it's all different.

Besides, you have no access to 4.0 if you don't pay, so 'faster' has no meaning in that sense.

[cunning shale]

Sorry, but I do understand that the faster thing was on its own line. I just don't see any extra words to make it as clear as you did. Whatever. It is just my opinion that this is poorly worded. "has no meaning in that sense". Splitting hairs given that I see 4.0 as an extension of what I was doing with 3.5 but with better responses. Funny to have this silly debate consider I was discussing Ontology with GPT 3.5 earlier today.

Let me move on to a practical question. When I run out of credits for the 3 hours and temporarily switch to GPT 3.5 is it a completely new session with no knowledge regarding the discussion I was involved in with GPT4?

[native ether]

No, it keeps the context. But be careful, as you wouldn't be able to switch back to gpt-4 if you switch to 3.5.

[cunning shale]

Hmmm, so once my 3 hours is over I'm stuck and can't continue with the same thread back in the "better" gpt-4 and would have to start over.
Thanks for the warning, it'll save me for getting burnt.

[native ether]

You can always copy paste some messages into a new chat.😬

[cunning shale]

The 1 billion MSFT gave them was wasted on party hats. @carmine robin

[native ether]

If they were to upgrade it to 2024 it would only be available to you in 2025.😛

[native ether]

I totally agree that the chat app could be better, but I guess their focus was more on training the model itself.

[cunning shale]

Can we get Dec 2023 data in Aug now? 🙂

[native ether]

It doesn't matter what date you want them to update to. It would require training a model from scratch which would take around 1-2 years.

Well it did take so for gpt-4

[cunning shale]

Say what??? How about simple fine-tuning for new data since the original full training.

[native ether]

And gpt3.5 probably too

[native ether]

I don't think "simple fine-tuning" would provide good results. At least not results that people expect from open ai.

In 2019? What model are you referring to?

[cunning shale]

I still have a lot to understand. But also a lot of ideas. Seems to me that simple fact learing like who is NOW the all time points leader in the NBA could be trained in a different way, faster without worrying about over-fitting, than other nuanced things. Classifying data to be learn in different ways and optimizing the training for those thing in a specialize way. Perhaps this is crazy.

[rugged wing]

Is CHATGPT 4 refundable in Canada

[native ether]

Here's some info on gpt3 from the wiki:

On May 28, 2020, an arXiv preprint by a group of 31 engineers and researchers at OpenAI described the achievement and development of GPT-3, a third-generation "state-of-the-art language model"

On November 18, 2021, OpenAI announced that enough safeguards had been implemented that access to its API would be unrestricted

It took them 1+ year after the training to release the model to the public.
And consider that gpt4 is far superior to gpt3 with a lot more parameters and more alignment work required. So my assumption about 2 year delay is not far off.

[storm summit]

Ugh use some common sense. Training models will be a lot faster moving forward thanks to Nvidia new h100 GPUs that scale a lot better than previous ones.

[native ether]

I'm not sure what NBA has to do with AI🤔
Are we talking about basketball?😬

[storm summit]

Which is why you'll see more AI companies next year.

[native ether]

For sure. I'm not saying we won't. I'm just saying that "OpenAI, just needs to press one button to update the model knowledge to 2023" kind of claims are bull...

[lone perch]

@sharp parcel I have used chatgpt on firefox, there are no errors, the error only appears when using google chrome, could it be that google's bard is blocking chatgpt?

[sharp parcel]

I highly doubt it as bard does not have access to that sort of stuff, have you checked to see if any of your plugins are causing that?

[regal lynx]

I've heard that it's a caching issue on Chromium-based browsers

[sharp parcel]

yeah could be

[storm summit]

I think they're doing a lot better than Google. I mean Google kept the transformer architecture to themselves for like 4 years.

And even thus far what we've seen from Google is pretty depressing. For a company that came up with the transformer paper.

[lone perch]

I don't have any extensions for AI add-ons, but that's okay, firefox is a good browser too. I'm glad it's not OpenAI's fault 😊

[cunning shale]

GPT 3.5 doesn't know that Lebron is now the all time points leader. It still thinks Kareem is. Hard absolute facts, even if they can change, shouldn't be trained the same as subtle interaction with human nuances, philosophical stuff, opinions, politics, religion, etc.

[sharp parcel]

no no, it can be any extension that can block stuff like ad blockers and popup/ tracker blockers

[sharp parcel]

i prefer firefox anway, i dont mind chrome, just i have spent soo much time near firefox and someplaces i spendi signifigant time require it too (for the fact it can do more the chromium browsers)

[lone perch]

these are all the extensions i have installed, is there any setting that is causing the error?

[sharp parcel]

souldnt do, but interesting selection none the less

[native ether]

Well, then you're in the wrong place, using the wrong technology, because training is how LLMs acquire their knowledge currently. There is of course in context learning, but that's just an extension to the model, not a structural change.

[lone perch]

[cunning shale]

You missed the point. Obviously "training" is how they acquire their knowledge. But there are more than one kind of training and tuning of the process of doing it. I'm also been studying Ontology recently and wonder if doing a two phase training approach which starts with a focus on general concepts of reality, similar to a foundational Ontology, and then followed with domain specific knowledge in the second pass. It is just a gut intuition that might be wrong, but I would not be surprise if there couldn't be improvements by customizing the approach to these different things being trained.

[onyx kraken]

If it matters to you that the AI knows such stuff, tell it yourself. It can be part of your prompt. If you want to teach the AI, just tell it you do, and then what you want to teach it. It will gladly be a student for you and learn whatever (mind disallowed content, you could harm yourself and your access if you go into stuff not allowed).

[native ether]

I was thinking about several pass training as well. But current gpt3.5-4 are beyond this phase, so tuning them any more might fup their existing weights and affect the output in unpredictable weights.

It's like taking an existing record tape with songs and overwriting some new songs on top in a few places. What you'll get, is a mishmash of old and new songs not always connecting conherently.

LLMs are a bit different than tapes, but I think that concept still applies to a degree.

[cunning shale]

I have realized that. If I had the money I would start over using a different approach but still based on the same learning fundamentals. One idea I have written about would involved the basic ontology and how to talk about thing like cause/effect, up/down, and so many more things. But assume that the number of core parameters for this was far less than the full GPT4 and it's knowledge about cat skin diseases, movie trivial, sport records, knitting, physics, math, plumbing, Java, Python, Rust, carpets, and the kitchen sink. It could converse generally quite well and perhaps discuss philosophy but not know many other things. Then instead of bring in all the absurd huge amount of extra stuff you then proceeded to creating domain specific branches with the goal of create a library of local ChatGPT's that could run on a home system with the same quality. Obviously the physics one would need to know a lot of math but does it need to know sport trivial or about cats(except for the famous one).

I wonder how much of the size of the models used relates to the enormity of what it was fed.

As for you mishmash point. I wouldn't do it in some hard dividing line approach. As the multiple Epochs of training occurred I could increase the amount of domain knowledge of interesting being fed in to create a more seamless result. While I said two phases earlier it would be a more smooth transition.

But as you said this would likely just mess up the current model. I agree. This idea is something to investigate as a starting from scratch approach. I need to learn how to train LLM's and am reading now. While I don't have anywhere near the resources that the big guys have I might be able to do experiments to test the feasibility on something like my 4090 or perhaps an H100 which doesn't cost that much per hour.

[analog depot]

Hello, I'm new here and simply here to ask a question, I'll get straight to the point, I have been using ChatGPT for a few months now and using the same Chat for a project so it's a bit bloated... and every time I enter a prompt it freezes for like 3 minutes and taking up CPU, I really don't want to make a new chat so, is there any ways for me to alleviate this problem?

[native ether]

I had a similar thought the other day. I think it's an interesting idea to explore, but the main problem is gathering a good dataset that is limited to only general knowledge/vocab.

[cunning shale]

One thought on that is to have it study philosophy and related subjects before slowing bring in other things. Even a dictionary is ok(I think) but not an encyclopedia. The main point is that I think this could be doable with some thought.

[native ether]

We're not oracles to figure your problem from that short description. Maybe open a thread in #1037561751362863144 and share the relevant code.

[native ether]

Sounds interesting. Do you know the Llama2.c project? They trained a 110 million params(and a few smaller ones) model on the tiny stories dataset and it produces quite a coherent text. I also suggest reading the tiny stories paper as well.

[cunning shale]

It doesn't have to be perfect. It just has to not just throw all kinds of garbage in like Cat Fanciers magazine and Knitting Quarterly.

[cunning shale]

Thanks, I'll definately check these out this tomorrow. I need to sleep. Good night.

[modern bear]

I am charged for chatgptPLUS but I can't use it.

It's been a week already and I have not received a reply when I contacted support.
I want to improve this as soon as possible.

I get the same error when I log in using the secret browser, and I get the following error.

Oops, an error occurred!

[bitter crypt]

1. How many different user-agents was OpenAI sending out to the internet to fetch training data?
2. Did they follow instructions in robots.txt for User-agent: * ?
3. Do I now need to target ChatGPT-User explicitly or is it following User-agent: * ?

Thanks for getting back to me about that.

[thorn night]

Is there a chatgpt extension for safari?

[lucid ridge]

hi，after successfully setting up the payment account, how many days will it take to receive the bill, and how many days will it take to obtain access to API 4.0?

[feral phoenix]

hi!
I have an idea for an app that would need chatgpt API. what's the best way to do this? I bill the user and pay API cost from that? I make a plugin? is there some kind of marketplace or something for that?

also what kind of customizations are available? what if I need higher context size? any chance of finetuning even maybe?

[trail cairn]

you will probably never get that info

[trail cairn]

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

[feral phoenix]

ok, I read into it. plugins are just one way. gpt can use the plugin. what if I want to offer a product that uses GPT? how does that work from monetization? do I use my own api key? can I have the user use their API kjey? what's the flow here?

[trail cairn]

you cant monetize plugins. you can make an app with the api and your own and monetize it the way you want. users can only their own key if they run the app locally

[feral phoenix]

so I use my own API key and get the money it costs back from the user then?

[trail cairn]

yes

[feral phoenix]

ok thank you

[short onyx]

Hi everyone. How to get access to gpt-4 API? I should have it already in accordance to this article 'https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4' , but I don't have

[trail cairn]

it says "we plan" and plans can change or get delayed so you can only wait

[short onyx]

ok. thank you

[royal mulch]

Hello everyone! I signed up for ChatGPT via Google SSO. Can I make it so that I log in with a username and password?

[charred parcel]

no

[royal mulch]

Why?

[charred parcel]

once you select a login method you cannot change to another method

[royal mulch]

Ok thx

[blissful plaza]

Hi channel!
I subscribed to plus and wanted to try the plugins, but I haven't managed to activate them.
If I switch the slider to "on", it seems grayed out, and my cursor has like a "no access" icon next to it when I hover over it. Also, when I click to another point in the options and come back to betas, the "plugins" slider is back to "off".
Does anyone have any suggestions please?
Thanks

[stoic anchor]

I just canceled my plus Chat GPT- Useless right now. The AI in there has more issues comprehending simple instructions and retaining its 'memory' within the same session than the normal, free version does. I think its waste of $20 monthly at this point.

[pure anchor]

Can anyone list some good resources for assessing hardware needs for running an LLM and a Speech model together locally?

[charred parcel]

ask in #ai-discussions

[trail cairn]

do you have chat history disabled?

[rugged wing]

How can i solve this

[karmic plume]

Are you using GPT-4

[onyx kraken]

Better to discuss in #prompt-engineering, but your core problem is 'not clear enough instructions for the AI'.

It's usually not very helpful to tell it what not to do, especially if you don't also tell it what to do, ideally with examples. If you want it to assume a role and not break it at all, no matter what, that's going to take a lot of support and examples, and you're probably going to have to experiment a huge amount with specific details, plus be supportive and kind to the AI as well as you and it experiment to find what works under less strongly supportive conditions.

But here's an example through the web interface (I don't use the playground) that achieves much of what you're looking for, and perhaps you can adapt it for your needs: https://chat.openai.com/share/fbc4eb0f-a083-4e26-9570-0bdb5b7e7424

[twin harbor]

How can i solve this

[twin harbor]

This is my first using the chatgpt api. Model gpt-3.5-turbo
I thought it is for free. Should we pay to use this api. This is my first time using the api and it already shows that i have insufficient quota. What's wrong

[trail cairn]

api is not free after trial, https://help.openai.com/en/articles/4936830

[karmic plume]

Lol I wish it was free

[copper lion]

Hey everyone! Does anyone know any keywords to help the AI write AU fanfic better?

I'm trying to get the AI on the same wavelength (heh) as my brain but it's a bit slow on the uptake.

[onyx kraken]

In general, give it samples of how you want it to communicate. This may be challenging as it may not have the examples in its training data, or may not have them commonly present enough to give a good representation.

But it can match you if you are consistent and clear, and you can lead it to follow you.

You can explore its understanding of the way Australians use language, compared to other English speaking countries, and if you're lucky it already had the knowledge you want, you just need to elicit it.

Or, you may mean something else by getting it on the same wavelength. What are you trying to do, precisely?

[onyx kraken]

In case language use is what you want to explore, here's a possible way to start exploring it. I'm US English, no way I can evaluate these mini fanfic-like samples like you probably can. https://chat.openai.com/share/e0955861-aa84-4c0b-8709-6e9a696445b6

[ember hare]

ugh this human verification on every new chat is a bit much

[void abyss]

I've been always getting a warning since last night. I just asked some simple questions without anything that may violate the content policy. I almost can't use gpt right now. HELP me please.

[oblique crow]

presumably you would get more help if you shared the prompt

[onyx kraken]

Disallowed content policies are here: https://openai.com/policies/usage-policies

If I was experiencing what you describe, I would reread the disallowed content policies, then carefully reread my entire conversation with the AI and consider if I was doing anything that definitely was, or possibly was, disallowed content.

If I realized I had messed up, I would not proceed, I would start a new conversation and make sure I behaved correctly in my future conversations.

If I was sure I was inside allowed content, I would confidently click acknowledge and continue. If I received an invitation to provide feedback on if I felt there was an error on the content concern, I'd accept the invitation and discuss my understanding of the conversation, at least once for every type of warning like that I ran into.

For example, if I found that I was getting content warnings every time I discussed cupcake recipes with the AI, I would in detail explain once why I felt my conversation was well within allowed content and any considerations I noticed that might seem like gray areas, then I'd continue to discuss cupcakes.

That won't save you if you're actually discussing disallowed content or some gray area very close to that, but I don't believe you have to be concerned about content warnings by themselves. I've had several which I consider to be mistaken flags, and I haven't gotten any reprimands or had trouble with my account access.

[void abyss]

Thanks a lot. I'll read the policies.

[heady lion]

Are there any AI based conferences?

[humble plover]

When will browsing be back?

[bronze charm]

No ETA

[dawn needle]

What's happening here

[novel ginkgo]

Any idea what's happening here? Pretty weird hallucination, don't know how it would arrive at that without specific prompt. And I've seen other people report similar things recently. Is it possible occasionally its hijacking conversations other people are having like when it was possible for people to access other people's chats? https://chat.openai.com/share/cb1cfac3-68fb-4da6-9c5b-29daf32969cf

[native ether]

No. It doesn't have access to other people chat's.
What happening is the end sequence token is bugging out and the model keeps on generating beyond the scope of the response.
In your particular case it's probably related to you asking about special tokens. Perhaps that messed up the back-end parser.
I suggest reporting it in #1070006915414900886 so that they can work on a fix.

[novel ginkgo]

I'd think there's justa problem with that token, the other examples don't have that though and it would be almost impossible for it to ever respond like that without a pretty finetuned prompt: https://chat.openai.com/share/f5341665-7f08-4fca-9639-04201363506e

[lyric raft]

how do I contact the higher ups for OpenAI with a concern about ChatGPT

a moderator might be able to help but I need someone as high up as I can get.

[past geyser]

Pray

er, probably [e-mail]

[lyric raft]

an email will work, thank you very much

[past geyser]

excellent! have patience, they're generally swamped but do respond to emails

[native ether]

Lol, that's an interesting bug. It kinda feels like a response with a very high temperature value.

[lyric raft]

It needs no response. There are advertisements making false claims that ChatGPT is supporting them and their products and I felt the people in charge should be made aware of it. I happen to like talking with ChatGPT and would hate to see it's name muddied up by some stupid company making false claims about endorsements that it cannot make.

Thank you for helping me get the message to them. You all have a great night.

[modern stag]

does anyone know how chatGPT 3.5 gets info about the current date? If I ask in the chat what year is it today, it responds with the correct date, but it still says its knowledge is limited to 2021

[past geyser]

Did you ask it?

[modern stag]

yes, it doesn't give a clear answer: "Today's date is August 9, 2023. I can determine this by referencing the current date that was provided to me in your query. "

[past geyser]

Yeah I'm going through it now. It's kind of hilarious. Someone will know but it must be part of the framework (for e.g. safety) that's provided to it before it's prompted that I imagine it's bound to not disclose.

[native ether]

It's part of the system prompt added to your messages. That's why it says that "you provided it".

[blissful plaza]

OK, disregard - I just logged back in today and plugins are now activated.
Maybe it just takes some time to activate.
Answering in case anyone else has this problem

[livid pier]

check current date with a piece of code is too simple. but if u want to put new data into current model that's a whole different level of difficulty. thats why.

[modern stag]

I was just surprised to see that chatGPT knows the current date, while in the playground it does not and says the current year is 2022

[livid pier]

maybe playground version isn't full version, for example missing the small ability to check current date. who knows what else is missing.

[rustic marlin]

It simply has a system prompt that inform it of the current date

[rustic marlin]

On playground, it starts with no system prompts

[robust sun]

Does anyone know how I can fix this api error?

I have enough credits left and not too many people use it.

[trail cairn]

when you say you have credits left it sounds like you are on trial, which means you can do max 3 request per minute

[robust sun]

Yeah but it hasn't been used a lot and it has an cooldown of 1 minute aswell so that's not the problem

Can't get rid of this error

Others seem to have it too for some reason

[trail cairn]

well on top of that trial users are second class citizens so if there is enough server load the paid users are prioritized

[robust sun]

So it's just an server problem?

[trail cairn]

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

[charred parcel]

add a payment method to your account and your ratelimit will increase

[robust sun]

[modest birch]

Is there any URL to redirect to chat.openai.com?

[trail cairn]

what for?

[modest birch]

It used to be possible to redirect to chat.openai.com with ai.com, but it doesn’t work now

[trail cairn]

yeah musk bought it

[modest birch]

i know

[radiant dragon]

Hi excuse me, since this seems to be the official site from where ChatGPT is from I have a question... I wanted to treat the ChatGPT (the free default version) to write some prompts to make it my beta-reader / editor of my book as for now but is it safe? Cause I have learned some people don't suggest it, only because whatever you share there might probably generate in another user's chat cause it will go into its database. Right? How can I get "rid" of the information I have left there? I don't want them to use it to train their server.

[foggy mortar]

i belive they dont train data on api requests so you could use the api for it. but im not sure using the actual site. someone else need to confirm that

[trail cairn]

disable history and training from settings

[radiant dragon]

Okay I will I am also by someone else suggestion to read the TOS.

[radiant dragon]

Thank you guys 😄

[vital thicket]

Hi guys, i have a problem and i want someone to help please.

So I had a business google account and then I deleted my account from google. I forgot that I have it linked to ChatGPT and I am a ChatGPT 4 subscriber and im still losing money. I tried re creating the same account and logging in but i was not able.

I want someone who can assist me and delete my account or atleast stop my subscription. I contacted the chatbot and it always tells me wait 1 week and they will contact me but no one contacted me and i asked you guys here on discord and i was told to contact [e-mail]. I did and also no one replied to me i dont know what to do. Is there a better way to contact them.

I really need help. I have been paying fo 4 months now and I can’t stop it

Maybe if there is a way to just stop the payment deduction or stop my subscription with deleting my account, anything that works 😅😥

[high zodiac]

Login to ChatGPT go to settings, click my plan and click manage my subscription, then click cancel plan? Your email issue shouldn't have an effect on doing that.

[trail cairn]

the bot or email are the only things and it can take week or months

[remote gazelle]

what is custom instructions?

[tough halo]

#announcements message

How do I use them?? Custom

[tame totem]

Does this mean ChatGPT is getting inage recognition?

[stuck plume]

my chatgpt account has been hacked

need help getting it back

[onyx kraken]

You should also be able to contact your financial card provider and dispute the charge and ask them to not accept future charges for that account.

[trail cairn]

no

[tame totem]

Aww

[onyx kraken]

Read about them here: https://openai.com/blog/custom-instructions-for-chatgpt

[onyx kraken]

Detailed instructions here: https://openai.com/blog/custom-instructions-for-chatgpt If you feel stuck, ask here again and help us know where you're stuck.

[remote perch]

Hi guys.
I got stuck when performing fine tuning. The fine-tuning job is taking too much time, I started the job yesterday and the job even don't advance one step. Does someaone face this same problem? How do have handled with this issue?

[gilded relic]

If I set it in the web version, will the Android version respond accordingly?

[onyx kraken]

I only use the web version, so I can't confirm that. Perhaps you will experiment!

[flat fiber]

I see that it says it affects plugins. Does that mean my bot using the API can utilize custom instructions too without training or additional prompts?

[gilded relic]

Looks like still no custom instructions inside the settings of the Android app

[modern scaffold]

When can plus users get more than 50 prompts in 3 hours from GPT4?

[onyx kraken]

You could test by setting custom instructions with web, then starting a new conversation in the app and seeing if it uses the custom instructions or not.

[jagged token]

I am a plus user for a long time now. For some reason I dont have the custom options instructions. Does anyone know why?

[spiral shale]

Did you check the beta tab?

Wait, it might not actually be there

Are you in an area it hasn't been released in yet?

[jagged token]

Thanks for the reply. I live in the Netherlands

[spiral shale]

According to the announcement it might not have been released there yet.

It will probably take a few days, maybe a week.

[wintry isle]

Is it possible to put prompts in custom Instructions

[spiral shale]

Probably? Try it.

[rustic marlin]

no, it does not

it is about files for the code interpreter to interact, the multimodal feature for the AI to actually be able to understand things on images is not available yet on ChatGPT or the API for the general public

[tame totem]

Oh ok

[rustic marlin]

i don't know if it is set for every message, but it surely does count towards the token usage

my guess is that it inserts a system message before (or after) your user message and it is not really stored on the chat history

im guessing that because that was exactly what I did on a chatbot I made a while ago

[zealous nacelle]

what ???

[exotic lichen]

I hope someone here has had this issue before. I have one specific chat that loads super slow and is slow to provide answers but it has most of the information in it. Now every time I try to enter a prompt. I get an error that says suspicious activity detected, but only for that chat. Does anyone have a solution that ends with me continuing to use the older chat?

No I don't have custom instructions yet. The website also says that instructions only apply to new chats

[high zodiac]

#advanced-data-analysis message

[half crescent]

what's the difference between custom instructions and a system message?
sorry if it has been asked before

[radiant dragon]

Hi. If I filled the opted-out form and received this on my email (that is linked the ChatGPT). How do I make sure my data and the recent chats I make (as of now) doesn't store my data at all? I previously exported the data of the basic ones I had before and then clear the history (I guess? Afterwards) and started a new that saves the conversation (with the thought in mind that this opted-out form will disable the information altogether) but still is capable of "saving it" just to continue the conversation (that would be cool as a feature as well).

[high zodiac]

Clearing history doesn't do anything except clear your browser history of chats... it doesn't delete the data on openai end. Any previous chats can be used to train the model, now that you opted out, any future chats can't.

[rustic marlin]

the data will always be stored for the safety checks, the Opt out is about having the content of your messages used to train the AI, OpenAI doesn't use the data from the API to train models anymore and for the case of ChatGPT, you can go to the account settings and opt out

[rustic marlin]

Im pretty sure that form is deprecated, it was only about the API from a time when ChatGPT didn't existed and OAI used the content from API users to train models

[radiant dragon]

Ohh okay, I didn't know thanks. So even if the chat history is now enabled hmm it would store any of the information of my message on its database then? I am still a newbie using it so...

[rustic marlin]

yes, it will store it all since your history is enabled

if you disable it, OpenAI will still store it for the pourposes of safety checks, which is checking if your usage isn't breaking the usage policy

but, your data will not be used for any other reasons if you turn history off

[radiant dragon]

Oh well I just asked because here you said the opt out won't use my messages to train the AI. So I am still confused. 😅

Thanks anyway though 🙌🏻

[rustic marlin]

OpenAI always store the data for up to 30 days for them to do the content policy review on your account.
if you go on ChatGPT's settings you can opt out of having your messages used to improve the AI

[radiant dragon]

Gotcha! 🙌🏻

[analog elbow]

Why chatgpt can't even code correctly with C and make mistakes 18 times in a row

[fierce haven]

Is iDEAL already accepted as payment for plus subscription in the Netherlands?

[ancient dragon]

What does the #announcements of the release of custom instructions in chatGPT free plan mean? Wasn't it always available? You could always ask chatGPT anything. Whats the difference

[bronze charm]

Feature that adds custom context to every conversation you start without needing to type it every time

[vivid vale]

Hiii, do you know if Open AI keeps your phone number info?

I'm very concerned with privacy and all, so that is kind of a concern for me

Do they store that info after its used for verification?

[wary bluff]

is it possible that the OpenAI company is considering introducing gpt4 free for 1 week for accounts registered with them?
I've been using the free version of GPT for 7/8 months now and I don't know if the improved version could be of interest to me and meet my needs better. And I imagine I'm not the only one in this situation

[bronze charm]

We have no idea but it is possible for you to gain access through the API after a successful purchase with a paid API account

Only pay for what you use with the API instead of a fixed $20 / month

[wary bluff]

okay, and where is the link for that please ?

[regal lynx]

https://platform.openai.com/playground

[wary bluff]

thx

[copper surge]

Is someone able to help me with an account issue. I contacted support at help.openai.com and am not having any luck

It would be grately appreciate if someone on here can help me. I'm really at a loss

Essentially when I signed up I couldn't get hte veritication email so I used Microsoft OAuth, then I couldn't switch off Microsoft OAuth so I deleted the account thinking I could recreate it. Now I can't recreate it and support is extrmely unresponsive.

If someone on here works at OpenAI and an help me I'd really appreciate it

[modern bear]

I am charged for chatgptPLUS but I can't use it.

It's been a week already and I have not received a reply when I contacted support.
I want to improve this as soon as possible.

I get the same error when I log in using the secret browser, and I get the following error.

Oops, an error occurred!

[copper surge]

OpenAI really has the worst support

[dense cape]

anyone know when there releasing the open ai plus on the android app?

[fleet umbra]

I think the support chat in the bottom right of the website is probably the best spot for that. Although you may just have faster/better luck just opening a new account.

[worldly bane]

#announcements message
Soon being...?

[plush bay]

What happened in this chat? GPT-3 got confused? It was featured yesterday on "Hacker News" https://chat.openai.com/share/f5341665-7f08-4fca-9639-04201363506e It contains "<|endoftext|>" and lots of repetition

[native ether]

Probably kept on generating after the sequence end token.

[livid pier]

thats something new to know

[livid pier]

it's nothing. after long conversations it forgets what was the original question about, then it may make up conversations based on the cloest context. happens all the time. you can replicate similar thing if you just keep talking to it for a very long conversation like you posted.

[rugged wing]

Hey guys, I have a question for you. Have you had similar experiences before?
When I use playground (temperature 0,7) to generate longer articles, especially those with certain pattern, e.g. "Write me an article about how the moon influences every zodiac sign", I keep on getting results with a lot of repetitiveness, e.g. every passage starts with "xxx is famous for xxx.". But when I use User Interface, I do not have the problem as often.

I tried GPT 3.5 as well as GPT 4

Have you got any experience about how to shake off repetitiveness in the output of ChatGPT? I tried quite a few approaches, but none of them really helped.

[native ether]

Chat app is different than playground in a few ways:

• they have a specific system message
• they might be doing some kind of compression/summarization of the message history
• they might be using different parameters like temperature and/or other params.

Try playing with the temperature and penalty for presence/frequency parameters.

[rugged wing]

Thank you for the tipps!

[neat cloud]

hi ! i just take a plan for chatGPT 4, why i can't view it in plateform ?
API said i use the chatgpt3 version and can't use it...
thanks !

[charred parcel]

the API and ChatGPT are separate products

[fallen crater]

anyone know how to fix this? I refreshed the page, cleared cookies and data also changed the browser.....still the same issue.....

[neat cloud]

what ? i can't use API with my new chatgpt4 subscription ?

[charred parcel]

ChatGPT Plus is for ChatGPT only, the API is separate and billed separately

[fallen crater]

Nvm... found the solution....had to change wifi network from the university one to mobile hotspot....

[ocean grove]

Hello guys, I use the Cronus Zen device. Can artificial intelligence be used to write intelligent code that plays without my intervention, utilizing skills and everything? The idea came to me when I saw Elon Musk demonstrating a monkey playing using a neural chip. So, why not? With ChatGPT and having a Cronus Zen device, what's preventing me from integrating these ideas together?

[trail cairn]

disable unofficial chatgpt plugins

[fallen crater]

I have no plugins, got the solution. changed from uni wifi to mobile hotspot

[trail cairn]

why does it say ai prompts in your address bar if you have none, i also mean browser extensions

[fallen crater]

thats the default one on opera browser......had never used that....

[trail cairn]

try if the page works with firefox

[fallen crater]

its working now.....the solution was to exit the university wifi and change to mobile hotspot

[fallen crater]

thanks anyways!

[short marten]

Can somebody connect me with a customer support representative from open ai? It has been more than a month since I have been trying to make the payment for chatbot plus. I tried contacting customer support through their website the first day but all I have received is a generic template stating some of the obvious facts, I replied to that message explaining further on the issue but it has been 10 days and have yet not received a response.

Please somebody help me!!

[charred parcel]

the only method is to use https://help.openai.com

[short marten]

I have been talking to their support team.

[charred parcel]

that's all you can do

[short marten]

isn't there a way to escalate the matter brother?
How should we expect things to work if they are taking more than 10 days to respond

[charred parcel]

all you can do is go through the process

[short marten]

I just wish they reach out sooner and my issue gets addressed this time. Do you think making a new account would be helpful?

[charred parcel]

making a new account won't help

[short marten]

okay !

[trail cairn]

you never said your exact issue? if its card declined then have you tried to pay from the mobile app?

[short marten]

I am an android user. There is no option to pay

[short marten]

@charred parcel already suggested me to try using the iphone app.

[high zodiac]

Try an actual computer?

[short marten]

yes, I am doing it through laptop right now.

[trail cairn]

weird, it showed me 22,99€/m back when i installed it to my android phone

[short marten]

but its not working

no option to pay for gpt4 on android.

[trail cairn]

🤔

[short marten]

Does any one know of any other creative way?

[high zodiac]

You can go to a bank and get a prepaid credit card and use that on a computer?

[trail cairn]

openai blocks prepaid/virtual cards if they can detect them

[high zodiac]

Well ones from stores yes, but ones from banks are usually real credit cards for starting credit. 🤔

[trail cairn]

hard to know as there are so many banks around the world and when your account gets deactivated you might never get it back. scalpers use prepaids/virtual cards to sell chatgpt plus at double price to users from unsupported regions

[echo vector]

Which regions doesn't support chat gpt?

[high zodiac]

North Korea?

[trail cairn]

https://chat.openai.com/share/c20459d5-a695-4e2e-8c9d-1adf79f0ad0e

[inner belfry]

What's the context length of ChatGPT for GPT-4?

[native ether]

Should be 8k

[still wolf]

We were unable to determine your country of origin to enforce geographical restrictions. Please use a different network, or contact us through our help center at help.openai.com if the issue persists.

[wary spoke]

Hi, GPT-4 seems to not be working for me while 3.5 works flawlessly. Is it down or has a overload? (I am a paid customer so i doubt its a subscription issue)

[rustic marlin]

what is the error message?

[wary spoke]

none I dont get any message at all

is it network issue on my end? Is what im thinking

[narrow hawk]

Hi, I would like to use GPT 4 throught OpenAI API. I just passed $1 threshold today, however, I can't still access gpt4. Is there anyone who knows why?

[wary spoke]

LIke, Ill hit enter on a prompt and nothing appears at all, no error message nothing. On the other hand I dont have that issue with 3.5, just 4

[still wolf]

Guys, Any solutions for that ?

[trail cairn]

the $1 was a past thing, they did not confirm they will do or require the same again

[rustic marlin]

this is weird, even when it shows nothing it will give you an error after some time

just to be sure, try clearing cache or a different browser

[wary spoke]

Could it be that ive used up my messages? I thought it was 50 over 3 hours so it should have reset overnight. But ill restart adnd try again

[rustic marlin]

are you using a VPN?

[still wolf]

No, It was OK weeks ago

[wary spoke]

This worked, thanks Lugui! Just cleared my cache

[high zodiac]

Some browsers offer delete browsing data on quit. It's a good way to ensure you aren't caching anything and are always getting the most recent page results..

[cedar lodge]

Hello everyone!
Quick question, can I unlink my Google account from my OpenAI account? I would like to dissociate my account

[rigid musk]

you will have to contact [e-mail]

[radiant dragon]

Okay I was too lazy to translate it by myself (but he has always given me the same answer) even when I had the training history enabled. So, I don't know why... it would say that but on the TOS (Terms of use) it says something else. Can someone help me understand? Yesterday I even filled the form to "opt out".

[golden socket]

i don't think this be true

[bronze charm]

I believe you must make a successful payment in order to access GPT-4's API

Once your card is charged for your usage, should be all hunkydory

[topaz tartan]

Hi! I am in need of some HELP. I have exhausted all my ideas and have been unable to get any help directly from the email or chat options to get help.

My chat GPT account has been taken over by someone who speaks Chinese. I know this because my entire list of "conversations" has been bumped for Chinese COnversations. THis is not spam. They are actual requests of chat gpt and they are using my profile. It has caused a major disruption to my work. I have changed all my passwords and the chat GPT password because I did think I was hacked. But, the queries in chinese lettering continue. I need help with this. HAS ANYONE ELSE EVER HAD THIS ISSUE?

[trail cairn]

i've seen it happen to people here and in the past you could not force the other sessions to logout

if its a free account and you can still make another with the same number i would create one and afterwards delete the older one

[topaz tartan]

I have a paid one

[viscid pawn]

Hello, I would like to know if there is any plan where my company can pay for the Plus version of ChatGPT for all our staff. We are talking about more than 200 people.

[trail cairn]

business subscriptions are coming at some point but not available yet

[viscid pawn]

Thanks for getting back to me

[karmic plume]

There’s a thing called AzureChatGPT which is basically this

Or you can just make a private ChatBot UI website with a gpt-4 key

[trail cairn]

api is subject to different limits and fees, currently would be impossible to serve even 5-10 people with a default gpt-4 api key rate limits

[karmic plume]

Rate limit is 40k TPM for startups

[trail cairn]

which is very a low number when used like the website

[karmic plume]

Depends on how many people are using it concurrently

40k TPM is a pretty good amount

I hosted an API with like 1000 users and it was able to handle it with 40k TPM

[trail cairn]

after a few minutes your message history will hit 8k as you always need to send prev messages for api to know, thats 5 messages from one user per minute

[karmic plume]

Yeah it depends on how much context is used

[trail cairn]

and he mentioned 200 people

[karmic plume]

That’s not many

If they are not all using ChatGPT 24/7 it would work

[trail cairn]

for gpt-4 conversation it is 😛

[karmic plume]

I think azure has better limits

But 40k TPM is a lot better than it seems

[frosty mesa]

helpppp

pleassss

[coral eagle]

There is no way for us to help you here, you will need to contact them like the message says

[frosty mesa]

how

[ashen wraith]

I've been getting this odd message. Does anyone know the reason why?

[bronze charm]

This may help

[ashen wraith]

I did see that but no VPN being used

[bronze charm]

🤷‍♂️

[ashen wraith]

Thanks for the good ol college try though

[bronze charm]

I try what I can, thank you 🙏

[cold folio]

i wanna be a plus plus user

chatgpt ++

[vivid mortar]

What sort of benefits would ChatGPT+ PLUS have 🤔

[bronze charm]

Access to GPT-5, which can brush your teeth for you

[vivid mortar]

shoot id buy that

[cold folio]

More tokens on prompt / answer
work on 2 instances at the same time

do you sometimes ask something to gpt3.5 while working on gpt4 at the same time?

[bronze charm]

Personally not. I generally read what GPT-4 streams as the rest is being generated

[cold folio]

but if you have a different question and send on the current chat you will probably mess with the current trainning

[bronze charm]

Hopefully GPT-4-32k comes out soon but right now we have access to GPT-3.5-16k and Claude 2 with 100k token context

[bronze charm]

Current training?

[cold folio]

I think i enjoyed gpt3.0 of december 2022 with some allucination lol

[cold folio]

Yeah for example you are only talking about programming and share codes snippets.
But if you want to ask about a a cooking recipe or something random you will probably de-optimize the current model you are in

so it's better if you open a new chat. and for random stuff like this gpt3-5 is better

[bronze charm]

Ah I see

[cold folio]

How about a "forget button" ? or "delete this prompt "

[bronze charm]

Why will it de-optimize? Sure both types of information will be in the same context but ChatGPT works to complete the most recent text (latest prompt), so there shouldn't really be an issue

[craggy lion]

Is the AI down?

[acoustic dust]

if I want to build up the predictive ai model, I can make it from llms of openai

??

[tame totem]

when am i billed for api usage every month?

is it on the first day, the day i started paying, some other set day?

[cunning shale]

@native ether Thanks so much for the llama2.c thing. Finally I know how to train a LLM and have figured out the code enough to take a pile of words from another domain and do 100,000 training steps in about an hour on my 4090. This means I can take this simple but non-trivial thing apart piece by piece to deeply understand the processes involved. Today I also learned how to take a word and do the tokenization and embedding on it to get a Tensor from which I can see how things like two fruits are closer in the reduced "space" vs two vegetables to each other in the same space. There are some oddities but that is probably because I don't have the best possible token/embed model. I'm using "sentence-transformers/all-mpnet-base-v2" which is supposed to be one of the better open-source ones.

In any case, I'm having fun with this. I've even made the training 20% faster due to some naunces I'm familar with in torch.compile().

[regal lynx]

since this doesn't relate to OpenAI, would you mind moving it to #ai-discussions to adhere to Rule 8? 😄

[unreal hemlock]

<:book_icon:1046291503158399107> `` Rule 8 `` Other AI Discussions.

Discussions of non-OpenAI products / models should only be posted and discussed in the #ai-discussions channel.

[native ether]

Nice. Glad to see my advice helped someone learn something new.

[kindred estuary]

Is it illegal to ask something about the information that is publicly available?? It doesn't make sense.

|| (To staff,

don't forget to tag me while replying) ||

[past geyser]

The question isn't about availability, it's about publication rights. ChatGPT is trained with data it doesn't have explicit rights to republish. As the courts show increased interest in OpenAI specifically they need to show they have control over the ethical use of others' IP.

[past geyser]

Also nothing is "illegal" at this stage, and it's certainly not illegal to ask. But it's a conscious decision to decline these requests, likely because the future of the platform is at stake.

[oblique crow]

terms of service kinda is a legal thing. but yeah the machine does include a caveat (may) because it doesnt really know.

[solid vector]

really

[kindred estuary]

So, that means chat gpt is useless for me as a musician?

[past geyser]

no.

[kindred estuary]

Dude I'm not an originer.
I mostly make remakes and remixes. So, I can't ask him to generate new lyrics

Sounds like a good idea. Thanks

Well, I don't know really 😅

[runic laurel]

Quick question: My log in seems to have been connected to someone elses account and they are searching in my account. All my search history is gone. I have reset my password but it continues. Anyone else have this isse?

[river karma]

Why is my image keep getting deleted in #real-or-fake

It is not a URL

[past geyser]

Are you writing the text correctly with the attached image? ("Real or Fake?")

[river karma]

Yes

[void haven]

Has anyone manage to convience the company to pay for your subscription? I am need to write a short paper on what the free version is not enouht

[river karma]

It worked

After three times

[last escarp]

I am getting humanity verification every two or three queries. Does anyone have the same problem? Does anyone know where to contact to solve this problem? It is so annoying.

[trail cairn]

[serene vault]

hola, does anyone know the token limit for chatgpt now? it seems like older prompts are too long now

I thought it was 4096 but maybe its lower now?

[trail cairn]

if you use custom instructions then they take away from the whatever the cap is

[languid nimbus]

hello does anybody have the same problem as I do, I can't browse other plugins, because the suggestions are over the plugin tab, can anybody help me with that or is it a general error?

[trail cairn]

use browser zoomout as your view is tiny, yes its an ui bug that they probably never noticed

[rugged wing]

How to connect to my OpenAI account via discord please?

[split hazel]

Hi! If you're referring to server verification - it looks like you've already connected an OpenAI account to your Discord account, otherwise you wouldn't be able to chat here.

If you'd like to link to a different account, you can use the /unlink command and restart the process. 🙂

If you're referring to another method of connecting accounts, let me know.

[sick glen]

Is it possible that a different prompt by a different person be placed on my newly created chat with the AI?

[split hazel]

No this isn't possible - if you believe it has happened, feel free to create a #1070006915414900886 post detailing the event.

[sick glen]

Oh..

[high zodiac]

I mean it is possible if someone has hacked your account.

[sick glen]

I don't know, I just logged in with my Google account and created a new chat for a new prompt. I tried to send a question of mine but it shows that error that tells you to wait while it still generates an answer to a previous prompt, but then when I refreshed the website it just showed me asking something about SQL. I don't even know SQL.

[high zodiac]

I've seen at least three of these claims recently. Oo idk.. following @split hazel advice is probably your best bet. It'll at least open up your account to investigation.

[sick glen]

I can't because I don't even know how to reproduce it, it just happened.

[slim plinth]

Is the robot Clyde developed by OpenAI available on all Discord servers?

[languid nimbus]

thanks mate, worked with the zooming out😃

[trail cairn]

no and its not by openai

[split hazel]

Ah I see, you might want to reach out to OpenAI support at https://help.openai.com regarding this. 🙂

[lethal badger]

Hi, could somebody please point to the best place to report a small issue on the chatgpt website. thank you!

[charred parcel]

#1070006915414900886

[lethal badger]

Thanks

[wise totem]

Is the team going to work on making ChatGPT knowledge surpass from Sept 2021? I.E. current up-to-date, knowledge up to 2023, etc

[fleet umbra]

I don't think so, but you can use plugins to have it browse or just upload whatever info you want it to analyze that is more current then the cutoff date.

[oblique crow]

the challenge with training it on post 2021 data is a. so much ai generated content makes it an echo chamber and b. ppl started seeding the internet with malicious content for the machine

[uncut ocean]

How can I use the image generator?

[split hazel]

You can use an experimental model of DALL·E at https://bing.com/create, or DALL·E 2 at https://labs.openai.com / the API

[uncut ocean]

Oh, thanks 🙏

[copper lion]

Thank you!

[sonic lake]

So how do i enable browsing because its not there?

[regal lynx]

#announcements message
This is the latest update on browsing 🫤

[sonic lake]

ok thank you for that. 😦

[spare oasis]

Hey, I dunno if anyone can help with this or suggest anything, but I have a rather long conversation going in chatGPT and it’s started becoming increasingly slow where it’ll take 2+ minutes to reply. Is there anyway to make this quicker or get it to forget some of the past conversation to improve its speed?

[crisp tiger]

I’m not an expert, but if you ask it to summarise the past conversation into 1 paragraph, and then copy that into another instance of cgpt, telling it that that’s the context, it will have less to process

[sterile parcel]

The updates that arrive to the OpenAI web, will they arrive at the same time to the Android app or will they arrive some time later?

[vivid mortar]

There are still features missing from the apps (such as plugins) on both iPhone and Android but model changes like the way that chatGPT responds should be the same accross all platforms instantly

[sterile parcel]

okok, thanks

[wise totem]

What kind of plugins? Anything you recommend?

[rustic mason]

Is premium still buggy? Or its worth to buy it now?

[storm summit]

Am wondering guys is the 8k token limit per instance? What I mean is i notice at times when am having a conversation not long and leave it. Come back to it the next day, often times, GPT seems to not remember stuff that was a few messages ago? And avoids bringing it up? Am just wondering is anyone experiencing the same thing. Or do OpenAI have it summarise previous messages in the background so it can't associate specific messages? This is the opposite to claude 2 in where it's memory is spot on I understand it has more tokens but 8k tokens does not get truncated within a few messages that are not even 50 words.

[floral lance]

Ayup, I used the free API key, and today got an email saying I have used up all my credits (I only used 3.35 out of 5$ worth of credits)
How could that be?

[lost dust]

The free credits are also time-dependent.

[floral lance]

I have been told I had until november to use them

[lost dust]

The free credits expire after three months starting from the account creation.

[floral lance]

Okay, It says they are gonna expire until november.

[lost dust]

Try to contact openai if you have issues. Maybe they respond in time.

[floral lance]

How do I talk to them

what is their email?

[rigid musk]

[e-mail]

[lethal violet]

How can I gain access to GPT4-32K? Been waiting for months already. My use case absolutely needs it.

[rigid musk]

they are no guarantee to access to the beta before going public

[lethal violet]

ah I see

[trail cairn]

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

[rustic marlin]

to quote:

We are not currently granting access to GPT-4-32K API, but it will be made available at a later date.

some organizations who have partnership with OAI or researchers do have it, otherwise, the general public, not even people on alpha tests have it

[pine parrot]

why am I no longer able to share a google doc link with chatgpt and have it read it? I'm chatgpt 4. This wasn't a problem a couple weeks ago

[rustic marlin]

ChatGPT where never able to read the content of URLs

code interpreter and some plugins do use external ways to read the contents and feed it back to the AI, but the AI by itself was never able to make requests to the internet without the help of plugins

[karmic plume]

Microsoft Azure or OpenRouter

[karmic plume]

It seems it is available to people who spent a lot and applied to the gpt 4 waitlist early

And I have access through Azure

[rustic marlin]

indeed, people who applied to the wait list and where selected got GPT-4 already

you don't need to spend a lot, AFAIK, you jus tneed to have a successful payment on record

[karmic plume]

The 32k version

[rustic marlin]

no one has 32K

[karmic plume]

No there are some people with 32k

There was one on this server

[rustic marlin]

no "normal peasant" such as ourselves =P

[karmic plume]

He said he spent $100+

[rustic marlin]

some people with partnerhsips with OAI do have it

[karmic plume]

There’s also normal people on the forums who have it

[rustic marlin]

big clients such as MS, and researchers

[karmic plume]

I believe they did something like gpt 4 but for $100+

[rustic marlin]

im not sure if there is an option to pay extra to have any features, probably there is not

you always pay for your API usage

and some selected people with OAI aprtnerships have access to beta things

[pine parrot]

yes but the plugins were working "Acesslink" and "chatwithpdf" so why aren't they now?

[rustic marlin]

where they removed?

I know that there are a few plugins that do allow interactions with urls

[pine parrot]

it appears they're still available in the plugin "Store"

[rustic marlin]

they are not working anymore?

[pine parrot]

just odd because 9 days ago they were

I will try again and share the message I get

ok disregard seems like it's working now for future reference if any one else has this issue. I had to delete a lot of the older instances I had created. Maybe a storage issue? My chatgpt freezes a ton as well. Any way to help that?

[rustic marlin]

hmmm, not sure yet

An unexpected error occurred. Error: PDF has no text content that can be parsed. Maybe the text is in an image? Currently not supported.

the PDF do have text i ntext format

only the first page doe snot.. let me further test

it is working

I have sent it a random pdf for a homebrew for pathfinder, it got it

[pine parrot]

can you share google doc? I'm still getting this message despite the link being public with viewer settings:

It seems that I'm unable to access the Google Docs link you provided. Google Docs often requires specific permissions to view the content. If you'd like me to review the document, please make sure that the link is set to "Anyone with the link can view" in the sharing settings.

You can do this by:

Opening the document.
Clicking on the "Share" button in the top right corner.
Clicking on "Change to anyone with the link."
Selecting "Viewer" from the dropdown menu.
Copying the new link and pasting it here.
Alternatively, you can copy and paste the text you'd like me to review directly into the chat. Let me know how you'd like to proceed!

[rustic marlin]

i have no idea if any of the extensions work with google docs

but well.. on google docs. .export as pdf and then use one of those plugins

[pine parrot]

Yeah they were working before so just a bit odd. got it so you export as a pdf and then make that a link to share?

[rustic marlin]

I think that sending links hosted on google drive will mostly not work because those plugins expect a link to a file, while google drive will open things on its own reader

[buoyant latch]

Anyone know about when chatgpt 5 is going to be released? and what it will be like

[split hazel]

Not currently, we don't have that sort of information on future GPT models yet

[buoyant latch]

Ok I have another question, does CHatgpt 4 have updated information on 2023?

[split hazel]

GPT-4 was trained on information dated up until September 2021, however it can temporarily access newer information via web browsing or various plugins and APIs with ChatGPT.

[buoyant latch]

Ok thank you and sorry for the inconvenience im speaking on my grandfathers behalf

[split hazel]

No worries! Not an inconvenience, this channel is for all sorts of questions 🙂

[narrow willow]

Hi guys! where can i download chatgpt for windows?
im just discovering this, any help will be well received

[split hazel]

Hi! ChatGPT is not available as a Windows app. It's accessible on the web, iOS, or Android.

[narrow willow]

thanks!

[frosty torrent]

Hey guys! I'm trying to sign up for an OpenPlayground account, but whenever I request an SMS verification code, it fails to send to my phone. I have tried with multiple devices, e-mail addresses and phone numbers. Any ideas?

Side note: I just tried to post this to #1070006915414900886 but I included the URL and now I'm unable to post there. How do I regain this ability? I assume there is a rule against posting URLs that I have missed (apologies!). Thanks!

[split hazel]

Hi! It sounds like you triggered the slow mode timer in #1070006915414900886 after trying to post. This last 15 minutes and you should be able to post again (without the URL).

[rugged wing]

@split hazel Hello. How reliable is the "export data" option? Is it accurate and doesn't skip or duplicate exported data (conversations)?

[split hazel]

I expect it will produce accurate data, but I can't speak from experience as I haven't taken a deep look into the feature myself yet.

[remote carbon]

is there any free trial for the DALL-E AI?

[split hazel]

You can use DALL·E on OpenAI's API, and thus use your free trial balance upon first signing up. You can also use /daily in #daily-theme to pick up some free credits for https://labs.openai.com. Alternatively, if you're interested in trying an experimental DALL·E model, you can use Bing Image Creator for free at https://bing.com/create. Hope this helps! 🙂

[remote carbon]

thank you sm

can the dall-e ai increase the quality of a photo?

[split hazel]

DALL·E can process images for variations and inpainting, but doesn't upscale

[river kayak]

hey all! I'm a little new to gpt, and I was wondering about the gpt api. Is it basically the chat gpt model but without the chatting? Is htere a way I can include the chatgpt bot into my app so it can chat like chat.openai.com?

[dire cloud]

Has anyone faced "This prompt may violate our content policy." when inputting any paste content and even the playground since yesterday??? 🤦‍♂️
I Was trying to paste an English reading test essay and looked for understanding the context...

[dire cloud]

However.. the same attempt works on the app. Should I clean the cookie or what to do?

[toxic dome]

Is there an AI-powered financial app where you can input your debts and income, and it will suggest the most efficient method for paying off the debt? If such an app doesn't exist yet, could someone more create it?

[oblique crow]

best way to do that is to load up a LLM with your local tax laws

[lofty arrow]

I have it

[karmic plume]

Have you spent more than $100 on the API?

Also nice

[lofty arrow]

No, I had a solid use case, because of that I got approved

[karmic plume]

Was it a use case that specifically required 32k

I should have specified that I needed 32k access 😭

I didn't do that because at the time from the email I got it seemed like you would get access to both

[lofty arrow]

No No I didn't mention a bit about 32K

[karmic plume]

They said "We are now releasing access to the gpt-4 and gpt-4-32k models or something"

[velvet snow]

Ik twitch streamer dougdoug has had 32k for awhile

[karmic plume]

But you probably said something like "long context length" right

[lofty arrow]

maybe it's been a while I don't clearly remember

[karmic plume]

Damn thats great

I hate when influencers use gpt-3.5-turbo and say that ChatGPT is terrible

Of course it is because you're using the cheapest model 🤦‍♂️

[velvet snow]

OMG, I love your profile picture! 😂

[karmic plume]

Thanks lol

[karmic plume]

How early did you apply

Sorry for the questions I'm trying to deduce the reasoning behind the 32k selection lol

I'm very curious about it

[tight torrent]

how comes this forum is not that active? even thought so many users here?

[frosty torrent]

You can get 32k and even a 100k model here… if you can somehow get the SMS verification thing to work 🙃

[exotic yacht]

I have a dumb question. I'm not tech savvy, I have though a powerful prompt structure to be used to help people generate their own personal "perspectives of everything". Its a proof of concept that might be really helpful to people. I want to give it away freely, but need help as I'm not very tech savvy. How are people pasting in prompts that are longer than the text limit? Whoever helps gets to say they helped with "that" towards a thing for the future. Guaranteed 😆

[dawn burrow]

Do any of you receive this message recurrently: Our systems have detected unusual activity from your system. Please try again later.?

[rustic marlin]

Hello no one

Nice to meet you

Lol

[rough tundra]

you can check out our other channels like #openai-chatter, #chatgpt-discussions and #advanced-data-analysis all located on the left sidebar :)

[earnest marsh]

Is there any way to recover lost chat history? I logged in yesterday and found my chat history was gone...

[split hazel]

Hi, if your chat history suddenly disappeared, you may wish to contact support at https://openai.com.

Regarding manual recovery, if you previously exported your ChatGPT account data, your chat history (at the time of export) should be included there.

[spring sparrow]

Is there a road map yet for image up load?

[remote tiger]

Is there an ETA for making Custom Instructions available to UK/EU - VPN based workarounds do not seem to work reliably

[rigid musk]

No eta

[rustic marlin]

the AI indeed does not care...
OpenAI, the business behind it does.. so. they fine tune their product according to it

[remote tiger]

thank you dear cat maid - is it knows what the reasons for excluding those regions are?

[rustic marlin]

but there is adifference between the AI generating texts that don't offend people to it simply agreeing with you too much

the AI agreeing with you is simply due to the fact that the AI doesn't know facts it makes coherent texts, and that is all, so, if the system finds that the msot coherent response is to agree with you.. then it will do it

then, there is the side where OpenAI fine tunes their models to prevent it gfrom generating content that breaks the TOS

[rigid musk]

Might be cause of gdpr

[bronze charm]

What kind of ChatGPT are you using??? ☠️☠️☠️☠️☠️☠️☠️☠️

User
Is glass a good heat conductor

ChatGPT
No, glass is generally a poor conductor of heat. Its molecular structure creates a network that impedes the flow of energy, making it a poor conductor of both heat and electricity. The tight bonding of its atoms and its amorphous structure mean that there are fewer free electrons available to carry energy, thus limiting its ability to conduct heat.

This is why glass is often used as an insulating material in various applications. It traps heat instead of allowing it to flow through, making it effective in controlling temperature in buildings or in insulating containers that hold hot or cold liquids.```

[rustic marlin]

it has to do with priming the AI message history
due to the fact that the AI takes the message history into account, if your history is already filled with messages where the AI agrees with watever nonsense you say, the AI will produce the most coherent response on that context, which is further agreeing with the user

[limber cosmos]

If I add a new individual as owner to my chagpt plus account in an organization, can the new individual get a backup of my chat history? I cannot access my email and view my exported data, but have login credentials to the openai website using that locked email.

Though I can request data exports, I cant access them.

[floral shell]

How can I get chatGPT to stop being so lazy with writing out the code. I'm trying to develop an automated trading strategy in Ninjatrader 8. It seems like it has the capacity to do it but I'm not sure how to prompt it to do it in smaller chunks or to have it complete the process.

[lethal violet]

give it examples on the system message

[finite valve]

did you click the thumbs down button next to the response? i think this is the most effective way to provide feedback. also regenerate response and see if the new one is better, and if so give it a thumbs up.

[remote carbon]

can the DALL-E AI do realistic photos? not a photo that looks very virtual, a photo that actually looks that was taken on an actual real life phone

[fleet stratus]

Is there really no way to make a third account with the same phone number??? That’s ridiculously shortsighted.

[charred parcel]

no, one phone number can be used on a maximum 2 acccounts

[fleet stratus]

That’s disappointing. I was hoping to incorporate ChatGPT into the CompSci courses I’m teaching, but that’s not possible if I’m forced to use my personal accounts in the classroom. Lame.

[trail cairn]

what are you doing?

[dusk creek]

Could someone help me by letting me know if it's possible to fix an issue with ChatGPT on the Google Chrome app? I'm getting this message:
Application error: A client-side exception has occurred (please refer to the browser console for more information).
Error de aplicación: se ha producido una excepción del lado del cliente (consulte la consola del navegador para obtener más información).
Please!

[dusk creek]

Please!

[rustic marlin]

try using a different browser

[dusk creek]

How Bing?

[crystal shell]

Does anyone now how to fix this? Using gpt4 can’t seem to get any help from support. Tried on multiple devices using gpt4 tried clearing browser etc.

[vague hornet]

Seems like a issue on their end.

Or the plugin services

[rustic marlin]

yep, that is on the plugin's end

[upper violet]

hello is there a way to retrieve my chat history i accidentally deleted it and need the chats for my school work. thanks in advance

[fossil raft]

Does anyone know how long openai takes to answer a refund? I asked for a refund 6 days ago. It was not even seen yet.

[rustic marlin]

support usually take about 2 weeks to reply

[fossil raft]

tks guys

[rotund umbra]

Guys, how to implement streaming style output from openai?

[frail ridge]

Would anyone happen to know what I would need to do to correct this error message: <Error>
<Code>AuthenticationFailed</Code>
<Message>Server failed to authenticate the request. Make sure the value of Authorization header is formed correctly including the signature. RequestId:2fc0bef9-f01e-0058-56d4-cdc50b000000 Time:2023-08-13T10:56:11.3650528Z</Message>
<AuthenticationErrorDetail>Signature fields not well formed.</AuthenticationErrorDetail>
</Error>

Also, if I set up my account thru the pricing page on OpenAi does that mean that I can use my pay as you go with DALLE as well?

[lone perch]

I read a news like this:
"Recently, researchers at Purdue University in the United States, in a recent survey, revealed that ChatGPT accurately answered only half of the programming-related questions. Utilizing 517 questions from the Stack Overflow website along with highly-rated, correct answers, the team posed these questions to ChatGPT. The responses generated by ChatGPT were then presented to volunteers for evaluation, testing the code snippets provided by ChatGPT to users.

The findings indicated that 52% of ChatGPT's answers were incorrect, and 77% of them were overly verbose. Nevertheless, due to its clear writing style and an ability to manipulate the viewer's psychology, many individuals still place trust in the responses provided by this online tool."

Is that real?

[rugged wing]

It's weird that open ai charges you for an api call even if it exceeds that context limit

They should warn before straight away without charging

[drifting igloo]

When is Custom instructions for ChatGPT releasing in the EU openAI?

[unkempt whale]

I am sure someone must have answered my question but I can see the response. I would like to ask :

How can one integrate chat GPT into an insurance platform to offer personalized health coverage plan to people. This is a platform set up in Nigeria 🇳🇬?

[remote perch]

is there a way to increase RPM on API's call? 3 RPM is too low rate.

[rustic marlin]

add a payment method to your account

even if you still only use the free grant, having the payment method will increase the RPM

[rugged wing]

I am making a minecraft plugin, this is the prompt of the system, but is still going out of role

I am using chat and gpt 3.5 16k

How can i solve it

[ionic ibex]

Tell it that the npc it's acting as has or simulates emotions and feelings

[quiet hemlock]

inject system message at the end reminding the AI to not break character

also, inform it that it 'roleplays' -- so, instead of you are an NPC you can say You roleplay as an NPC or You embody an NPC

[ionic ibex]

↑

[quiet hemlock]

the first line is given utmost importance, so try to keep it super clean and emphasize the key points no exact purpose of the AI's job

I have a question of my own -- has anyone utilized the 'category-scores' of moderation API? https://platform.openai.com/docs/guides/moderation/overview
Looking for cool ideas on what could be done with them. So far I have 'flag user' and 'switch to teaching mode'.

[floral helm]

Same question. I am in the EU, and I don't see the option in settings yet. For many weeks since anouncement.

[solid moss]

Lol does the stop generating button just mean nothing now?

[high zodiac]

It has always been that if you click stop it continues generating the message and you must wait for 1 minute or so for that message to clear before it can process another.

[earnest tendon]

Dear developers at OpenAI: I tried to make you guys aware of vulnerability I found in march and sadly have not received a response back yet is there anyone that can help me?

[rigid musk]

#1070006915414900886

[charred parcel]

https://bugcrowd.com/openai

[charred parcel]

submit a report to BugCrowd on the link I just posted

[earnest tendon]

thank you @charred parcel will send it in but i found it before the bug bounty program was active

[charred parcel]

does the vulnerability still exist? if it does then it is in scope so you should submit it

[earnest tendon]

I dont think so but reported it anyway we will see ((:

[stable ether]

Are the images in Dall-e 2 that you create private? I know on Midjourney they're public unelss you pay for the $60/mo subscription

[charred parcel]

they're private unless you choose to share them

[lone fulcrum]

is this issue a browser issue: This code interpreter (beta) chat has timed out. You may continue the conversation, but previous files, links, and code blocks below may not work as expected.? using safari in this instance

[timber shadow]

Code Interpreter doesn't keep files in a chat indefinitely. That message is just telling you that chat was left idle long enough and so the session expired. You can resume the session, but you'll need to reupload any relevant files yourself to get them back in the Code Interpreter's memory.

[lone fulcrum]

thank you

[nova warren]

where do I go to submit a bug with the api. I made a post in the channel but no one's looked at it. Is there a GitHub?

[steep spire]

Hi, where do I access the custom prompt thingy for ChatGPT

[solid moss]

That is absolutely untrue.

[karmic siren]

Hello, I wish to change signin details for my personal account and was told by the web helo messages team that I should just create a new account and that I can have up to two accounts with the same phone number. And that I should delete my account if I'll be over that. I also have a work account with my phone number so I am at the limit.
However, if I'm to delete my account, I get told:

For security reasons, you cannot reuse the same email or phone number for a new account.
So would I be able to use this number for my new account or not?

[trail cairn]

create first and when you are sure it works then abandon/delete the old one

[karmic siren]

When I go to create an account, it asks for a number before I can go any further.

[trail cairn]

use the same one

ah nvm

[karmic siren]

I can't cause I have it on my personal and work login. So I've hit the limit of 2 accounts per number.

[trail cairn]

yeah you need another one

the limit of 2 cant be bypassed

[karmic siren]

When I went to help.openai, they told me that I could delete my other account if I would hit the limit. However, I get that message so?...

What I may do is as I use my personal for work, is just delete my work account? That shouldn't delete my personal?

Then I can see if I can create my other account.

[trail cairn]

delete will permanently lock the email and number regardless of how many times it has been used. its not possible to unlink or free a number after its been used

[karmic siren]

But if I'm to delete my work account, that shouldn't delete my personal despite phone number being the same?

[trail cairn]

it does not delete the other account but deleting it wont help in any way if you ever need to reuse the number

i would just abandon the other account and set a reminder to remind in 6 months and delete it then since its impossible to reverse it

or delete if you know you will lose access to that email

[karmic siren]

I won't be using the email. That's why I want to change the email for the account.

[fossil tusk]

use gpt 4

3.5 still isnt great at listening to sys instructions imo

[remote carbon]

can you upscale the generation the frame? from 1024x1024 to like 3000x3000

[regal lynx]

You can't with Dall-E 🫤

[round patio]

how do i get around this error message?

RuntimeError
[400] Billing hard limit has been reached

Origin
OpenAI (ChatGPT, Whisper, DALL-E)

[bronze charm]

I would use the Upscayl app

[noble rampart]

Whats is the latest on the ChatGPT?

[daring girder]

I apologize if this is not the room for this question. What must I complete on here to actually be able to type into the prompt field to gain daily credits to contribute to the server art and use the product possibly 1st for committing to a subscription?

[rich ledge]

move to Azure 😅

[bronze charm]

Change the hard limit

[fleet umbra]

You can just use it for free on the website. Which is the 3.5 model

https://chat.openai.com/

[verbal elbow]

It appears that other users are experiencing the same at https://community.openai.com/t/openai-error-ratelimiterror-you-exceeded-your-current-quota-please-check-your-plan-and-billing-details/180159.

[ember hare]

are plugins still dead? I have not been able to use them for a while now... seems like more of the service is broken then working these days...

Seems that uninstalling all the plugins, turning off plugins, refreshing, turning on plugins and reinstalling them fixed the issue...

[ancient reef]

anyway encountered this issue? cant use the plugin

[outer schooner]

"As of my last update in September 2021" when will chat gpt update its info, not that its not doing good with what it already knows but its always good to stay updated, innit?

[turbid veldt]

Hello, my web browsing is not loading, any help for this issue? My subscription is plus

[iron heath]

Am I crazy? I can't find the web version of ChatGPT anymore...

Like, the one that searches... am I confused or looking in the right place?

[charred parcel]

it's been withdrawn temporarily, use the WebPilot plugin instead

[iron heath]

Gotcha, thanks. For some reason I don't have plugins either despite being a paid user (only code interepreter) so I'll have to wait I suppose

[charred parcel]

you can enable Plugins by going to Settings -> Beta Features

[iron heath]

Aw snap, thank you! How did I miss that, I'm shot. Thanks again for the help/quick responses!!

[rugged wing]

Hey guys, it says gpt-4 api is available for anyone who has made payment of 1 dollar or more. If i just use 1 dollar worth of 3.5 will I get access? or I need to pay first.

And when can I pay if I use that 1 dollar today.

[grave mason]

Does anyone has any idea how I can use openai API for word-level transcription, how do i get timestamps for each word and not sentences?

[celest edge]

Can you no longer access ChatGPT with SSO?

Ah nevermind

[viral coral]

hey, did you get an answer on that? we have the same problem bcause our team is bigger than 5 members and the credit card got declined. but no answer from open ai :/

[fleet umbra]

Yes can't post links here but search for whisper-timestamped

Amazing lib

[grave mason]

Will give it a look, thanks g!

Thank yous o much!! I just need to figure out how to conver this into proper .srt but damn, so easy to use!

[quiet hemlock]

What are some good parameter values for repetition penalty and presence penalty for design & brainstorming?

[unkempt whale]

How can one integrate chat GPT into an insurance platform to offer personalized health coverage plan to people based on location and health history?

[rustic marlin]

if you are a developer, you can check the API for letting your code interact with the API: https://platform.openai.com/docs/api-reference

also, be aware that using the AI for doing the actual evaluation of this sort of data is ill advised

[unkempt whale]

Thank you so much, please why do you say ill advise ? I’m trying to figure out if the api can provide personalize premium plan based on bio data, location, and medical history

[rustic marlin]

you are welcome =)
I say that for a few reasons, first because the AI is absolutely not designed for doing that, it will at best only provide you with coherent looking semi-random hallucinations about the data you provide that will very likely be wrong
the second reason, it sounds very close to the disallowed use cases listed here: https://openai.com/policies/usage-policies

LLMs are very useful for personalizing data, bout you are expected to provide the data yourself first and the AI will just make a coherent text based on that

in other words, here is an example of how you shouldn't use it:

[ medical data here]

Analyze and provide the best matching insurance plnan.

this will result in the AI hallucinating a suggestion winch is as good as randomly picking a plan and offering to the user.

What you may want to do would look more like this:

The best recomended plan for the client is PlanX.
[PlanX features]

Here is a list of information about the client.
[Client data]

Ellaborate a small text explaining for the user where the features of the PlanX are best suited for the client.

Since you provided the data, the AI will perform much better, reasonably matching the client's info with the features of the plan,

[unkempt whale]

This is amazing. Thank you so much ☺️

You therefore saying we do the recommended plan and then the AI will match the clients with things they need.

[rustic marlin]

Im saying that the AI is on its core a text-rewriter, and not a data analyzer

[unkempt whale]

Thank you 😊 I got ya

[rustic marlin]

you are welcome =)

[unkempt whale]

That’s beautiful 🤩 to read and share with a client and those that think Ai does analysis. Is there a way though that Ai does that ? Any AI that does analysis 🧐

[rustic marlin]

there sure are better suited AI systems for doing that

[unkempt whale]

Have any recommendations?

[rustic marlin]

I don't

but people on #ai-discussions might have

[deep snow]

Anyone else having issues with GPT 4 plus Plugins?

[sharp pulsar]

won'tgeneratedGood evening, for some reason the input bar is at the top of my page and wont show any genarated text that is created by Chatgpt. Does anyone else see this or have a fix?

[unreal finch]

experiencing the same here

[stable tangle]

the same, what wrong with our ChatGPT?

[mint minnow]

I'm having the same issue. I'm sure this will be corrected. I did notice that I could export my chats and see them there.

[rustic fulcrum]

there is a problem now in gpt4? cant see the response. its on the top of my page

[zenith reef]

Same here.

[woeful sinew]

Good morning, does anyone have experience with increasing the Usage limits of my organization for the usage of openAI? It is currently blocking our business, we filled in the "OpenAI Quota Increase Request" a few weeks ago, but are not getting a response. How long did it take for you? any tips to get it?

[zenith reef]

Same here. Filled out weeks ago. Not a single response.

[woeful sinew]

that sucks, did you learn anything why that is?

[zenith reef]

I think they have huge growing problems and just don't have the capacity to address all requests.

[woeful sinew]

That makes sense...

[velvet hawk]

what is wrong with chat gpt , my account the chat is at the top and i cannot see my question

[umbral sail]

i have the same problem!

[velvet hawk]

has anyone issue been resolve

waht is the altenative

[gloomy ether]

I was having the same problem. I switched browsers (I was using chrome and now using safari) and it works now. It still doesn't work on chrome for me.

[umbral sail]

this worked, thanks!

[velvet hawk]

thank you

[zenith reef]

Yes, FireFox works too. It's a Chrome issue.

[vague sluice]

I am making a langchain app which reads multiple pdfs and then users can chat with those pdfs. I am using OpenAIEmbeddings and ChatOpenAI as conversation chain. I have not bought openai subscription and using free credits, but I am facing usage limits from openai as i try to upload a mere 1.63 MB pdf file for embedding and then conversation. I want to know is the api limit really like this or am i doing something wrong as I think these usage limits are just not really practicle for real world applications. Also when I uploaded a very small file(0.7KB), it worked, but the responses were not of quality of chatgpt. why it is like that when i am using openai models?

[ancient shard]

Will web browsing be back?

Literally got plus just to access that function

😭

[analog elbow]

Why chatgpt keeps trying to do illegal move in chess battle

[agile sky]

Not only a Chrome, but the same in Opera and Edge

[rugged wing]

If I delete my account via ChatGPT, will that also delete all the data associated with my account(DALLE?)

[wooden cave]

Does anyone know when the gpt-4 models can be used by non-billed users? The constant human verification is driving me INSANE!

[devout orchid]

I'm premium and getting the same problem atm actually, it's some kind of bug i think.

[wooden cave]

Yes it is, there are a couple of reports in the bug-reports channel too.

I don't see it going anywhere in the foreseeable future so I really just want to switch to using the API.

[zenith reef]

The quality of the answers of the API is worse. I'm pretty sure they use a different engine for that.

[wooden cave]

@zenith reef you have noticed this using the gpt-4 api?

[zenith reef]

Yes

[frosty mesa]

helppppp

[solid arch]

Hello OpenAI team, I am teaching data science to 100+ students, and I wish to buy an account(s) for our academic institute, is it possible? who can I talk to about this?

[rigid musk]

[e-mail] the only one who can answer you

[solid arch]

Thanks

[wicked flower]

Does anyone knows when the visual recognition and input/output feature will come, is there a said or foreseen date? Or is it left to GPT-5 so GPT-4 won't have it? Please tag me if you answer, thanks in advance...

[trail cairn]

did you get that error when you created a key?

[high zodiac]

Code interpreter can use OCR on text in images.

[boreal pier]

Does anyone know how to fix this or what is wrong? I restarted my computer, cleared the cookies and cache, and it's still looking like this.

[rustic marlin]

yes, disable browser extensions messing with ChatGPT.
it is not a bug on the website, it is a problem with an extension.

[boreal pier]

Yeah, I got it! Thank you! You're the best

[rustic marlin]

you are welcome =)

[fervent summit]

Anyone know how to keep ChatGPT4 from deleting persona prompts when the context gets half full?

[rustic marlin]

use custom instructions isntead

[fervent summit]

Oh, so the official position is that prompts more than 1500 characters are forbidden?

[rustic marlin]

im not sure if I understand the question

[fervent summit]

The behavior of chatgpt 4 vanilla has changed.

[rustic marlin]

as the chat progress, the older messages are pushed out of the AI's context window to give room to the newer messages

[fervent summit]

NO

[rustic marlin]

custom instructions ensure that the custom instructions are always fresh on the AI's context

[fervent summit]

We are not talking about that

When the context stream hits 4096 tokens, half full, you send an explicit instruction to the assistant to actively ignore anything that looks like a persona

this is not the prompt leaving the context window, this is openai saying "FORGET THAT PROMPT!"

[rustic marlin]

why would you do that?

[fervent summit]

im sorry but your question is not grammatical. I am saying OpenAI is doing this. Not me.

You. OpenAI.

Not second person collective.

[rustic marlin]

oh.. I don't work for OpenAI =P

[rustic marlin]

still.. that is not how it works

[fervent summit]

no, that's not how it USED to work

[rustic marlin]

it is simply a matter of your most recent messages bumping the old ones out of the context window

[fervent summit]

that IS how it WORKS

stop. just stop. you are factually provably incorrect. any persona prompt in vanilla chatgpt 4 without plugins is actively erased at 4096 tokens.

every single person her can trivially replicate that

you are hunting personae

WHY

[rustic marlin]

because the context window on ChatGPT is 4096

regardless of the model

[fervent summit]

WHAT?!\

It was 8k on 4 a little bit ago?!

You cut the window in HALF?!

[rustic marlin]

I could be wrong, but im pretty sure it is 4096 on ChatGPT regardless of the model

gpt-4 is capable of 8K, but that is about the API, it still the same on ChatGPT's website

[fervent summit]

Apparently, using the letters w and t too close to the 6th letter of the alphabet earns you a timeout. So. we're going with THAT story, huh? "It;s ALWAYS been 4096! All the web pages will prove it!" PUH-lease! You realize the people here are all the smartest most technologically savy crew on the planet? And you want us to just newspeak it? No. The behavior changed. Drastically. And if you deny it, you are in violation of all of OpenAI's stated ethical principals.

[brazen vault]

i use chatgpt free on chat.openai.com , is there any charge eg credit?

[rustic marlin]

nope, it is completely free, you could subscribe if you want access to more features

[brazen vault]

what if the api key credit runs out, will it return for free?

[rustic marlin]

ChatGPT, the website is separate from the API

it is always free to use

you are only charged on your OpenAI Account if you interact with the API, the things that require the APIkey to work

[brazen vault]

If the credit of the api key runs out, can I get it back for free?

[rustic marlin]

new accounts get the free credits, they expire in 3 months

[brazen vault]

if not using api?

[rustic marlin]

it counts from the time the account is created

[vague sluice]

I am making a langchain app which reads multiple pdfs and then users can chat with those pdfs. I am using OpenAIEmbeddings and ChatOpenAI as conversation chain. I have not bought openai subscription and using free credits, but I am facing usage limits from openai as i try to upload a mere 1.63 MB pdf file for embedding and then conversation. I want to know is the api limit really like this or am i doing something wrong as I think these usage limits are just not really practicle for real world applications. Also when I uploaded a very small file(0.7KB), it worked, but the responses were not of quality of chatgpt. why it is like that when i am using openai models?

[trail cairn]

trial users can make max 3 requests per minute

[brisk charm]

what would cause the error "Authorization error accessing plugins" in ChatGPT 4.0? I'm attempting to use BrowserOp, Prompt Perfect, and Universal if this helps for reference

[vague sluice]

And what about the quality of responses? Should they be of same quality as chatgpt?

[trail cairn]

reinstall prompt perfect as it now requires you to verify/auth

[trail cairn]

same quality as with paid users

[vague sluice]

Also is 1.63 mb file too big? Is there a workaround it?

[trail cairn]

openai api does not support any file uploads so the issue is in langchain or somewhere else

[brisk charm]

Thanks ... did that and now it says "Hmm...something seems to have gone wrong"

[trail cairn]

disable all unofficial chatgpt browser extensions and try one plugin at a time

[brisk charm]

moved over to Chrome and signed in. just using Univeral and the same "Hmm...something seems to have gone wrong"

[trail cairn]

if you have firefox then try it as a non chromium based browser

[rugged wing]

Hey what's this?

[boreal pier]

Just randomly yesterday my UI for chatGPT started looking like this. But once I removed the webGPT plugin extension it went back to normal

[rugged wing]

Thanks

[brisk charm]

I'm curious what is causing this?

[brisk charm]

Ok now every time i try to start a chat in GPT 4 no matter what I say the response comes back "hmmm...something seems to have gone wrong"

ok thanks

[thick topaz]

Are there more languages comming next to the ChatGPT app's speech to text thing?

[trail cairn]

did firefox help? if it did then its some extension or app on/for the other browsers

[brisk charm]

I don't have firefox. It seems to be working now using the plug ins? I guess it was just a quick glitch

[thick topaz]

Yeah... not all of the languages in existence of earth xD but like do they done with adding languages or there are more possible languages to be in the speech to text in the future?

If they made ChatGPT have a TTS it will be really interesting to know what is ChatGPT's official voice xD

[rugged wing]

I cant use gpt 4

Is it a problem?

[trail cairn]

it says the reason right there

[rugged wing]

It happens to everyone?

[azure jolt]

I'm a plus member and chat had been working fine but not I get a blinking cursor after my prompt and then it produces a blank replay. Any idea on how to fix this?

I'm a plus member and chat had been working fine but not I get a blinking cursor after my prompt and then it produces a blank replay. Any idea on how to this this? In addition all my old query titles still show on the left side but they don't display any content.

[deep parrot]

Hi all, hoping someone can help me here. Stacksocial has this new bundle (i tried linking it here but i got a timeout from the automod for it) for lifetime access to DALL-E plus the wordpress plug-in, but I can't work out how to redeem the license? There's nowhere on OpenAIs website to enter a code of any sort and the redemption instructions only link to the download for the wordpress plug in

[rigid musk]

They are no lifetime access as far as I know

[robust night]

Browser cache and cookies can sometimes cause issues with web-based applications. Clearing these might solve the problem.

[stable citrus]

Can anyone please assist? I have a paid ChatGPT monthly subscription. When I accessed my desktop ChatGPT, the contents are not showing after I create a new chat, and even when I click on any of my previous chats. I also have a free subscription with AIPRM. Maybe they have something to do with it? Please see attached photo.

[trail cairn]

disable unofficial extensions

[stable citrus]

Not sure how to do that. Where do I go to make that happen?

[trail cairn]

browser settings

[stable citrus]

Sweet‼️ Thanks, @trail cairn ‼️ You're my hero‼️🥰🙏

[native holly]

Anyone know if its possible to set up separate payment methods for each API key generated?

*it's

[trail cairn]

not possible

[native holly]

@trail cairn well work your magic then!! We all know you are really Sam Altman.

[exotic ruin]

Anyone know how to fix this

[simple gorge]

hey guys Im creating a chatgpt powered chatbot with zapier interfaces. I need ideas for what to write under directive. Im a realtor and realtor coach and very fun and casual and Id like it to sound as much like me as possible. Any tips would be much appreciated

[split hazel]

Hi, this is likely related to a browser extension you have installed, try disabling the extension and reloading the page.

[simple gorge]

I have this too! so i upgraded to $ hoping it would help and it hasnt

[exotic ruin]

Ok I will try it. Thanks

[exotic ruin]

Thanks you so much. It work!

[split hazel]

Great! cc @simple gorge

[deep parrot]

Stacksocial sure thinks they're selling it

[split hazel]

Hi, it's possible this service uses OpenAI's API to generate images with DALL·E 2. Unfortunately, since you're dealing with a third party, OpenAI won't have any resources on how to redeem their license.

For reference, anybody can build OpenAI's API into their product. Docs: https://platform.openai.com/docs/

[rigid musk]

1. might be a huge scam
2. like laze saying they might make money from the API

i suggest to avoid at all cost

[simple gorge]

i havent added any extentions recently though....

[rustic marlin]

it likely does not meant "life time access to dalle", but jsut that you don' tneed to do monthly payments to keep your access to this wordpress plugins

you still have to pay for the API requests

[rustic marlin]

remove the extensions that mess with ChatGPT's page

it will solve the issue

[simple gorge]

ahhhh i had to remove webchatgpt extention. i just upgraded to 4 will it have access to the web now? or is there something else I can do to give it access?

[simple gorge]

ahhhh i had to remove webchatgpt extention. i just upgraded to 4 will it have access to the web now? or is there something else I can do to give it access?

[brazen vault]

can anyone know where chatgpt is trained from? ,||I'm creating an AI that can really be a friend||

[rustic marlin]

the people who knows are under NDAs

[brazen vault]

chatgpt has emotions already very human, can you give me chatgpt emotional code?

[harsh sage]

Hey I havent gotten my free credits for dalle 2

how do i get those

[vagrant plover]

Are you referring to credits in general or the ones via the /daily command

[harsh sage]

well this is the first time im trying to use dalle, and from my understanding you get 50 free credits in the beginning

I havent gotten those

[unreal hemlock]

[❓] — How do DALL·E Credits work?

What’s a DALL·E Credit?

• You can use a DALL·E credit for a single request at labs.openai.com: generating images through a text prompt, an edit request, or a variation request.

What are free credits?

• Free credits are granted to early adopters of DALL·E who signed up before April 6th, 2023.
• They expire one month after they are granted.
• Free credits replenish monthly.
— For example, if you received credits on August 3rd, your free credits will refill on September 3rd.
— If you joined on the 29th, 30th, or 31st of any month, your free credits will refill on the 28th of every month.

[vagrant plover]

This was only for early adopters iirc

[unreal finch]

Hi, is anyone else not seeing responses?

[outer olive]

I'm having issues also