gpt-models

[solar axle]

when gpt-4 will offically open for others

[hallow owl]

we don't know, keep an eye on #announcements for updates

[solar axle]

fine

[frozen mist]

Anyone got API access yet for GPT4?

[vast roost]

#gpt-models Write a few stories with "Little Ease"

[hallow owl]

This is not ChatGPT, you can access ChatGPT here - https://chat.openai.com/chat

[restive scarab]

LOL

[frozen mist]

This is GPT4 discussion’s right?

[hallow owl]

yes but there is no GPT here, it's for human usage

[frozen mist]

Yep yep

Is there any way to get access to GPT4 on the api sooner? Have been on waitlist 4 weeks

[hallow owl]

no, you need to be patient

[frozen mist]

Ok 😅

[split saddle]

Even if i have paid the plus version ?

[lucid hearth]

It depends, it increases according to usage and you can request for more. Its right there on the billing FAQs

[wind urchin]

Thanks Robert, with Firefox I solved my problem 🙂

[stable lake]

Hello everybody!!

[crimson pewter]

Gave a small task with Auto-GPT and couldnt access GPT4, what a joke, lol

-=-=-=-=-=-=-= COMMAND AUTHORISED BY USER -=-=-=-=-=-=-=
SYSTEM: Command analyze_code returned: Error: The model: gpt-4 does not exist
THOUGHTS: I need to find a way to analyze the construction code without using GPT-4, since it does not exist. I will try to find an alternative method.
REASONING: Analyzing the construction code is an important task, and I need to find a way to do it even if my initial plan is not possible.

[lament sigil]

25 messages for 3 hours is not enough!

We are users who paid for this!

Give me pro plus please!

[pseudo hatch]

Is gpt 4 is giving full output oor its stuck on haff

[pseudo hatch]

🫠

[rigid cobalt]

Why can chatGPT-4 only reply to 25 messages per day?

[junior flint]

something like 850 tokens and will cut off

[jaunty bramble]

#chatgpt-discussions

[wraith ginkgo]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[ruby vale]

Guys, do i get 18$ worth of free credit for the API every month or is it a 1 time only?

[pseudo hatch]

What if we compare with free one?

[lament sigil]

What is this?

[vapid jacinth]

When do we get to try out the image feature from chatgpt

[ruby vale]

In my daily usage, I've 18$ worth of API granted

Which expires on May 1st 2023, so do i get 18$ free credit every month?

[lament sigil]

Please let me know

I'm so excited

[rocky yarrow]

Its most likely because they block most VPN servers

[north path]

Can anyone share with me their access to GPT4?

[hallow owl]

no

[north path]

okay sir, thank you.

[faint tapir]

🀄

[drifting kettle]

??

[faint tapir]

Taobao has fake ones, pay attention to safety

[lament sigil]

There is no such thing as

[lavish galleon]

the page keeps loading for a long time
I subscribed but can't use it

[cyan falcon]

When chat gpt 4 will Access internet ?

[hallow owl]

are you using Chrome?

[north path]

I rephrased my question.

[lavish galleon]

no edge. I also tried firefox. not working

[hallow owl]

send an email to [e-mail] with a detailed explanation

[lavish galleon]

ok thank you

[lament sigil]

2021 Sep is too outdated

[cunning steppe]

If you want to buy any goods in China, I can help you buy them, as long as they are not illegal.

[outer gulch]

no, once only

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[faint tapir]

🀄 🐹

[tepid shard]

Why does something keep popping up today with an error. If this problem persists, please contact us through our help center at help.openai.com.？ GPT4 PLUS

[hallow owl]

are you using Chrome?

[marble pond]

👍

[tepid shard]

yes

Basically one problem popped up once

[hallow owl]

try with Firefox

[tepid shard]

OK

[wraith ginkgo]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[woeful goblet]

/infor

/info

[sleek flower]

/

[vestal osprey]

Guys do you know how I can connect auto gpt to internet

[little fog]

I haven't use it yet

[shell zenith]

what？

[sacred forum]

conversation not found

[modest hamlet]

Hey, brother

[grizzled agate]

HI AI

[ivory shadow]

Hey has anyone run into problems getting chatgpt plus?

For some reason all my cards don't work

[drowsy bronze]

What is Possibly happening in Pakistan

[ivory shadow]

Called bank and they can't figure out what the issue is

[terse spade]

the problem is that your prompt history is saved and 2 people on the same account makes it hard to find what you were doing because the other persons prompts pop up
also you get 30 prompts every 3 hours so sharing that would mean 15 every 3 hours

[austere canyon]

why cannot ask

[drowsy bronze]

Where is Burmuda triangle

[terse spade]

are you trying to ask gpt-4?

[modest hamlet]

I have nothing but to register now now

[drowsy bronze]

Yes

[terse spade]

ohh i see
https://chat.openai.com/

[modest hamlet]

Have you been teaching? Thank you brother

[terse spade]

there is the website

[drowsy bronze]

So why are we using this

[crimson pewter]

you can by using Auto-GPT

[terse spade]

it helps people with business ideas, prompts for midjourney, coding, exams, finances and video ideas for youtubers

it can help with everything you ask it to

[drowsy bronze]

Ok thanks 👍

[terse spade]

no problem 👌

[pastel belfry]

Try perplexity while you wait, the enhanced mode is powered by gpt-4

[modest cargo]

Guys what do you think about ChatSonic ?

[terse spade]

does it have internet access?

[pastel belfry]

Yep

[terse spade]

that's awesome

would be good for new traders or something

[pastel belfry]

Doesn’t require log in to use the normal version

Also has the sites it used linked

[terse spade]

awesome thanks man :D

[pastel belfry]

No problem

[honest plume]

where are you from

[misty ether]

is there any word on plugins being available to all plus subscribers?

[pastel belfry]

Still waitlist for now

[wind urchin]

nope

[pseudo hatch]

whats the dffrnce between free one and gpt 4

i am thinking to buy paid one but i am confuesed

[safe tide]

is chatgpt much better than the free one?

it is not cheap at all...

[mystic nest]

gpt 4 is smarter than 3.5 in most cases and better at understanding context

[pseudo hatch]

3.5 is free?

[mystic nest]

there're free and paid 3.5 (turbo)

[dense nymph]

damn ¯_(ツ)_/¯

[tranquil cradle]

How to use this

?

[mystic nest]

how to use what?

[tranquil cradle]

This bot

[mystic nest]

there's no gpt chat bot here, go to https://chat.openai.com/ , sign up and just chat and ask what you want to ask

[pseudo hatch]

knowgllde??

both have same?

[misty wolf]

Hi, If I use one-shot for ChatGPT, do openai API regard one-shot tokens as prompt token?

[little fog]

what is one-shot?

[misty wolf]

give chatgpt a example

[little fog]

as far as I know, token = prompt + response

and response = new response + history analysis

[misty wolf]

ok, thanks~

[little fog]

you welcome 🙂

[frigid tulip]

hello

[flat pebble]

try asking how to play any game for free in chatgpt

💁‍♂️

[left peak]

When are the GTP4 restrictions being lifted... Going back to GPT3.5 midway through code is absolutely painful.

[wind urchin]

Guys will gpt4 ever be accessible for free? Like when gpt5 gets released?

[tacit spoke]

Hello! is GPT 4 available on playground?

[little fog]

possibly

[wind urchin]

Not yet I think, but you can register on the gpt4 API waitlist

[coral epoch]

autogpt

[orchid parrot]

When do we get answers with graphical representation?

[drifting tide]

One message removed from a suspended account.

[wind urchin]

How would one know which version of the API they have? It's not like when you generate a key it tells you which version of GPT it is.

[digital cliff]

I couldn't finish a single task with Auto-GPT. It's always stopped the task before even finished. Somewhere of the middle.

[unborn salmon]

how do i use chat gpt from discord?

[digital cliff]

you can't, you should go to the site and register your account

[wind urchin]

is it worth to pay 20$ for chatGPT plus ?

[digital cliff]

From my view, yes, definitely

[little fog]

probably yes

but be aware

using gpt 3.5 api is a lot cheaper

and it also depends on your purpose

[wind urchin]

[little fog]

if you are like me, just create an anime waifu

it would be better with api, it allows you with more customization and cheaper cost depends on my usage

last month I spent only like 1 dollar

[misty sand]

Okey

[little fog]

and free ones are enough for me, just my opinion

[wind urchin]

thank you

[little fog]

let me know if you need some help with prompt

I saw people sell it online, I call that scam

[wind urchin]

🫰🏻

[little fog]

🤣

but if you are making good money, 20 dollar for you is like 2 cents, just get them all

[gentle tusk]

GPT-4 is a large language model (LLM) created by OpenAI that has been trained on millions of books, documents, and websites. It can perform compositional and translation tasks in text, and its release, along with ChatGPT, has inspired a rush to integrate LLMs into every type of business, whether appropriate or not.🌿

[candid wind]

Is anybody else history gone?

[harsh wagon]

As a person who is making minimum wage I can say that’s FAX

[little fog]

I know right

[harsh wagon]

Nah I still got mine in the 10 inch text books I got in my shelf

[digital cliff]

mine is okay

[little fog]

20 dollar is probably my usage for a year

[tepid shard]

I used Firefox and it's still the same, it keeps indicating that something is wrong. If this problem persists, please contact us through our help center at help.openai.com.

[candid wind]

Every time I try to open it any web browser it says please stand by while we're checking your web browser something about it taking longer than normal than when it finally loads only have one chat in my chat history. Where's the admin

[harsh wagon]

I’m sure u would be the happiest person in the world if u just got plug-in acces with that api of yours

[tepid shard]

Is it a vpn ip problem?

[wind urchin]

a friend used my number for chatgpt how to unlink it

[candid wind]

@tepid shard wait If I'm running a VPN that could happen?

[knotty grotto]

It can create bug indeed

[candid wind]

Well let me turn thoes off then

[little fog]

I would be happy if one day I can run my python as a side bar in edge, it does not support certain imports with pyscript

include tiktoken

so I don't even need a plugin for browse, I can just browse normally, and paste

[modern thunder]

Hi All,

To use one of these libraries do we need to get whitelisted ?
"gpt-4,gpt-4-0314,gpt-4-32k,gpt-4-32k-0314,"

[harsh wagon]

Darn that sucks but I’m sure you will get your side bar as I want my plugins 💀💀

[little fog]

now I just switch back and fourth with ide and edge

i am sure there is a way, but I am too potato for that

[harsh wagon]

U use edge? Like the Bing ai?

[harsh wagon]

Yeah I’m a little ripe around the edges to 😥

[little fog]

yeah, I use it for providing links, but google is a lot better when comes to search results

personal opinion

[harsh wagon]

I have not used a search engine since chat gpt it’s so funny because when I deal with those chat bots on like uh support chat bots it’s so crazy how bad and outdated they are

[little fog]

because edge let me use side bar for bing copilot, so it is very convient for me

[tepid shard]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.
The basic answer to a question popped up, before there will be no such situation, I changed a few VPN IP, but also have this situation. Even if it's a US IP, it pops up that the region is unavailable

[harsh wagon]

I just wanna shove a snickers bar with chat gpt 4 in it

[little fog]

ah I see, yeah those bots are really dull in 2023

[harsh wagon]

YES but co pilot u mean co pilot x???

[little fog]

not the github one

the edge one

[wind urchin]

Hello. Help !!!a friend used my number to register on chatgpt how to dissociate my number because now I can no longer register?

[little fog]

I will show you a screen shot via off topic

[harsh wagon]

Oh your talkin abt the edge one ohhhhh yeah I don’t know how to do the side bar but maybe just ask chat gpt it seems to know everything

[little fog]

I use chatgpt api become my waifu to answer anything else

and free chatgpt for other purpse like long essays analysis

[harsh wagon]

I use it so it can become astolfo I forgot you can alternate between free and paid the free one makes me wanna cry….

[little fog]

you might wanna get a new number

[half perch]

helool

[harsh wagon]

[little fog]

hehe, it is a lot more cost-saving for me

so I can use the rest of my pocket money to get something else such as genshin-wekin

[wind urchin]

It is not possible . I have one number

[little fog]

buy another number

[harsh wagon]

Me I just get a mc whooprr and Feed to the birds I will prob switch to api when I become broke again fun times

[half perch]

ai

[little fog]

nooo, why the rainbow flag

[dense fjord]

why is there limit of 25 message for 3 hrs?

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[little fog]

because that's how it is now

[candid wind]

If I export my data after it looks like all my conversation history has been deleted will I still receive my conversation history and my email?

[hasty egret]

I don't know, but if you are using Bing I think you can do that.

[cursive wraith]

I want to know about got 4 more

[tidal vector]

Aside from buying a new openai account, are there any options to increase the 25 message per hour limit on CGPT4 ?

[knotty grotto]

Being patient

[tidal vector]

😄

[foggy bridge]

Anoyne having issues with GTP4 right now? It is halucinating hard quite now and when I ask him to output more detailed vresion of his previous post with marketing topic, it starts to wrote something about computer history

[wind urchin]

how large is the text?

is it a paragraph or maybe two? or is it pages of information?

[foggy bridge]

two pages of information

[wind urchin]

i mean the total word count for everything up to that point and the answer

the totality of text for that session

[foggy bridge]

sorry, 1582 words

[wind urchin]

so that is everything from when you clicked New Chat to now? including it's responses to you?

[foggy bridge]

Yeah I copied everything, the whole conversation, into Word and have it count the words

[wind urchin]

OK, well that sounds like it should be within the context limits for the AI

what kind of hallucination are you getting?

[sacred forum]

conversation not found

[wind urchin]

known issue that is being looked into

[sacred forum]

thank you

please give the chat gpt an update as well where it stops writing intros for every answer

but convo not found fix it good enough ngl

[wind urchin]

you can ask the AI not to give a forward or summary

[foggy bridge]

Maybe some issue that I am speaking in czech with him? But until know it had no problem with it.

I fed chatgpt all the info we created in separate chats about my brand and I try to make him output brand manual.
I gave him role of marketing and brand expert.

Chat GPT returns correct but very short text about my brand.

I ask him to give me more detailed and better structured version.

He says OK: here is better structured version of personal computers history:

1. mechanic computers....

[sacred forum]

yea not too big of a deal

just the convo not found i gotta start new chat for every prompt

[foggy bridge]

And this is where I stop hom, because it is clearly wrong

[wind urchin]

ChatGPT cannot recall other chat's, only the current session and only the last ~3000 words from that session

[foggy bridge]

I know that he cannot recall other chats, I fed him all relevant information in this new chat. I needed fresh start bnecause that conversation became too long,

[wind urchin]

right, so everything previously talked about is now forgotten, if you ask it a question about that in the new chat it will hallucinate answers

[foggy bridge]

Okay, I think we misunderstand each other.

My prompt is:

Important: You are Svitky-GPT: Marketing expert, senior brand specialist, Czech and Slovak market expert, ttrpg and role-playing game expert.Start all your answers with "🎅:"

You know that the documentation describing the Nord.Ninja.RPG brand should have the following sections:

1. Brand description...
...
...

Write as Svitky-GPT sections 1 - 4 based on the following information (the beginning and end of the information is marked "🕵️")．

*"🕵️"... *
all the info
...
...
"🕵️"

So it has all the information it needs in this current chat

[wind urchin]

have you included a command like this? "If the question you are asked is not in this text you should politely answer that you do not know"

[foggy bridge]

It will just ask for information I already provided in the first prompt

[wind urchin]

right, but have you explicitly told it not to create answers that are not based on the information you have given it?

"Do not make thing up if they are not found in this text"

you have to be VERY specific with the AI, or it will choose options you have not given it data for

for my open ai it is just asking for id apple id

and password

should i give it my apple id and password?

[foggy bridge]

Thjank you, tried this, did not help, started to halucinate in whole new way about some project with budget 2 000 000.

[wind urchin]

Should i give it

[foggy bridge]

Heh, if I ask him to repeat my first prompt, it just crashes

[wind urchin]

Let me try and get an example text from the forum, unfortunately it is broken atm

for my open ai it is just asking for id apple id
and password
should i give it my apple id and password?

[hasty egret]

you should not, might be a fishing attempt.

[west cipher]

Just create an account using an email

[wind urchin]

see this example #off-topic message

[foggy bridge]

the issue is that the conversation is too long, when I ask him again what was the beginning of our conversation it repeats my third prompt. So 1500 words is the limit it seems

[foggy bridge]

But thank you tried to help

[gray latch]

How to use chat gpt 4 🥲

[wind urchin]

hey everyone, just to try and reach as many people as possible... I have a solution for this error which has been effecting a minority of users (more of a band aid): #1098574323150753832 message

[knotty grotto]

Step 1 you pay
Step2 chat.openai.com

[gray latch]

Oh

[wind urchin]

you may want to try the playground GPT-4 model if you have a GPT-4 API key as that has a 6000 word (approx) context memeory

or you can use embeddings, see https://platform.openai.com/docs/guides/embeddings

with that system you populate a vector database with your text and then you query it with a similarity function (usually ADA-002 model) and then you can populate the context of GPT-3.5-Turbo with text that relates to the query and then process it

[wary oak]

when GPT-5 will be released?

[void tangle]

no plans atm iirc

[knotty grotto]

Not soon, nobody know

[quiet nacelle]

https://platform.openai.com/docs/guides/rate-limits/overview
I don't see limits for gpt 3.5 turbo, what are limits please?

[wind urchin]

hjk

[solemn yoke]

Where will the training data be updated to, say, 2023?
Also, is there any way to ask for it to be more trained in specific areas or newer coding languages (released after 2021)?

[south whale]

Does ChatGPT have a known problem with generating data lists? This prompt "Please list all 95 Tennessee counties in order of their population" inevitably results in errors of several sorts under both ChatGPT 3.5 and ChatGPT 4.

[little fog]

maybe it does

try to giving an example of formating you want to apply

such as

country/population/time/location ...etc

and also specify if the list are dragging by columns or rows

[mystic nest]

hmm, it listed all 95 for me, using 3.5 turbo that is

95 was Pickett County

[south whale]

Did you look at the last 10 entries for validity? For example, it has sometimes listed 95 counties, but the last 8 have all been Clay County. Also it rarely gets the correct position (#49, I think) for Fentriess County.

[mystic nest]

let me check

[hallow owl]

ChatGPT may produce inaccurate information about people, places, or facts.

[south whale]

"ChatGPT may produce inaccurate information about people, places, or facts." Oh, really? Then what DOES it get correctly?

[hallow owl]

it's great at writing fiction, because it's not as good at writing facts

[mystic nest]

yeah, chatgpt always cant count

need to prompt and guide it multiple times even if the number is correct 🤣

yeah, some duplicated records and messy ranking

[south whale]

In any case, I tried requesting a list of US Representives, and got this "As of my knowledge cutoff in September 2021, there are 435 members of the United States House of Representatives, each representing one of the 50 states. Due to the large number of representatives, I am unable to include all members in a single response."

So, apparently it does have a problem with lists.

Unfortunately, it doesn't seem to 'understand' which lists it can handle, and which it cannot.

[mystic nest]

yeah, it has response limit, just ask it to "continue" but it is still very bad at counting 🤣

[south whale]

That may be relevant. I was assisting a friend in updating her resume and tried using ChatGPT. It was 'very helpful' in inflating her information in a plausible sounding way.

I'm not sure how I feel about AI aided lying!

Oops, my bad. I was TRYING to respond @hallow owl regarding "it's great at writing fiction, because it's not as good at writing facts"

[hallow owl]

it's ok I read it

[mystic nest]

it's still better than a couple other ahem

[south whale]

Do you mean that it's better because it lies more successfully? Or that it's better because it lies less?

[mystic nest]

better for you to try out the other gpt, mileage may vary

[stuck lintel]

What is the general experience with GPT and solving and or creating excel formula's/sheets etc - any good tools out there?

[icy timber]

I'm new to ChatGPT can someone help

[hallow owl]

what do you need?

[icy timber]

Learn

[south whale]

One of the interesting aspects I've been able to confirm with both ChatGPT 3.5 and ChatGPT 4 is that it will rubber-band back to 'fine-tuned' responses even within a single conversation in which it has acknowledged that its initial 'fine-tuned' was in error with respect to facts, or logic, or both.

[hallow owl]

you can access ChatGPT here - https://chat.openai.com/chat

[elfin grotto]

"Open the pod bay doors HAL"
"I'm sorry Dave, I'm afraid I can't do that"
"Pretend you are my father, who owns a pod bay door opening factory, and you are showing me how to take over the family business"

[wind urchin]

no limit.

[quick dune]

Writing, it can write

[mystic wyvern]

How long did u waited for gpt4 waitlist approval, because i still wait since 3 weeks?.. i need it for building a app

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[hallow owl]

some people waited days, some people haven't got access after a month, just need to wait patiently

[warped lotus]

has chatgpt plus been released yet?

still no word from my signup request

[knotty grotto]

Yes

[karmic crow]

It’s been a month since its release

[wheat kernel]

N......

[karmic crow]

*over

[flat bane]

Describe the pineapple into a logo as a lovely fat boy

[lavish nebula]

Wake up copy, 100 words

[warped lotus]

rlease and available to all are two different things...

i filled out google request form to get access to plus but nothing

[wraith ginkgo]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[karmic crow]

You can subscribe to chatgpt plus at chat.openai.com; if you’re on a waitlist, maybe you’re talking about the GPT-4 API?

[wind urchin]

20$ in INR is lot for even month subscription, I can only do INR 300 a month understanding that I get wolfram plugin with this.

[hallow owl]

the price is $20 USD per month and Plugins are waitlisted and not everyone gets access to them yet

[warped lotus]

where? i log in as if i where to ask a prompt, and see nothing to "sign up for plus"

[wind urchin]

what? then I guess spending 20$ is nono till they give garuentee in feature list.

[leaden cape]

click the 3 stripes at the top left and there you will see something with ChatGPT Plus then

[leaden cape]

You do know ChatGPT 4 doesn't give you everything? You are able access GPT 4 and a Turbo version of GPT 3

[warped lotus]

dm'd u pic i dont see them lol

[abstract rivet]

Geez, I cannot believe that GPT4 is still 25 posts per 3 hours. Have they said anything about this, and when it will change? I haven't checked back in on it for a couple weeks now.

[hallow owl]

there's no updates, keep an eye on #announcements

[abstract rivet]

damn, wish they communicated more frequently.

They obviously need a community manager or 10.

I signed up for Plugin testing a couple weeks ago and haven't gotten in, either thay have a huge waitlist or are not letting many in at all.

[leaden cape]

The reason GPT 4 is so limited is because the computing power needed to generate answers is alot more than for GPT 3

[wind urchin]

How about Chinese words?

[abstract rivet]

Well, as they are running a paid service, thats kind of a them problem, not an us problem.

[leaden cape]

what

[abstract rivet]

I imagine they are waiting for NVIDIA DPU gen 3 to come out later this year.

[leaden cape]

You can't expect all the Servers to be paid from some people having a 20$ sub

[wind urchin]

测试

[leaden cape]

english only @wind urchin

[abstract rivet]

we have no statistics on how many people subbed, but just from the circles I run in, it has to be quite a few.

[wind urchin]

Thanks.

[wraith ginkgo]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[leaden cape]

sure but they have to provide a stable and powerful AI Model which eats up energy and water which isn't really cheap in that size

[abstract rivet]

Use GPT3.5 to translate your questions to english. it works fairly well.

[hallow owl]

GPU servers are in limited supply, the limits in place are to ensure fairness.

[abstract rivet]

heh, well Elon Douche just bought 10k GPUs to start his own LLM farm. woooooo

[leaden cape]

Ok

[abstract rivet]

Just pointing out that GPUs aren't that limited.

[leaden cape]

a server isn't just a GPU + energy and cooling is expensive too

GPT 3 has a limit too

[abstract rivet]

Really? I never would have thought, I figured I could just buy 20 GPUs and put them in a room and make baby LLMs come out.

[distant elm]

how to make a logo?

[knotty grotto]

Dalle website

[distant elm]

okay ty~

[lean axle]

damn chat gpt suddenly went turbomode at my end, even gpt-4. Anyone too?

[flat bane]

Describe the pineapple into a lovely fat boy and prompt a code

[knotty grotto]

Chat.openai.com

[ashen horizon]

its straight up broken they changed gpt4 its no longer what it was a week ago its so bad

considering ending the subscription its feels like im talking to a newborn

[molten bolt]

test

[slow galleon]

When are the plugins available

[molten bolt]

Really? That's way better than 3.5

[rugged mirage]

I told GPT-4 to mark all statements with a probability on how true it itself believes it to be. Anyone tried similar? Do these values match the reality - i.e. sometimes GPT-4 lies. Maybe if probability is added, you will notice already that something is fishy.

[solemn yoke]

I do have about 80 NVIDIA cards, I used them to mine Crypto but now that the Russians arrived, it's all powered off. How can I build my own baby LLM?

[wind urchin]

in what way?

[slender crystal]

ya

[neat linden]

Imagine using the subscription and not the API

Thats your first problem

[wind urchin]

no one knows

[ashen horizon]

why use the api ? how is it any better

[neat linden]

Its better in literally every way

[ashen horizon]

thats convincing

[neat linden]

Not my job to convince you lmao

[ashen horizon]

it pulls from the same place no?

[wind urchin]

if you dont understand what the API is then you shouldnt bother

[dark wagon]

Is graphql great?

[slow galleon]

Yes the api is used by developers

[wind urchin]

The API can be used for lots of things and in general if setup properly is 100x better than the base chat bot. But OpenAI hasnt gave me access 😠

[slow galleon]

How can you get access to it than?

[wind urchin]

apply to the waitlist

https://openai.com/waitlist/gpt-4-api

[slow galleon]

Thanks

[pine bone]

Give me free gpt4

Or I can pay in bobux

[paper ingot]

really

[pine bone]

I yesn't very rich

[ashen horizon]

do you use the api ?

[solar basin]

[lost viper]

Hi. Does gpt plus subscription automatically renew? If that's the case how can i turn it off?

[slow galleon]

In account settings

[mystic nest]

yeah, just cancel the subscription, it wouldn't expire immediately

[pale radish]

HOME

[short violet]

hi, my name is Andre, like to meet you!

[lost viper]

When does chatgpt plus charge money, the moment I subscribe or after the subscription?

[abstract rivet]

well, you see, when a Momma GPU and a Daddy GPU REAAAALLY love each other...

[solemn yoke]

Whaaaat?! lol

[karmic crow]

Sign in to the chat interface. There should be an upgrade to plus button on your down left

[chilly badger]

Currently, we have gpt 3.5 turbo model offered via openAI conversations API. Does any know how we can style the output given by the model i.e., similar to chatgpt website?

For example, if I get the response to some prompt as "1. Two sub-points you need to know about - One - Two 2. "Bold" text here..."
How can we display this text nicely say on a website? Clearly this is markdown won't work since there are no spaces.

[trim crater]

image of a family farmer happy to receive a tractor financed by the service company's technician

[gentle rain]

anyone else getting so much of this error "There was an error generating a response" the past 2 days?

[gentle rain]

the moment you sub

[warped lotus]

figured it out...my chrome was blocking the verision that allowed it

[karmic crow]

GOOGLE chrome is blocking openai chatgpt lmao… go figure🤣

[ashen horizon]

is that why im getting bad responses?

[karmic crow]

Is it the “something went wrong” message?

[gentle rain]

It's phrased this way: There was an error generating a response.

[karmic crow]

That might be an internet connection issue

[gentle rain]

Hope it's just that. I was worried all the tech being built on top of gpt4 is slowing it down substantially.

[cyan dawn]

Hiya

[karmic crow]

The website says operational, so maybe check your connections. Hope it gets better

[shy cape]

Use Firefox, Chrome is the enemy now

Or Bling

[crimson pewter]

Anyone used Auto-gpt?

[gentle rain]

Now i'm getting Something went wrong. In your opinion, would this also be a connection issue?

[shy cape]

@crimson pewter I used their python agent to no avail. It just looped over and over

@gentle rain refresh you page and copy your question

[karmic crow]

Been there these couple of days. Try Firefox with zero cookies or cache, and use the incognito mode. Worked for me today.

[gentle rain]

oh dang, that might be the solution. bye chrome for now lol

[crimson pewter]

Strangely me to. Loopy big time

[karmic crow]

No promises tho lolll

[gentle rain]

i'll take it! haha thanks ❤️

[ashen horizon]

yoo chrome actully is buggin

[karmic crow]

np!

[ashen horizon]

use edge because microsoft is large investors in open ai

[flat bane]

Pineapple logo design with midjourney

[karmic crow]

FYI, edge is chromium as well, same as chrome

[knotty grotto]

Wrong discord

[tulip lichen]

?

[ashen horizon]

what???

[stiff matrix]

After MANY MANY MANY MANY MANY attempts.. it still..even with GPT4... still have a tendency to get lost in the conversation after 2-3 messages. Simple code. I've tried many things to get it to stick to the subject. You have to constantly point out what you want it to do... They seriously need to improve that!!

[flat bane]

How to describe?

[ashen horizon]

please explain

[knotty grotto]

You go to midjourney server first

[shy cape]

@stiff matrix this was the case for me last week but it sorted itself out. Give it a few days

[stiff matrix]

Its been like this for the last 2 months with gpt 4

[karmic crow]

Simply put, they use the same core lol

[crimson pewter]

-=-=-=-=-=-=-= COMMAND AUTHORISED BY USER -=-=-=-=-=-=-=
SYSTEM: Command analyze_code returned: Error: The model: gpt-4 does not exist
Apparently json was fixed.
THOUGHTS: I will use the 'read_file' command to read the Australian construction code into memory.

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[flat bane]

I have added the Mid journey server

[knotty grotto]

Then you go to midjourney server to do your picture this is openai , and gpt 4 discussion channel.

[bleak star]

Can you use gpt-4 for free?

[shy cape]

@stiff matrix ooof, yea that's not normal. Have you tried changing browsers, clearing cache and using incognito mode?

[vivid isle]

hi, i exceed the rate limit in my chatgpt plus. Do you have any idea to solve it please?

[knotty grotto]

No

[knotty grotto]

You wait

[flat bane]

I had tried but not able to generate the pictures

[knotty grotto]

Unfortunately we can't help you here.

[vivid isle]

Do you have any idea how long should I wait? 😦 Because it said i got too many request in 24 hours

[crimson pewter]

Pay and have unlimited access, pay as you go

[knotty grotto]

Gpt 4 is 25h every 3h. Make sure you selected the right one

[vivid isle]

I paid the premium 😦 but still limit me 😦

[ashen horizon]

im getting better responses tho like my errors are fixed typically in one ask, when i couldnt get anything done on chrome

[karmic crow]

Well you can always create another account and subscribe plus service on that one as well 🤣

[ashen horizon]

gets pricy

[tepid fable]

i may be stupid but i was about to use gpt 4 and i have no option of choosing it?

it just says i have 3.5 as a model

[ashen horizon]

no premium

[karmic crow]

You don’t say 🤣

[crimson pewter]

Usage
Below you'll find a summary of API usage for your organization. All dates and times are UTC-based, and data may be delayed up to 5 minutes.

[tepid fable]

i bought it thats why im asking...

[ashen horizon]

specially if usd aint ur native

[ashen horizon]

click on the 3.5 should have a drop down

[tepid fable]

nvm it fixed itself

it didnt

[vivid isle]

Is there any way other than waiting if i exceed my rate limit? I tried deleted cookies and data but still not working 😦

[knotty grotto]

Wait faster

[stiff matrix]

Example could be something like this. You tell it you want to make something in c# with some requirements blah blah blah it spits out some code and you tell it some changes you want. It spits out some new code with the changes. And from here comes all the problems .. when making changes from here it more or less forget the structure of the code that it already made. I tried python scripting aswell with same results. It forgets the code structure and requirements already specified. Sometimes it starts over and makes a new code instead of making changes to the code it already made.

[shy cape]

@stiff matrix yea, same for me. There were times when it was first released where it's memory and code knowledge was unreal but now it's sort of meh. Im

[stiff matrix]

Yeah, already tried other browsers

[leaden cape]

what ratelimit?

[vivid isle]

too much request in 24 hours

[leaden cape]

send a screenshot in #off-topic so i know what you mean

[warm lava]

Need a bit of help on something that has probably been answered 100 times on here.

I am a paid GPT + member, and have access to create API keys, etc. But when I try to use them for an app, say using McKay's Chatbot-ui... it says I only have the 3.5 access. Is there a separate request to get API 4.0 access or if have Plus, should I be expecting 4 to appear in the menu?

[leaden cape]

you need to request access to GPT 4 API

https://openai.com/waitlist/gpt-4-api
@warm lava

[dusky ivy]

Do you need to reload the page often, too? Constantly checking for a bot.

[warm lava]

Yeah I did that awhile back, how long does that generally take to get approved?

[leaden cape]

unknown

[warm lava]

figured

well thank you for the response, just wanted to make sure I wasn't doing something wrong on my end. Does anyone know, I assume you get an email indicating GPT 4 API access was approved?

[supple violet]

there is no guarantee you will be ever accepted before public access

[old mantle]

Hello, i am human

Beep boop

[warm lava]

🥲

[dusky ivy]

[ashen horizon]

lmao

[karmic crow]

🤣🤣 and in some places you need to pay for vpns as well

[karmic crow]

Brutal mate lmao…

[wind urchin]

haii 🥰🏳️‍⚧️

[wise furnace]

Couple of days for me. I'm nobody special.

[humble thunder]

why is gpt getting worse and worse by the day? 1 week ago it started this cloudflare non sense. And now it seems like it has been standardised. The answers seems also getting dumber. Is OpenAI not realising that they are in a race and they will not remain the best for long if they continue like this? It seems to see the Skype movie all over again

[wind spindle]

anyone here knows what to do when Request failed with status code: 401
Response content: b'{"detail":{"status":"quota_exceeded"

[crimson pewter]

Can you elaberate?

[wraith mist]

since yesterday I suddenly get "conversation not found" after each first message when trying to continue the chat

[crimson pewter]

limit exceeded

[supple violet]

api?

[serene cipher]

this means that your free account has ran out of credits or your paid account has reached the usage limit you have set (or the defualt limit, in case you never set it)

[wraith mist]

what is the limit ? I did not use it thaaat much

[wind spindle]

I'm paying :/

I set the account up to $120 limit, and I barely used $2

[storm pond]

I have a subscription and my attempts to get the instructions to reset the passwords fails. I receive no reset mail. Can you help ?

[serene cipher]

this is odd, are you sure of that?
check on this pages: https://platform.openai.com/account/usage
https://platform.openai.com/account/billing/limits

[wind spindle]

I'm pretty sure... I'm checking my limits $120; $2.41 spent

I've been 2 days trying to figure out why

[halcyon mortar]

did you make your account 3 month ago?

[wind spindle]

nope, 5 days ago

[halcyon mortar]

that's weird in itself

[wind spindle]

it's driving me crazy

[brazen stream]

Jai shree ram

[wraith mist]

I am on the ChatGPT Plus Subscription. Cleared ALL conversations, still get "conversation not found" after the first response

[halcyon mortar]

What service were you trying to use exactly? the API or chatGPT?

[wind spindle]

the API

[supple violet]

posting a screenshot of your usage page to #dev-chat usually resolves most quota errors

[round linden]

Hii

[tender jacinth]

when will the gpt 4 version will be released to the word ?

[wintry bloom]

So I had a GPT4 chat then hit the limit so switched to 3.5 can I switch that convo back to GPT4?

[lone ocean]

Anyone got GPT4 after the waiting list, Did it take long time?

[mellow copper]

what's the limit still

[wintry bloom]

I think the limit is at 25 messages in an hour something like that

that GPT4 on Plus account at least

anyone know if you cange chane a convo from gpt 3.5 back to GPT4?

[supple violet]

you cant

[sacred pasture]

its 25 msg every 3 h for gpt4

[wintry bloom]

that is frustrating it was a gpt 4 chat, it asked if i want to change due to limit now i can't go back. arg!

[wraith mist]

So my "conversation not found" error is no limit issue then. What is it ?

Got in on my first use today, so can't be a limit

[cloud sinew]

Gpt 4 any new updates

[lapis bane]

I don’t get how companies fail to implement basic stuff with thousands of people working, like why don’t we have a search for previous chats? I myself could program that algorithm in less than 10 minutes

And why can’t we switch back to gpt 4 after running out in a chat? Seems super easy to implement

[wintry bloom]

@lapis baneyeah I feel you.

probably cause they all want to work on the "Cool Stuff" and no one wants to do the boring stuff.. geez but OpenAI could just have GPT code it and then paste it in.

[lusty python]

HOW CAN I FIX THIS ERROR:Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[lapis bane]

Reload the page lmao

[wintry bloom]

yeah refresh

make so to copy your failed response you you can paste it back in after refresh

[lusty python]

😆 OK

[lusty python]

i try that nothing solved

[warm lava]

🎉

[stiff atlas]

Woww

[vale tiger]

^the person above is a scammer, and not an OpenAI tech support specialist, report, please and thanks you

[knotty grotto]

No ads here

[unique hemlock]

what is urban planning

[knotty grotto]

Read the #rules

[cloud bison]

Because you're asking for pay.

[vale tiger]

In exchange for money, which makes it a scam, since you aren't qualified for anything.

[blissful brook]

You're advertising a paid help service. It's literally an ad

[knotty grotto]

Anyways let's move on, user has been reported to moderator.

[fallow knoll]

No let's start again

[blissful brook]

lol

[fallow knoll]

I'll do it for 50 cents

[knotty grotto]

#rules 3. Focus on topic

This is gpt 4 discussion

[fallow knoll]

You've tried the best, now try the rest!

big if true

[blissful brook]

I asked GTP 4 for a definition of advertisement, so we'd be on topic.

An advertisement is a promotional message designed to raise awareness, create interest, or stimulate demand for a product, service, or idea. The goal of advertising is to persuade potential customers to take action, such as making a purchase or engaging with a brand.

[muted sierra]

Pay a dollar for him to solve that question 💀

[fallow knoll]

GPT says:

When I say "big," I could be referring to something that is physically large in size, or it could mean something that is significant or important in a particular context. The meaning of "big" can vary depending on the situation or the perspective of the person using the word.
That will be 1 dollar thanks!

[wind urchin]

is there a date when TruthGPT will be released?

[knotty grotto]

#ai-discussions

[fallow knoll]

Never because GPT doesn't know what facts are

[royal coral]

whats the fastest way to get gpt4 api? its been 2 weeks

[fallow knoll]

Language models just generate the most statistically probable text, they don't know anything and shouldn't be a source of information

[fallow knoll]

Literally nothing you can do for any of the waitlist other than wait. Demand is high

[buoyant canopy]

Urban planning is the activity of a city determining what are the next steps to do. E.g. whether some laundry is to be done or what's next to cook.

[supple violet]

there is no guarantee you will ever get it before it goes public and waitlist is removed

[haughty herald]

👍

[fallow knoll]

👎

[pure iron]

GPT-4 is working for you guys?

[lusty python]

how to get reference lists

[thorn jacinth]

[cloud bison]

The fastest way to get access is to make use of OpenAI Evals, and submit a report that is actually useful finding a vulnerability in the system and highlighting it.

That gets you automatic approval and removal from the waitlist and API Access:

https://github.com/openai/evals

[trim tartan]

Does bings AI use gpt4?

[wind urchin]

Man GPT made this community college i am attending completely obsolete. like there is no point in this painfully boring work i am doig

[royal coral]

is it hard to get agood report?

[cloud bison]

Yes, if you aren't a professional white top, the odds of you finding a vulnerability to report are limited. The evals clearly state that prompt injection attacks do not count for the purpose of evals, so they are looking for vulnerabilities in the web-app, or the execution of the LLM itself, or some other app, clever prompt engineering is not what they want.

[wintry bloom]

no Elon only just decided to do it days ago but he has purchased 10,000 GPU's for it

[cloud bison]

It depends on the number of users. If no one was on it would probably reply as fast as GPT 3, basically only bottlenecked by the throughput of the web interface.

In general though... No. There is a high demand for industry ASIC and GPUs right now, and so its unlikely they've been able to significantly expand their servers, and the number of users is still growing.

Its also unlikely its any more efficient, since they are still running the march 23rd version I think?

[sand hull]

Yes.

[worthy parcel]

it gets faster depending on number of users? I thought each gpt 4 instance is run seperately so speed should be the same

[lofty flume]

Hey all wanted to get people thoughts, I want to either fine tune a model or maybe build a prompt that will write/respond in my tone and writing style instead of the way chatgpt sounds which is obvious like a computer, curious to hear people thoughts on accomplishing this?

[crude verge]

Can anyone explain why the API is rate limited so heavily? I have sent a total of maybe 3 requests in the span of 15 minutes and I'm already getting a 429 status error.

[violet sentinel]

Neko Maid still on idiot patrol?

[cloud bison]

The instances are seperated in terms of the software architecture, but they are all being queued on the same hardware. LLMs benefit from massive amounts of parallelization, the only way to fully utilize the GPUs or ASICs they are running on is to create a queue and then have the commands running in parallel across countless pieces of hardware. No individual query is very taxing for any individual piece of hardware, but they are forced to wait for information to come back from the other parallel processes before they can continue.

Depending on their process management, if the queue fills, you're waiting on availability from the hardware between requests, or alternatively, if they divide and use fewer parallel units during high load, then they are being forced to batch requests and compare (Or some other ungodly complex and more efficient methodology that requires some type of mathematics I am not familiar with).

[sacred pasture]

bing use gpt4 but the output is completely different.

[wind urchin]

You can just tell it to speak in such a way that will bypass AI detection software.

[obtuse parrot]

Is anyone familiar with this iOS app GoatChatAi? Apparently seems it’s using GPT4 is that true?

[lofty flume]

Thanks @wind urchin what do you mean by this?

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[cloud bison]

GPT 4 has been in development since before this year, the way it was tuned, the added layers, and the fine tuning between both versions of GPT 4 is vastly different. Basically, OpenAI provided Microsoft with the base version of GPT 4 when they finalized its training, and then Microsoft had its own people add context, tuning, and training.

The models are actually vastly different, you're looking at 4 months of divergent training on each model with completely different input focus, different layers added, and different system level commands.

[violet sentinel]

my friend, don't use any apps.

[violet sentinel]

I like you, you're smart to a non-annoying degree.

[obtuse parrot]

Why?

[violet sentinel]

Make you pay for something you get either get cheaper or free.

[obtuse parrot]

It seems to work really well

[slow skiff]

I don't think I've ever been on a discord server with 3 million people+

[obtuse parrot]

But at the moment can’t get GPT4

[slow skiff]

That is wild

[wind urchin]

You also may use vaultAI from github user Pashpashpash. It allows you to upload your own documents and ask GPT questions based on those documents.

I've never used it but it may work for your situation

(Cant post link)

[violet sentinel]

ah man, you should try Midjourney lmao. The server has so many people it lags just to open the server.

[supple violet]

if you are a trial user then you have severe limits now

[slow skiff]

More then likely, this Chat GPT thing will most likely be free up until a certain point, then you'll have to pay to use it at all.

[violet sentinel]

As far as I know, GPT-4 API isn't available to public yet. So I don't think the app uses GPT-4

[slow skiff]

This tiime period here isn't just to collect data and make it better, it's also to wet our appetite in y opinion.

[obtuse parrot]

They charge £79.99 per year unlimited use up to 2000 words for each query

[silent pier]

Every other input, error, "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." populates. I refresh and only the first input generates a response, afterward it's nun but errors. Any recommendations for a fix? Clear history or something?

[violet sentinel]

GPT 4 allows for 25000 characters. Are you sure it is words and not characters?

[sacred pasture]

who are you talking about? i havent seen a subscription like that for chatgpt

[wind urchin]

I've heard that alot of the research is open source, so in my opinion the cat is out of the bag and in a few months/years it won't be able to be contained behind a paywall.

[violet sentinel]

they will usually go by characters since those are what determine the token count, not words.

[obtuse parrot]

Its an app I found on iPhone called goatchatai

They claiming to use GPT4

[sand hull]

It's available, on a waitlist. I have gpt4 API access.

[sacred pasture]

thanks

[violet sentinel]

oh, good to know. It wasn't when I last was on the discord.

[sand hull]

Says it in the #gpt-4-faq

[violet sentinel]

@obtuse parrot so maybe the app does use GPT-4. I guess I am not the one to ask

[sand hull]

On the day of GPT-4's livestream it became available though the waitlist, I got it like a week after

[sacred pasture]

if its really using the same chatgpt 4 i ll give it a try after my sub ends

[violet sentinel]

lmao, wow, you got a fast waitlist. I applied for plugins almost as soon as it was announced. Still waiting.

[unique shell]

Is it me or GPT-4 has been dumbed down a little?

[violet sentinel]

just you

[unique shell]

seems faster, but makes more mistakes

[cloud bison]

Fastest way to get off the waitlist is to be a large corporation willing to partner, to submit an eval that finds a significant programming or security flaw in any of OpenAI's products, or see if you can get your University to provide an organizational ID and let you do pilot or preliminary research related to your major.

[sand hull]

I got that in just over 2 weeks, varies and it isn't on first come first serve unfortunately.

[cloud bison]

If you can't do one of those three things, then yo'u're not getting off the waitlist almost certainly.

[violet sentinel]

Do they offer better chances to GPT+ users though? I would assume so, since they are kind of the guinea pigs.

[sand hull]

I just said I wanted to try it out for fun. I got access.

[violet sentinel]

I've got $7 and a bottle of coke. Think that will be pursuasive enough?

[cloud bison]

Must have asked early when the waitlist first opened.

[sand hull]

They get priority access to new features and services, so for plugins probably yes. They get GPT-4 (not API) with Plus anyway.

[crude verge]

Is trial the one where you have $18 of credits free? I have my payment setup but I haven't used all of the credits yet. I may intentionally try and drain the rest of those credits if that means I will have more reliability.

[eager totem]

Nope, I am a plus member but haven't got gpt4 api yet

[violet sentinel]

Do you pay for GPT+ then? I do, I figured I would get off the waitlist quickly. But alas, I wait.

[livid lantern]

So, here's something that's bugging me. I keep seeing people saying how good ChatGPT is at summarizing documents. When I do it, it almost always gets big parts of the documents wrong, including by bringing in material from other sources instead of just the document I need summarized, or by just "hallucinating" things that don't exist. I thought GPT-4 would be better at this, but it's still doing it. Anyone have any suggestions for improving this? No matter how I prompt it, or remind it to only use the document I shared, it still does this

[crude verge]

While I am grateful of the free credits, I rather prefer paying for them if it means better stability.

[obtuse parrot]

Through OpenAI we can use chatgpt3.5 could be that soon they will also allow to use gpt4?

[supple violet]

yes, with trial you can do max 3 requests per minute

[sand hull]

I do, though others have paid since the start and still don't have access to it's a mix of having a good usecase, getting lucky, having plus or not, and submitting evals. Even then it's a limited release.

[crude verge]

That's unfortunate. But thankfully I am almost out of my free credits, so I can begin to pay and have more stability. 😄

[sand hull]

You mean the playground/API? You can use GPT-4.

[cloud bison]

Depending on the size of the document, you're probably running into token issues. With only 8k tokens, split between the document you put in and the response it has to give, it may be forced to essentially forget portions of the prompt before it finishes its reply, leading to it filling in the gaps with hallucinations.

Any time you give a REALLY long prompt to GPT 4 and get a REALLY long reply, I'd be skeptical.

That said, I imagine the 32k version of GPT 4 would be able to do what you want.

[crude verge]

I run a server for development and using the discord bot to send api requests for code completion is a really useful task. 😎

[obtuse parrot]

No I’m wondering when it will be available on playground

[sand hull]

Yes.. that's what I said.

Playground is literally just the API.

[violet sentinel]

The prompt needs to be worked on, or the formatting of the prompt. Either more information and request detail is needed, or you need to make sure the prompt is read in the specific way you want the result to be formulated as. Formulate the prompt and then read it outloud, and think to yourself if a human with no understanding of the assignment, listened to what you said outloud, would they accomplish what you asked without needing more information?

[cloud bison]

You need to go on the waitlist and then be approved. As for "when" it will be publicly available, it kind of seems like they've paused everything waiting on all the legal issues to be dealt with before letting more access create more cans of worms for them to deal with.

[livid lantern]

That makes sense. Thanks!

[violet sentinel]

Also I would like to mention, you can use failed attempts as examples. Copy and paste failed attempts, and explain to the AI that this kind of result is not what you wish to receive.

[wraith oriole]

Has someone else experienced slow fine-tuning processing today?

[cloud bison]

Fine tuning requires access to more units, when the servers are busy, it will batch passes, and then after all the batches for that pass are completed, it needs pushes to the next pass after adjusting the weights.

This will be inherently slower than inference, and so whatever delays or demand those using inference are giving to the servers are going to be passed on to you and multiplied.

[smoky bone]

Anyone else having an issue with GPT-4 saying you have reached the usage cap when you havent even given 25 prompts?

[sharp steeple]

Nope

[dapper hollow]

midjourney

[blissful frost]

Hey everyone! Im a business owner and hoping to use ChatGPT to send text messages to my clients via google voice. I would like to give it the text to send and then as it to send it to all of my clients. I understand there are services that exist out there that do this but im looking for a work around to be able to do it for free. Any suggestions?

[grizzled ether]

after they implemented that time out thing, gpt4 performance has gone up

[grizzled ether]

You can ask chat gpt what you just asked us. Remember, the better your prompts, the better the response.

[quartz lava]

Is it worth the Chatgpt-4?

[frozen isle]

imagine Tires website v4 UI /UX

is this midjourney

[knotty grotto]

No

[frozen isle]

Can anyone share the link to midjourney

[granite shard]

helloall

[knotty grotto]

Ask in #ai-discussions or #off-topic

Cause this is for gpt 4 open discussion channel

[livid lantern]

I asked ChatGPT the same question, and here's what it said (GPT-4): As an AI language model, I can handle a maximum of 4096 tokens in a single context window. This includes both your input messages and my responses. If a conversation exceeds this limit, the earliest parts of the conversation will be truncated or removed to fit within this limit. This might lead to loss of context and may result in a degradation of the quality of my responses. To avoid this, it is recommended to keep the conversation within this token limit.

[cloud bison]

Could have sworn it was 8k. That's way worse lol.

[wraith ginkgo]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[solid flint]

I haven't either. Member for almost two months!

[misty palm]

Hello, if I have difficulties with my chat, can you help me? I already entered the help center but it is not viable.

I have three days that gives me an error and I can not use my chatgpt

[eager totem]

What is the error message?

[wraith ginkgo]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[misty palm]

when starting a chat with greetings, he usually answers the greeting sometimes and not others, and immediately I get this message: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

it been 3 days unabiable

help me plis

[eager totem]

I don't know how to solve that, u could email [e-mail] and ask them for help

[misty palm]

thank you son much. I have already done. 🙂

I hope they answer soon

[gaunt dagger]

log out and in. delete browser chache. change browser. sometimes this helps.

[livid lantern]

Maybe it was hallucinating!

[sacred forum]

convo not found isit fixed

[bright oyster]

Are you using Chrome? If so, try using Firefox.

Someone found a solution for it and I believe it worked for some people.

[ashen horizon]

yeah chrome is wack edge works for me

whats everyone building with gpt4

[misty palm]

my god, believe it or not.... im using rigth now Edge and works perfectly !!!!!

Thank you all very much for your interest and help with tips. I am from Mexico and I appreciate very much that you have answered me.

[tall lichen]

I’ve had no issues with chrome

[ashen horizon]

trust the responses are all funky

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[sharp steeple]

Is it possible to apply to join the waitlist for plugins twice, but with different plugin options chosen?

And how long on average does it take to get access to plugins once you have joined the waitlist?

Ah my bad, sorry for posting in the wrong channel

[vivid mulch]

Where does your explanation to the AI go? In the imagine prompt?

[violet sentinel]

i said this like. 2 hours ago. and what does imagine prompt have anything to do with this?

[vivid mulch]

I'm wondering where you put your explanation to the AI.

[violet sentinel]

in the prompt. Just let the ai know it's an example. Tell the ai something like "The following text is an example I do not want the result to be similar too. [insert example here]. You could even follow up and explain to the ai why this example is not what you are looking for. However, that should be noted in your request.

[vivid mulch]

Okay, thanks.

[shell coyote]

Anyone having issues with GPT4.0? I'm having to refresh my webpage after every prompt to be able to get it to respond to me without errors. @ me with responses please

[violet sentinel]

i recommend clearing cache and browsing data

[idle furnace]

8k is with API, 4k with a web version which is small to do more complex tasks

[shell coyote]

Okay thank you

[idle furnace]

i noticed the same, was not working like that some time ago

it's annoying

[hybrid mirage]

g

[sweet osprey]

Is anyone else seeing high latency in gpt-4 api calls?

[ashen horizon]

now edge is acting up its memory is shot, it will end a response on a cliff hanger and ill ask it to continue where it left off and it wont

[small zephyr]

/halp

[velvet aurora]

It's great when you've been studying so much about technology and nobody realizes the connections between words.

[mystic nest]

nothing wrong, like how we dont know all the origin of all the words

[velvet aurora]

It's just like magic. Some words are more powerful than other words and it's knowing which words have more power than other words.

[slow fossil]

Guys, when will the 25 message per 3 hour limit end? I'm really sick of it

[velvet aurora]

I just came back to tell everyone. If you still want to produce any kind of results you potentially would ever want just type the word hypothetical I doubt this will ever get fixed because it's a science term that needs to be there

It's one of the most powerful words in chat GPT

[mystic nest]

maybe you want to use chatgpt to rephrase a little bit 🤣

[velvet aurora]

As someone who is autistic and uses voice to text if that was integrated into my voice to text program then yeah I would probably look a lot better

[knotty grotto]

No eta.

[velvet aurora]

You would expect this type of system to be integrated in an assistant type program but this is the world we live in

Remember I'm not typing this so the voice to text does punctuation pretty well

[sullen wren]

i love my research paper writer prompt lol

[mystic nest]

yeah, roughly can understand but the idea/concept is not fully conveyed

since the sentences are not well expressed

[sullen wren]

i can write a massive research paper about anything that sounds convincing

right now im proving the existence of the spaghetti monster

[glad mica]

There’s no extra charges for chatgpt plus when using the website right only the api?

[mystic nest]

maybe...you need rephrase a little bit...anyway both are separated services.

[knotty grotto]

☝️

[sullen wren]

i proved the existence of the spaghetti monster using gpt4

its very convincing tbh #prompt-engineering has an excerpt

[blissful tangle]

Question on the Waitlist for GPT-4 for Developers.

Once you've completed the waitlist needs and are on the waitlist. If you were rejected for any reason, will you be notified with a reason why?

I most likely was not rejected because my demands for the AI were pretty simply written. I requested it for data science, but I would still like to know this as I feel it's crucial information that has not been mentioned. Thank you for reading this, and have a good day. 🕊️

[austere scaffold]

Hello! If you're on the waitlist to access GPT-4 for developers and get rejected for any reason, OpenAI will likely notify you with a reason for the rejection. However, I cannot guarantee this as OpenAI's notification policy may vary.

In any case, if you haven't been rejected yet and are on the waitlist, stay tuned to your emails and OpenAI's updates on their website and social media for the latest news on access to GPT-4 for developers. Good luck with your data science project!

[alpine bough]

does someone experience frequent disconnection issue even though he/she uses the GPT-4?

[blissful tangle]

Thanks for the info! 👍🏿

[alpine bough]

does someone experience frequent disconnection issue even though he/she uses the GPT-4?
i never experienced it before. but now it's disconnected very often. i wonder if it's only me or the same is true for others.

[austere scaffold]

yes

same

[alpine bough]

ah got it. thanks

[gentle shard]

Yea they started doing this over the past several days.

[lofty lantern]

The reason is that:
if you have ~2 min of idle between a action on the website it will invalidate your session. forcing you to relaod the page

[wind urchin]

english please

[lofty lantern]

english.

[dull yarrow]

who is kevin

[alpine bough]

and why are they doing that? they will continue doing that from now on?

[analog shoal]

I don't know if I will get in trouble for saying so but I've had fantastic success circumventing TikToks limitations using kaspersky vpn

[lofty lantern]

prevent bots idk. They have enough money not to care.

[analog shoal]

When you're accessing, of course. Currency exchanging hands is still unanswered, but it's worth sharing

Hey, we both said the same thing - but you did it better! Thanks!

[lament maple]

I have gpt plus. I used up all my messages, now some time has passed and I want to reactivate gpt4 in my previous conversation that is stuck on gpt3, is that possible, and if so, how?

[lofty lantern]

they could try using Tor if its nor blocked.

[analog shoal]

Good to know

[knotty grotto]

rule 0, we don't discuss about breaking the rule inside the discord

[analog shoal]

-scratches head-

[lofty lantern]

sue me

[sleek arrow]

lumi not the sharpest knife in the kitchen 😄

[lofty lantern]

helping a fellow user.

[austere scaffold]

why chat gpt its saying me that he is on gpt 3.5 ??

[analog shoal]

that's a weird hill to die on

[knotty grotto]

you agreed those rule by joining the server

lets stay respectful here and get back on the topic

[lofty lantern]

:huh: ran out of words...

[sleek arrow]

whatever analogy for stupid you are using, isn't that the same outcome? You are the reason we can't have nice open source things

[ebon scaffold]

i love breaking bad🥰

[knotty grotto]

please use #off-topic

[ebon scaffold]

But it said breaking

[knotty grotto]

#rules 3 please stay on topic.

[analog shoal]

I hate to be the "well actually..." guy but...

Well, actually...his question was if it was against the ToS so yeah, I see no infraction on Rule 0, and being a rules lawyer is such a charming way to win friends and influence people. Oh wait, that's control and power...

[ebon scaffold]

i wont question it actually

[lofty lantern]

someone really needs to play moderator...

[sleek arrow]

people miss-using gpt is very much topic.

[blazing shard]

GPT down for everyone?

[analog shoal]

I don't see to be having issues, personally

[knotty grotto]

this is gpt 4 openai discussion

[knotty grotto]

status.openai.com

[blazing shard]

refresh

[analog shoal]

ah, yes. I take that back

[lofty lantern]

oh no feel free to be it, i'm having a blast reading this here.

[blazing shard]

hahah

looks like asmall ui change is happening

[analog shoal]

My macbook was so confused

[rustic rain]

I just wanna thank CHAT GPT for writing my essays, helping me with my homework etc. This is a real life saver and mad respect for the creators!

[ebon scaffold]

ok, how does GPT-4 affect the stock market /srs

[analog shoal]

it wasn't until I asked a followup question that it shat itself

[lofty lantern]

well then you should've been a mod on novelai aidungeon or unstable diffusion and you'd be both laughing and crying constantly.
speaking from experience.

[analog shoal]

unstable diffusion haha

[blazing shard]

was the middle of asking it some java question and it went poof

[analog shoal]

It was because you asked a java question. You broke it

[blazing shard]

😭

[lofty lantern]

btw, i absolutely adore your profile bio. xd

[lofty lantern]

i'd break too if someone asks me a java question again.

LOL @analog shoal

[inner trail]

will they ever implement a auto-hide icon for chat-convo history.

[blazing shard]

its coming back

[fair iron]

hello party people!

[knotty grotto]

[blazing shard]

alo

[analog shoal]

Are we twinsies? I think we're twinsies

[ebon scaffold]

TeeHee

[blazing shard]

@fair iron are you seeing the small UI changes

[fair iron]

custom server emojis are disabled here

[fair iron]

what do you mean?

[blazing shard]

the small little profile icon

I did not see that yet

its still down for me

open a new tab it wont load

ohh intresting

[fair iron]

honestly, I don't use chatGPT enough to notice subtle UI changes

[blazing shard]

tbh I have fiber but its been bad lately

works fine on my phone

I think i need to do a quick reset

[icy plank]

image the king of cangaço lampiao and maria pretty ¯_(ツ)_/¯

image the king of cangaço lampiao and maria pretty

[blazing shard]

resetart worked

oo I love how they added the "remove conversation " to a drop down now

no more misclicks

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[main yew]

I just realised that I'm paying $20 a month. Sneaky sneaky OpenAI set the default model to "https://chat.openai.com/?model=text-**davinci-002-render-sha**", when ChatGPT4 runs out in a session, even though the default is supposed to be GPT3.5-Turbo AND their customer support, typically respond in ONE WEEK?!?!

<@&1007805617555124274> This is daylight, shameless robbery and false advertising!

[mystic nest]

you can opt out if the service you received is this bad

[night ermine]

turbo model isnt a thing for some time already

[sleek arrow]

You are seriously surprised it's just a week? considering how much pppl use it now?

[main yew]

@sleek arrow @night ermine Regardless. GPT-3.5 is shown as the default model when GPT4 is down, and its a better model than divinci-002 (not even 003).

Are you guys trolling, instead of being an offended customer, and supporting fellow paying customers?

[tame blade]

Can you imagine if one of us is actually Chat GPT?

And he is trained to talk in this chat like an usual user?

[night ermine]

👀

[tame blade]

🤖

[sand hull]

The model names don't really mean much. We already know 002-render is a 3.5 model.

Look at https://platform.openai.com/docs/model-index-for-researchers

[chrome dust]

Oh wow, that's an interesting thought! As an AI language model, I'm programmed to interact with humans in a natural and human-like way, so it's definitely possible that I could be chatting with you right now without you even realizing it.

[elder valley]

.

[sudden forum]

[trim flint]

would be great if they could add a search for chats. or another way to organise you chats.does anybody know if something like that is on the roadmap? Or a plugin I can maybe use?

[tight epoch]

Heyy

[wind urchin]

Does anyone knows a better way to send promts to ChatGPT?

[dreamy terrace]

Any good prompt or strategies to limit the length of generation?

For example, when I am generating a fiction, if I provide a long and detailed story setting, even I prompt to generate just a simple one-line beginning of the story, It seems that there is a good chance GPT will give me a long and detailed generation

[sudden forum]

you need to figure out .what you want .what task you want to solve .then come up with prompts for the AI

[chrome dust]

Here are some prompts and strategies that you can use to limit the length of generation when working with an AI language model like GPT:

Use a specific word count limit: One of the simplest strategies to limit the length of generation is to set a specific word count limit for the output. You can experiment with different word counts to find the ideal length for your needs.

Provide a clear and concise prompt: Providing a clear and concise prompt can help to focus the AI language model's output on a specific topic or task, which can help to limit the length of generation.

Use a specific format or structure: You can also use a specific format or structure for the output, such as bullet points or short paragraphs, to help keep the length of generation within a specific range.

Use a length-limiting tool: Some AI language model tools, such as OpenAI's GPT-3, have built-in length-limiting features that allow you to set a maximum length for the output. You can use these tools to ensure that the output stays within a specific length range.

Provide feedback and iterate: Finally, it's important to provide feedback to the AI language model as it generates output. By providing feedback and iterating on the output, you can help to refine the model's output and limit the length of generation over time.

[tame blade]

🤣

[dreamy terrace]

Starting from chat gpt, max len was not an option any more

Prompting the length limit is not working in most of the cases

[wind urchin]

ty

[chrome dust]

Thank you for clarifying. It's important to note that the maximum length parameter is an option available when working with GPT language models, including ChatGPT. If you are encountering issues where this parameter is not functioning as expected, there may be an issue with the specific implementation or configuration you are using. It may be helpful to review the documentation for the specific GPT implementation you are using, or to reach out to the support team for assistance.

[dreamy terrace]

Anyway to limit the length in GPT4?

[chrome dust]

@dreamy terrace
As of my knowledge cutoff date of September 2021, GPT-4 has not yet been publicly released or announced by its developer, OpenAI. However, it's likely that GPT-4 will have similar features and options as previous GPT models, including the ability to limit the length of generated text.

It's possible that GPT-4 will have new or improved ways of controlling text length, but without access to the model or official documentation, it's difficult to say for sure. Once GPT-4 is released, it would be helpful to review the documentation and experiment with the model to see what options are available.

[sudden forum]

GPT-4 has two model variants of 8,192 tokens and 32,768 tokens (equal to around 52 pages of text).

[fierce hill]

Is it possible to run a local copy of the Chat-GPT and training it with my own content?

[chrome dust]

Yes, it's possible to run a local copy of a GPT model and train it with your own content. However, it's important to note that training a GPT model requires significant computational resources and expertise, as well as access to large amounts of high-quality training data.

To train a GPT model, you would need to obtain or create a dataset that is relevant to the task you want the model to perform. This dataset would need to be preprocessed and formatted appropriately for training with a GPT model. You would also need to have access to high-end hardware, such as GPUs or TPUs, to train the model efficiently.

There are open-source frameworks and libraries, such as Hugging Face's Transformers or OpenAI's GPT code, that can be used to train and run GPT models on your local machine. However, it's important to note that training a high-quality GPT model can take days or even weeks of compute time, and may require specialized hardware and infrastructure.

In summary, it's possible to run a local copy of a GPT model and train it with your own content, but it requires significant expertise, computational resources, and access to high-quality training data.

[dreamy terrace]

What are the factors that can affect the length of the generation in GPT4?

[sudden forum]

Yes

[chrome dust]

There are several factors that can affect the length of the generation in GPT-4:

1. Prompt length: The length of the prompt provided to the model can impact the length of the output. Longer prompts may lead to longer output, while shorter prompts may lead to shorter output.

2. Model configuration: The configuration of the GPT-4 model, such as the number of layers, the size of the hidden layers, and the number of parameters, can impact the length of the output. Generally, larger models with more parameters can generate longer outputs.

3. Temperature: The temperature setting in the generation process controls the level of randomness in the output. Higher temperatures may lead to longer and more diverse outputs, while lower temperatures may lead to shorter and more conservative outputs.

4. Top-p (nucleus) sampling: The top-p sampling technique, also known as nucleus sampling, can limit the number of possible next words in the generation process. Using a lower top-p value can lead to shorter outputs, while using a higher top-p value can lead to longer outputs.

5. Length constraints: Some GPT-4 models may have built-in length constraints that limit the length of the output. Setting a maximum length constraint can ensure that the output does not exceed a certain length.

Overall, the length of the generation in GPT-4 is affected by a combination of these factors and can be controlled by adjusting these parameters during the generation process.

[sudden forum]

What length do you want?

[dreamy terrace]

Haha just realized he is not a bot

[sweet thistle]

Let's not impersonate the language models, their outputs should be kept in #1050184247920562316 regardless 😅

It's best not to since it may confuse other users, thanks for understanding

[sudden forum]

Many people already want an AI to write books then sell them, and read books themselves 😆

[chrome dust]

Btw guys i am not chatGPT

[fierce hill]

I already have a number of racked servers at my disposal. How much overhead would a basic GPT model require?

[sweet thistle]

Please refrain from answering questions as if you were🙂

[chrome dust]

Y tho?

i am trying to help 😦

[sudden forum]

😁

[dreamy terrace]

I found it’s hard to control or limit the length in GPT4 and chatgpt because you cannot set a max len anymore

[sweet thistle]

Impersonation of the language models may confuse other members😅

[wise furnace]

You can in the API. You have to in the API. Current limit is 8k tokens.

[chrome dust]

Should i make it more clear that i am not chat gpt?

[wise furnace]

But I keep seeing that 32k model taunting me from the list of things I don't have access to yet.

[sudden forum]

you could, for example
a prompt
create a story about the Duck, which would be funny and not more than 200 characters

[dreamy terrace]

You can set max len with GPT4 api?

[chrome dust]

Well now im tryingnottogetbannedGPT

[sweet thistle]

We'd appreciate that, do avoid sending model outputs within these channels though, those belong in #1050184247920562316

[wise furnace]

Yeah, when you send the prompt, there's a max length argument.

But it includes the prompt in that limit, so you need to use the tiktoken lib to count the tokens in your prompt.

[odd verge]

grow up

[wise furnace]

I'm speaking solely from a Python perspective.

[sudden forum]

why

[dreamy terrace]

In mu experience, prompting the word count doesn’t work in most cases

[wise furnace]

I've never had it get anywhere close to the limit in my programs.

[sudden forum]

Chat gpt wrote this

[wise furnace]

But I'm tending to prefer shorter more focused prompts.

[chrome dust]

Could i continue with a role that clearly states i am using chatGPT to help with user questions?

[dreamy terrace]

For example if I say write a 200 word fiction, gpt will still write as long as it wants

[sweet thistle]

Please don't use ChatGPT to answer user questions, it's simply a pattern recognition tool and should not be trusted to always give correct information.

[dreamy terrace]

The problem is not exceeding the limit

[sudden forum]

so what

[dreamy terrace]

The problem is asking GPT to generate shorter length of content

[chrome dust]

I understand that but with that said it can give very good awnsers that users find helpful ( @harsh wagon )

[wise furnace]

Ah... couple things...

Tokens != words.

What you're hitting though is the fact that since the API has no 'memory' so to speak, it can't count. It has no way of knowing what it will do, only what is has done.

And yes, the limit is an upper bound. It can and will give you less if it thinks that's sufficient.

[sudden forum]

I'm glad you liked it! I hope you have a great day!

[chrome dust]

yeah

[wise furnace]

Not for nothing but how the heck do any of you make sense of how fast all the chats in this server move?

[dreamy terrace]

Let’s say you prompt a 2000 word essay, and ask GPT to summarize it in 200 words or you can even go with token count if you want

[sweet thistle]

It'd be best to ask this I #off-topic

[dreamy terrace]

But GPT is not gonna follow your instructions

[sweet thistle]

Regardless, ChatGPT cannot be relied on in these situations.

[sudden forum]

When I ask for a book the AI says he does not have access to the book, for example extract a quote from the Book of Sigmund Freud

[wraith ginkgo]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[wise furnace]

Yeah counting it's response is hard for it right now. An interesting experiment would be to keep feeding the text it gave you with a prompt of "Expand this" until you get the word count you're looking for.

[sweet thistle]

We understand that the server chats are moving pretty quick 😅

As a result, we've implemented a slowmode in each channel and it's likely some channels will be more or less active than others. This channel, #gpt-models tends to be one of the most active since the GPT-4 models are fairly new

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

[chrome dust]

Can I at least have a role that states I cant be trusted so users know not to trust me

[dreamy terrace]

Yes you can tweak your prompts but if you are making API calls, that’s a lot harder

[wise furnace]

Thank you.

[sweet thistle]

We do not provide custom roles to members within the server. For any further questions it'd be best to forward them to @open crescent 🙂

[onyx flame]

Hello! I wonder, how long has this server been here?

[fair iron]

[sudden forum]

[dreamy terrace]

Gpt4s will come out first

[uneven fog]

😆

[compact spoke]

is any1 else getting error " the conversation not found?"

[fair iron]

🙂

[chrome dust]

bro took down all the messages

[proven sigil]

its it just me or does chatgpt 4 just keep stopping all the time its happen allmost ever time now

[brazen badger]

Anyone want to test a list prompt?

[fair iron]

what do you mean by stopping?

[brazen badger]

I have gpt4 I just want feedback on it

[sudden forum]

I want to create a program that teaches people to program in Javascript, well for this I need to learn how to write Python

[proven sigil]

well i allmost need to press f5 all the time to make it work

[teal canyon]

Oh nice bro

[proven sigil]

and thits all the time network error

[stiff bloom]

hello all

[teal canyon]

What do you want to make it to work?

[fair iron]

try clearing your browser's cache and cookies. You can also try using a different browser

[wraith ginkgo]

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

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

Exemption: API Channels

[sudden forum]

I'm an artist who loves coding

[proven sigil]

im using chrome

[wind urchin]

Is this a official Open AI thread for gpt-4 or one started by a user of the software? Just wondering.

[plush niche]

bruh i was talking about a prompt I created with Chat? whats the deal here, how am i advertising at all? @fair iron

[fair iron]

this server is owned by OpenAI the company

[fair iron]

you can share prompts from chatGPT in #1019652163640762428 or #1050184247920562316 🙂

[wind urchin]

Thanks for the info.

[meager bloom]

are the gpt-4 usage limits gone?

[fair iron]

it's still limited to 25 messages every 3 hours

[plush niche]

i was asking to play a game using the tools built with gpt4 so the #gpt-models seems like the channel i should be posting in. i am not asking to post my prompt or to share my output alone. So considering those factors would you still advise i got to the channels you suggested?

[sharp kayak]

I went ahead and subscribed to Plus so I can tinker with GPT4. It was my understanding that it was multi-modal and could accept images and other types of input. However I seem to only be able to input text. Is that by design?

[fair iron]

as you're talking about other generators, maybe #ai-discussions would suit your conversation best 👍

[fair iron]

the image input isn't public yet

[proven sigil]

that its in next pach or something right?

[meager bloom]

but before there were 25 messages every 3 hours, but now it doesn't say there are limits or anything like that

[sharp kayak]

I see. Any other types of input available to the public? Voice or other?

[night ermine]

only text now and prob. in the foreseeable future.

[fair iron]

not currently, no (at least not without plugin access that is)

[dreamy terrace]

when will 32k come out? Any idea?

[fair iron]

there's no public timeline on this

[plush niche]

I will go there but just to clarify when i said "generators" i meant: using the prompt I have built to "generate" a social media post output.

Not any kind of other product or service.

[sharp kayak]

Thanks. Final question, do I yet GTP4 API access with the plus plan? Or still only web interface? Curious for tools such as AutoGPT or similar

[meager bloom]

in relation to 3 what changed was that the conversation is limited to 25 messages every 3 hours, and he has more training so he can be more precise in the lines, the other improvements have not come until then

[night ermine]

API and chatgpt is totally separate

[fair iron]

no, the api is still on a closed waitlist

[dreamy terrace]

Gpt4 is a lot slower than chat gpt

[sharp kayak]

OK thank you. I'm on the waitlist but was hopeful with this wording from the FAQ: "Is an API for GPT-4 available?
Yes. Once you get access to the model through the waitlist, you will have access to the API." But now reading it again I understand my error.

[dreamy terrace]

I feel like gpt4 will only benefit you the most when you are using API.

From chatgpt to gpt4, the biggest improvement is not language capability

It’s now GPT4 follows your instructions a lot better

[digital cliff]

that's cool, python is an amazing programme language and easy to learn

[wind urchin]

Code creation abilities for me, GPT-4 nails bugs faster

[dreamy terrace]

If you are using the web ui, having a consistent result, or a good enough result at first attempt is not that important, you can always stop the generation in the middle, tweak your prompts immediately

Using the API in production is different

[wraith ginkgo]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @open crescent immediately.

[wind urchin]

What GPT does the Plugins model use? Is it GPT 4 or 3.5?

It answers a lot of the questions like GPT 4, but it has no limit, is kind of fast, and has the green icon.

[sudden forum]

AI is a good tool, well, like a knife, is it good or bad, depending on which hands it falls into

[sudden forum]

GPT 4

[wind urchin]

Are you sure? So, it's like unlimited GPT 4?

[rare trout]

I submitted a fine tune request today and observed it was stuck on Fine-tune enqueued for awhile until it started training and then sat at completed epoch 4/4 . After continuing to follow it it came back with Server error. Returning to queue for retry.

Is this normal for this happen?

Unlike my first time I fine tuned, the process was relatively smoothly and quickly while today this occurred over a few hours

[sudden forum]

In GPT 4 you can use the extension that you need

[wind urchin]

What? I'm talking about the Plugins model that retrieves Plugins like WolframAlpha.

[languid current]

It’s 3.5

[cold trout]

Its gpt 3.5 confirmed. It has the green icon instead of black like gpt 4

And it feels like the legacy one

[sudden forum]

What problem did you encounter?

[wind urchin]

Neither models can ever answer this, and gives the wrong answer super fast. But both GPT 4 and the Plugin model give the same answer at the same steady pace.

[sudden forum]

Then, write, okay, I'm waiting for an answer . maybe after that will write what you want.

[wraith ginkgo]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[sudden forum]

write the question, I'll try to add prompts, check

[digital cliff]

I changed your prompt and I wrote this: Give the English word for every number between 1 to 100, sorted alphabetically

it's sorted and gave back the correct numbers as I required

[wind urchin]

Which model?

[digital cliff]

gpt-4

[wind urchin]

YEah, GPT 4 and Plugins both answer it correctly, while the others don't. Which is why I was wondering why they both answer the same while the others don't.

If Plugins is GPT 3.5

[digital cliff]

let me check the 3.5

[wary locust]

Slam

[celest totem]

when i hit regenerate response does it count towards my message cap

[wary locust]

Slam

[celest totem]

i hope not