chatgpt-discussions

[pine tulip]

every company is getting rich working with openai lol

[past fiber]

And why not its great company

[pine tulip]

Yea, but I feel bad for all the companies who were late to the game

[quaint pelican]

Nvidia sits at the AI throne at the end of the day

[past fiber]

I am glad openai was first

[pine tulip]

which we all know, is not real because it's AI

[quaint pelican]

Because they power up them all

[pine tulip]

I'm not necessarily

[quaint pelican]

Where there is AI, there is Nvidia in the back

[plucky timber]

Still got a nvidia xavier laying round

[quaint pelican]

They are even the ones that power up for example Adobe Firefly

Nvidia Picasso AI

[dense halo]

how do I use chat gpt 4 turbo?

[pine tulip]

just use chat gpt 4 really fast

or wait until you have it

[dense halo]

yeah but I wanna code plus it's up to date

[past fiber]

Well there is no competion

[pine tulip]

thats what bad about it

[past fiber]

Well

[pine tulip]

every field needs competition to grow

or to offer reasonable prices

[past fiber]

Ok yeah the price needs to go down

[zinc girder]

why dont i have the new features

i pay for plus

[vernal linden]

Hi, i've got a quick question, how can I acces to New ChatGPT interfaces, (I still have the old one)

[pine tulip]

it's rolling out, just wait

[sinful bobcat]

Anyone else get the link to the form about requesting higher cap usages of GPT4 for Plus users?

[normal quarry]

be patient, it's still rolling out

[zinc girder]

doing my best hahhaa

[past fiber]

But if you thing about it if it was cheap we wouldnt have the electricity to power it

Sorry for my english

[brisk path]

There is one?

[quaint pelican]

Exactly

[sinful bobcat]

Yep. I finally hit my cap and got a link to a form:

Interest Form: Higher Usage Cap for GPT-4
Thanks for your interest in GPT-4!

To give every Plus user a chance to try the model, we’re currently dynamically adjusting usage caps for GPT-4 as we learn more about demand and system performance.

We're also actively exploring ways for ChatGPT Plus subscribers to use GPT-4 in a less constrained manner; this may be in the form of a new subscription level for higher-level GPT-4 usage, or something else.

Please fill out this form if you'd like to stay posted.

[vernal linden]

Ok, but like it's gonna be available for everyone in the same time or is it sort by region ?

[normal quarry]

it's random

[dense halo]

Hey guys how do I access the gpt 4 turbo? I really wanna use it for coding purposrs

[pine tulip]

just read a few lines above

[vernal linden]

Ok, thanks !

[dense halo]

still rolling out?

[pine tulip]

yes

[dense halo]

damn I can't wait fr

[pine tulip]

can take a few days to weeks

[past fiber]

Well for me i would say insted of higher the limit i would love to have more tokens. I am ok with 50 per 3h

[normal quarry]

here -> https://platform.openai.com/playground?mode=chat&model=gpt-4-1106-preview

[brisk path]

I don't wanna pay any more than $20 😭

[glad ruin]

guys the 2 step authentication is not working in playground's login page

[past fiber]

If its better why not

[normal quarry]

it is. I just used it

[past fiber]

And it will be as a second option like a tier

[glad ruin]

I had to try 16 times when finally let me in

[normal quarry]

maybe you were mistyping the code?

[glad ruin]

16 times 😄

[normal quarry]

are you in?

[glad ruin]

as a coder lol

or maybe google authentication is not working

[brisk path]

It's quite expensive though

[normal quarry]

you said you were finally let in after 16 tries, are you actually in yet?

[brisk path]

Especially for something that YOU as the user have to find the use case for

[glad ruin]

yes

[brisk path]

TV subscription service? You watch TV. Duh.

Subscription for an online service in a video game? You're paying for online.

ChatGPT?...

[past fiber]

Not for that but i work as an IT so i find everyday a usecase

[brisk path]

You pay for access to a computer that can talk and use tools

The use case is more unclear, especially for the general public

[lunar arrow]

Could it be that we got Turbo already?

Or did they update GPT-4 non-turbo's database to April 2023?

[past fiber]

Yeah true for people who dont really work with computers so much is really expensive and you need to find use case

[upper walrus]

I am not able to send messages in new ChatGPt bots

[past fiber]

But when you do

Its so worth it

[upper walrus]

Is this normal ?

[silk flame]

I am the same

[upper walrus]

Thanks seems it is not rolled out for all properly yet

[silk flame]

yep, there are always some problems at the beginning of rolling out. We need to be patient.

[past fiber]

Not really i dont thing you have to be forced to pay monthly for a game or a TV with ads

[brisk path]

You aren't forced to pay here though????

[plucky timber]

If they add another tier and still fail to give it early access to features ill lol hard

[brisk path]

Free ChatGPT is a thing that exists

And the lucky free users are LITERALLY getting GPT-4 all tools

[past fiber]

Wait what?

They are getting gpt4?

[past fiber]

Yeah but when you you try gpt4 you know that gpt3 is not so smort

[dapper isle]

hello i need help, so i am thnking about buying the plus sub for the new models so in the open ai play ground in the free tire i can only see models with gpt3.5 tubo4k and 16k etc, so after taking the plus sub would i be able to gain gpt4 and gpt4 tubo? or do i have to invest money separately on the playground as well?

[past fiber]

Not exactly API and chatgpt are not related

[normal quarry]

GPT Plus and the API are separate products and billed separately

[dapper isle]

oo so buying plus sub will not add anything on the playgroud ?

[normal quarry]

no, a ChatGPT Sub gives you nothing in the API/Playground

[past fiber]

Well it new gpt4 turbo is still rolling out so maybe it will come to you later

[dapper isle]

wth

[past fiber]

But you see in the playground normal gpt4?

[dapper isle]

20+ $

[past fiber]

No in the playground (API)

[dapper isle]

no i dont, since i use the chatgpt for free i thought upgrading would let me use the gpt4 models in the palyground

[normal quarry]

for API Playground billing you go here -> https://platform.openai.com/account/billing/overview

[plucky timber]

The real play would be to give users a dollar grant a month

If they are ChatGPT subs

Now i have some expired grant from when I signed up lmao

[past fiber]

Yes for chatgpt yes but for API not. They are not the same thing. On one of them you pay every month and on playground you pay as you go

[dapper isle]

this is confusing damn

[normal quarry]

the API/Playground is one service, ChatGPT Plus is another service. They are not the same and one doesn't affect the other. They are billed separately

[paper kestrel]

Is using the GPT 4 API cheaper than paying for Chatgpt Plus every month??
Has anyone tried using only the API?

[normal quarry]

ChatGPT Plus is much cheaper for most people

[paper kestrel]

Okie, thank you friend Robert!

[past fiber]

Depens on how much you use it

[paper kestrel]

Thank you, I saw some people were debating so I was curious!

[dapper isle]

ooh, got it thank you sm

[sleek swift]

OpenAI gods please make the rollout faster

[azure dagger]

anyone find that in the new custom GPTs you can't actually hit submit?

[acoustic sinew]

Joined here today because of the turbo hype but I got the pro version last week and it's been so worth it!

[past fiber]

Do you mind if I pray with you?

[sleek swift]

Nah

[paper kestrel]

Is there a limit to the file size you can upload for the Custom GPT?

[sleek swift]

Actually I'm an atheist

[past fiber]

Thx

I belive in ai god

[sleek swift]

Oh yeah

[pine tulip]

But do you think, that there might be a higher power of some sort?

[paper kestrel]

Burden of proof is on the person making the claim!

[sleek swift]

Let's go to #off-topic or even DM. I don't want to spam this channel in case a big debate might happen

[pine tulip]

Aight!

[past fiber]

But I dont have the new futures either so its a bummer

[sleek swift]

We need more advanced AI

[dense halo]

When will I get the all features in 1 thing?

[sleek swift]

All the time

[past fiber]

Yes

I need my own AGI

[sleek swift]

Multiple please

[past fiber]

XD

[sleek swift]

That operate all together to make my crazy ideas

And really cheap

[past fiber]

Yes

[rich widget]

File uploading is broken for me in ChatGPT

[ionic thicket]

Is anyone else getting file upload errors on chatgpt?

[rich widget]

Yep

[ionic thicket]

Haha, ok. Nice timing

[sleek swift]

I can try to see if I get such an error

[ionic thicket]

Haven't been able to upload anything for several hours

Status page claims everything's fine and dandy

[rich widget]

I was uploading files about 45minutes ago but now it is not working

[sleek swift]

I don't get any error

[rich widget]

Unable to upload Full_Conversation_SPR.txt

[sleek swift]

Said it too early

It was loaded successfully but ada broke down

[past fiber]

Yeah everything kind of broken. Even that i asked normal chatgpt 4 if he can create images because i wanted to try if he switches automaticly it looks good he even said wait until i am finished but nothing ever came out

[near vine]

hello all, question! will all the plus subs be able to gain the access on gpt4 turbo?

[past fiber]

So he things he has dalle3 but he doesnt

[silk flame]

chatgpt 4 is so slow today, do they really call it "Turbo" ?

[past fiber]

From what i heard i thing its only for API not for chagpt

[past fiber]

Well there are lot of users right now so the servers cant keep up

[near vine]

wait what?? not for chatgpt? wha

will they roll out for the sub users?

[slim hemlock]

chatgpt is so slow since updates seems like went back to initial days of gpt-4, also removing options like, to just use gpt-4 instead of it skidding to use bing whenever it feels like is taking the customization away(saw the raw gpt-4 in official gpt’s), why make something that is working very well harder to use, also don’t know if this gpt-4 turbo is less performant as it can’t be that its cheaper and performs better than raw gpt-4

[past fiber]

Dont take my word for it but this is from my understanding

[maiden fractal]

That's also what I understand from the #announcements article

[solar rose]

Pretty sure they didn't, but them combining all the tools into one definitely made chatgpt 4 slower

[plucky timber]

I guess thats something else they should add to their advanced tier their thinking about

[plucky timber]

Access to any new feature on the API in ChatGPT

[past fiber]

I know they updated it to april 2023 but its not turbo

*Chatgpt

[past fiber]

I thing the problem is cousing too much people on right now and trying all the new things

[solar rose]

[pine tulip]

yes, but it's random

[past fiber]

He means the gpt4 turbo

[pine tulip]

yea, which I already have and some don't

[indigo coyote]

who was able to start a business using this chatbot?🧐

[pine tulip]

or do you mean the api?

[past fiber]

Chatgpt 4 turbo?

[pine tulip]

You can't really start a business by only using chatgpt. But you certainly can start it yourself, with chatgpt as a good little personal assistant

[pine tulip]

yup

[near vine]

damn i wasted my money then should have invested on the API

[past fiber]

He was asking if subs get chatgpt 4 turbo but we thing its only for API with 128k tokens

The subs will have normal chatgpt4 max 8k tokens

[dense halo]

Wait what about playground?

[pine tulip]

to be exact, there is no "chatgpt 4 turbo" there is chatgpt and gpt4 turbo. Some Chatgpt instances use GPT4, some others use GPT4-turbo already

[near vine]

you mean on the ones who uses the plus sub right?

[past fiber]

So you are saying there is chatgpt 4 turbo with 128k tokens?

[pine tulip]

some of them, including me, already have it, yes

[near vine]

WT

WOW

[serene grove]

Got it aswell ! Really nice.

[near vine]

WOAHHH

[dense halo]

Is "gpt-4-1106-preview" the same thing as GPT-4 turbo?

[pine tulip]

no, I am saying I am using chat gpt with gpt4 turbo. ChatGPT has a context length of about 8k tokens

[pine tulip]

yup

[serene grove]

Didn't they "upgraded" to 128k tokens, even for the web users?

[past fiber]

Ok so the only thing that changed is that he is up to date

[dense halo]

k

[serene grove]

~100k words limitation instead of 2k.

[past fiber]

April 2023

[pine tulip]

and the behaviour, which could be due to less parameters or something else they changed

[past fiber]

Yeah so only the year changed

[pine tulip]

I don't think so, but I don't know for certain

[past fiber]

If yoh need more then 8k tokens you need to use API

[past fiber]

I dont thing either

[pine tulip]

ChatGPT is ChatGPT, Api is Api. Both use either GPT4 or GPT4-turbo. Context length is greater with the Api though

[past fiber]

Well if you thing about it gpt3.5 turbo is only in playground and then there is chatgpt 3.5 fpr free users

[near vine]

then technically since they made gpt4 turbo 128k token cheaper than gpt4 in the api i think then investing on the api would be beneficial isnt it so?

than the plus ones right

[past fiber]

Well you pay as you go

[near vine]

yeah but

[pine tulip]

Not quite right. Free users in ChatGPT use GPT 3.5, but paid users can either use GPT 3.5-turbo or GPT4/GPT4-turbo

[past fiber]

Oh my god yeah sorry you are right

[serene grove]

Nevermind, web chatGPT+ is still limited to 4K tokens, which is roughly 2-3k words.
The playground thingy uses the API if i'm not mistaken, thus having up to 64k or even more (128k), which can take up to ~100'000 words for a single input.

[past fiber]

Yeah most of the people want more tokens now

[novel parcel]

why my assistant dont save in playground?

[near vine]

you need to save it manually i think

[past fiber]

Isnt Chatgpt limit 8k tokens?

[novel parcel]

yeah but it's still loading and dont work

[past fiber]

8k input and 4k output

[pine tulip]

yea everyone be stating something different. We don't know for certain I guess

[pine tulip]

oh, you sure?

[near vine]

oh the servers are slow becuse of the upgrade they changed the whole UI added new models to it and because of the heavy use it is expected

[serene grove]

yes, 8k input and 8k output, if i'm not wrong

[past fiber]

Well not exactly

[pine tulip]

if I ask it, it states 4k

lol

[past fiber]

Yeah dont ask him

[serene grove]

By the way, i mainly use chatGPT, but is there any differences between chatGPT and https://platform.openai.com/playground?mode=chat ? GUI-aside

[past fiber]

That is something he doesnt know

[novel parcel]

im not sure i think something is wrong

[serene grove]

ChatGPT is not good about giving precise informations of itself

[novel parcel]

it says network error

idk why

[plucky timber]

You do not currently have access to this feature

[novel parcel]

I mean in playground

[plucky timber]

You do not currently have access to this feature

😂 okay ill stop

[novel parcel]

lol

[past fiber]

Yeah somewhere around there

[serene grove]

Bump.

[past fiber]

They dont really talk about the model specs

[brisk path]

They're technically the same model (of you, of course, use the latest GPT-3.5 or 4 model)

[past fiber]

The api is custom the chatgpt is more for chat with you but with the new gpts it shouldnt be a problem

[brisk path]

And they should respond almost the same

The difference is the utility. ChatGPT is a chatbot, and is meant to be a chatbot, while Playground is a testing ground for the API, which lets you integrate the models into other services

[serene grove]

Right. Thanks!

[pine tulip]

it lets you automate your pc

[plucky timber]

FridgeGpt

[pine tulip]

well not only, sure. but thats a usecase

[past fiber]

Wel we saw bingchat in windows

[serene grove]

Automate your Samsung Fridge™️ with GPT4 included.

[plucky timber]

"Emptying fridge"

Noooo!

[pine tulip]

some people did even more stuff with it, like changing almost everything on their pc, shopping amazon by voice and such

[past fiber]

Lightsgpt

[plucky timber]

"Good morning"

3 at night

[indigo coyote]

Will there be a custom subscription for GPT-4 turbo or will it stay the same?

[serene grove]

"Your midnight snack has been [locked], by your [wife]"

[pine tulip]

guess nothing will change

[past fiber]

Turbo with 128k tokens is only api right now

[plucky timber]

😢

[past fiber]

XD

[pine tulip]

I'm wondering how often some OpenAI Devs read this channel and can't contain their laugther

[indigo coyote]

will it be added to the chatgpt itself? If yes, when will it be available?

[novel parcel]

can someone check #ai-discussions i could not send any image here

[past fiber]

Dont thing it will
Its just too much tokens

[plucky timber]

Over time it will

[pine tulip]

got to #dev-chat

[plucky timber]

Its the ever developing loop of hardware software limitations

[past fiber]

Maybe 32k tokens

But i dont thing more then that

[plucky timber]

Maybe in 50 years 3200K tokens 🤷‍♂️

[novel parcel]

thx

[static wharf]

will openai quit calling itself "open"

[past fiber]

Yeahm then yes

[past fiber]

Why? They are open

[pine tulip]

more like 3200kkk tokens

[static wharf]

not 50 years more like 5 months at the way things are working

[indigo coyote]

I mean agents, using all the features in one chat room and the rest of the features they announced.

[plucky timber]

Hopefully man 😄

[static wharf]

they are for profit, don't open source models anymore, don't even reveal model info

[past fiber]

They have the best model on planet why would they open source it

[serene grove]

"Write me the Bible in the minecraft enchantement langage"

does it

[plucky timber]

I stopped reading it as Open in opensource and more as open in open a box

But yeah shame its less OS

[static wharf]

openai was originally formed to be non profit and to open source their work

[gleaming anvil]

how do you access gpts? i have pro...

[past fiber]

And even if they made it public you need atleast 500gb of ram to run it

[plucky timber]

It's prolly gonna warn you about potential real world effects 😂

[serene grove]

No? I never heard of them going open source. It's a private compagny and they want to keep their thing. Why giving away the source codes to other competitors?

[pine tulip]

They open source their findings. Just not the entire source code, which would be kinda dangerous.

[compact gyro]

it says 'coming soon' for most of users
and only accessible by rare few accounts

[past fiber]

Yeah but now they have the best model why would they let that go

[gleaming anvil]

bruhuhhhhhhhh

[plucky timber]

Wait while spamming the normal link of chat.openai with /gpts/editor appened

If it says no access welcome to the club

[static wharf]

bruh, search them up, originally a non profit organization

formed a few years ago

[gleaming anvil]

"You do not currently have access to this feature"

ffffffffffffffffffffffffffffffffff

[static wharf]

but then they hit it big and became for profit

[plucky timber]

@gleaming anvil I welcome you officially in the got scewwed by the queue club

[past fiber]

Yeah they were but if you were there as a ceo you wouldnt just open source gpt4

[gleaming anvil]

thats some bs

[static wharf]

bro thats the point

[serene grove]

"OpenAI was created as an open source (which is why I named it “Open” AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft." nevermind.
Well, not their entire fault, it's mostly because of Microsoft that they went closed source.

[static wharf]

there's not supposed to be a ceo

its non profit

`2015–2018: Non-profit beginnings

In December 2015, Sam Altman, Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel, Elon Musk, Amazon Web Services (AWS), Infosys, and YC Research announced[12] the formation of OpenAI and pledged over $1 billion to the venture. The actual collected total amount of contributions was only 130 million until 2019.[5] The organization stated it would "freely collaborate" with other institutions and researchers by making its patents and research open to the public.[13][14] OpenAI is headquartered at the Pioneer Building in Mission District, San Francisco.[15][16]`

[past fiber]

Well times are changing I love chatgpt4 so I dont really have a problem

[plucky timber]

Microsoft is the type to release entire OS updates just to handicap cracks for stuff like photoshop

[dense halo]

is gpt4 turbo better than regular gpt4

[serene grove]

Can't wait for Windows 12 integrating ChatGPT 🔥🔥
Jokes aside, windows 12 is really coming up "soon". I still remember when they said "Windows 10, the last Windows ever"

[past fiber]

And i wouldnt say maximum profit just to answer on your promt is expensive

[pine tulip]

I remember when they said windows 8.1 is the last windows ever

[plucky timber]

Clearly vista was the shi

[serene grove]

WXP/W7 :(

[plucky timber]

Yeah honestly that tbh

[pine tulip]

indeed, I loved to delete my recycle bin, so that I wasn't able to clear my recycle bin until I reinstalled windows

[plucky timber]

But hey I gotta admit they finally got me paying one way or another

Microsoft that is

[brisk path]

XP and 7 are overrated

[plucky timber]

Lmao

[brisk path]

Well, maybe not XP, but definitely 7

Lmao bro got automodded

[serene grove]

Well, tbh i used WXP, and went straight up to W10.

[plucky timber]

Discords automods is pure trash

[serene grove]

So idk about Vista or 7, but i heard that vista was a nightmare, and 7 was the "peak os"

[brisk path]

Wth

[pine tulip]

I'm back

[pine tulip]

yea 7 was the shiii

[plucky timber]

@serene grove Are you a company ? XD

Every company i know did that

waited years to upgrade

Sticking to XP , while 7 vist and 8 where there

Huge security risks 😂

[serene grove]

Probably /j

[maiden fractal]

I suggest for discussion not related to chargpt #off-topic . I am not a mod but before you get a warning :/

[serene grove]

Right, right.

I wonder if we'll soon be able to have several custom instructions?
For example, I have 2-3 long instructions of 1200-1500 characters each, and to switch from one custom instruction to another, we're forced to keep text files and copy paste + create a new chat...

[pine tulip]

thats basicly what the new "Own GPT" Feature does

[silk flame]

Guys, I have problems with the voice feature. Since the latest update, the chat seems to misinterpret non-verbal sounds such as coughs or sneezes as words and proceeds to transcribe them inaccurately. For instance, a simple cough was transcribed as “Thanks for watching”. Do you encounter the same bug?

[pine tulip]

You can feed it with your behavioural instructions and knowledge and save it for later use

[serene grove]

I access this thing via "playground", or where there be a "button" implemented in ChatGPT?

[solar rose]

Man, I'm getting absolutely destroyed by "Error in input stream"

[pine tulip]

You will be able to click on Explore in the side bar and create one from there. I can send you a screenshot

[serene grove]

Yeah, i am interested.

[pine tulip]

click here #off-topic message

[delicate sable]

Can we somehow achieve a "conversations folder" functionality?

[tame hornet]

A question, are they still rolling out the ChatGPT4-Turbo update? My application interface has not changed yet. Thx

[pine tulip]

That would be awesome. But then it would probably look like some sort of AI Note taking app

[pine tulip]

yes, could take days or weeks

[tame hornet]

Okay, I'll have to wait. Thank you very much

[pine tulip]

[loud vessel]

Still refreshing and waiting for the update to go through in my part of Canada....

[jaunty harness]

"The conversation is too long, please start a new one." I barely did anything

[pine tulip]

maybe it's not just your conversation that is too long

[jaunty harness]

is this page not the gpts?https://platform.openai.com/assistants

[eternal rock]

`` Server Tip: ``

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

[pine tulip]

no

this is https://chat.openai.com/gpts/discovery

[idle mason]

Anyone know how many tokens gpt plus currenty has access to?

[jaunty harness]

2000-4000ish

apparently it depends on the beta test

[idle mason]

Are we getting the 128k soon?

Or is that just the api?

[pine tulip]

just api

[feral zinc]

api

[past fiber]

Api

[idle mason]

bruh, thanks

4k is not enough for the chat.openai.com

[past fiber]

I thing its 8k

But yeah we get your point

Ist still forgeting

[idle mason]

Is anyone using another website that uses the api instead of the formal one?

[pine tulip]

when are the devs in the office?

I am wondering if they are already actively working on those issues

[feral zinc]

its 4.40am in SF

[tardy totem]

Is there anywhere we can see the current status of the service (instead of asking every time there is a problem)?

[pine tulip]

thats why I was asking lol

[eternal rock]

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.

[delicate sable]

At least to bring the conversations search functionality to Desktop 😢

[honest jay]

status.openai.com

[past fiber]

Its not midnight right now in USA?

[pine tulip]

the us has 5 time zones, or 9 depending on who you ask

[tardy totem]

Thanks

[past fiber]

Well

[delicate sable]

You have that tradeoff between sending everything categorized into certain conversations, and end up with gigantic conversations, or start new conversations for each question, and then everything is lost in the abyss of old conversations...

[past fiber]

Take one

[normal quarry]

it's 4:43am in SF now

[tardy totem]

Well ChatGPT isn't answering me anymore. 😦

[pine tulip]

GhostGPT

LeftonreadGPT

[tardy totem]

Yup. That.

Tried my API app and it works just fine. :/

My API app doesn't handle markdown though. Makes coding with it useless.

[past fiber]

when subs are having problems with gpt do you thing all the gpt enterprise customers have the same problem

Or they are running on dedicated servers?

[tardy totem]

My guess is no. A company like this have many server instances running.

Tons and tons of virtual servers.

[past fiber]

Well everything is running over azure

That is why i am asking

[tardy totem]

I'm not surprised.

I freaking hate working with Azure btw, but it probably fits them well.

[past fiber]

It is what it is

[tardy totem]

Yup.

Luckily I don't need to. 😛

[past fiber]

Did you tried one or you have enterprise version?

*Gpt

[tardy totem]

I've got Plus.

[sterile glacier]

Is there even a good alternative for azure?

[tardy totem]

On premise

So it turns out 3.5 works fine for me, but not 4

[dense halo]

They gonna wake up late, they probs had a goodnight last night

[tardy totem]

I switched to gpt4 in my app, and just as my app got its response Chatgpt started working again

[acoustic sinew]

Does anyone know how to go back to the non-turbo version of GPT 4? Someone on reddit mentioned it's possible to do in the explore section but I don't see the old version there

[delicate sable]

Did you treat her bad?

[dim kraken]

Do you gues can upload csv files or even ask to run a python code after update? Cause mine does not work this way(

[acoustic sinew]

Works for me but only half the time

[tardy totem]

Never. I can't even play a badguy in games. 😁

[dim kraken]

Thats weird. I also cant chat in any gpt chats in explorer section

[acoustic sinew]

probably being rolled out, I tried a few and they worked. Felt a bit buggy though

[dim kraken]

Well, thank you for respond. I will wait than

[tardy totem]

So, making a program with the help of GPT, and being able to show it the output is amazing!

The only problem with this is me playing around with it means I will have to work half the night. 😂

[rapid knot]

hi

[neat nimbus]

"An error has occurred please try again later" for a bit now... Just me?

[pine tulip]

no, please try again later

[tidal hedge]

Is it an intended behavior for a server load error on prompt to inmediately permanently change a GPT4 to a GPT3.5 chat, if unrelated to the 50 messages to 3 hour quota?

[untold seal]

Guys even for chatgpt-T inside chatgpt the token limit will remain the same and won’t increase???

[tidal hedge]

thinking I may do a bug report if unintended. Otherwise a suggestion for a toggle if I'd rather just wait than automatically convert the chat to 3.5

[mental rose]

how do you get access to GPTs? more precisely, how do you turn on developer mode?

[rotund geyser]

Hey folks

[eternal rock]

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 @languid valley immediately.

[normal quarry]

you only need to post once in one channel

[rotund geyser]

Are plugins dead now!?

After the new update?

[novel parcel]

anyone has trouble to save a gpt assistant in playground

[rotund geyser]

Omg 😱 plugins were my life

[gilded tartan]

It's not possible to add custom messages to the custom versions. I can only use the default questions. Anyone else having the same issue?

[novel parcel]

what do you want to do

i asked gpt4

go to api discussions

[tardy totem]

My chatgpt keeps getting network errors. They're probably doing update stuff right now

[patent mauve]

i believe so

[rotund geyser]

Damn

That stuff was so important

[tardy prawn]

Vision can’t sight-read.

[patent mauve]

is anyone else having issues with the GPTs? for example with the HotMods, I don't see a way to upload an image as it specifies. Or even with the creative writing coach, when I provide a passage to review and I click on the black up arrow, nothing happens

[normal quarry]

things are a little broken right now, sit tight and patiently wait while stuff gets back on track

[sharp pilot]

I believe it is system wide, those of us who have the new Explore section with the GPTs can't actually submit messages to them (apart from clicking the suggestions).
I think they released the UI before the feature was ready, or planned for a quick rollout that hasn't happened yet. Either way they kinda presented them as live at their talk when that's not quite true yet.

[novel parcel]

when can i expect the gpts

[novel parcel]

i also get this in playground

[mighty arch]

will GPT3.5 have access to use the image feature that was announced, like, september 25?

[pine tulip]

no

[mighty arch]

alright

[pine tulip]

only gpt4

[brisk path]

Image inputs are built into the GPT-4(V) model itself

Which means that without external tools, 3.5 cannot deal with images in any way

[delicate rover]

The OpenAI blog said that example GPTs are available today, but my interface has not been changed

[wanton gazelle]

Who is still missing the features?

[plucky timber]

I think entire europe

[dense halo]

GPTs when?

[elfin rain]

I'm still missing features

[pine tulip]

wait a few weeks, maybe months

[dense halo]

😭 pls no

[novel parcel]

they said this week

[brisk path]

days

[dense halo]

How many tho?

[tardy totem]

We've been without the features so far so we'll be fine until they arrive

[dense halo]

But I neeeeeeeeed them

[tardy totem]

Although I had begun experimenting with function calls before this was announced.

[swift widget]

grow up

[dense halo]

no u

[novel parcel]

same

[tardy totem]

I'm lucky. I have so much work to do that I don't have time to miss these functions. 😁

[untold seal]

Guys even for chatgpt-T inside chatgpt the token limit will remain the same and won’t increase???

[elfin rain]

Oh when I said I'm missing features, I mean features like my chat-gpt4 being able to access the internet

[dense halo]

Guys has the voice issue been fixed yet?

on ios

Is chat gpt down currently?

[brittle birch]

yeah they're available to me on the explorer but then trying to message after the first prompt nothing works, for me

[novel parcel]

do you have gpt plus

[elfin rain]

yes

[queen crow]

Why did they reduce the font size in the chatGPT app? It doesn’t look nice! The web interface is also bad now! Please revert to the old version

[pallid iron]

there appears a folder icon on ios, but the pdf file can not be selected after clicking on the icon, can not upload pdf files?

[dense halo]

Playground assistants file upload isn't working for me

[novel parcel]

it working for me but I can't save it

[dense halo]

Same

When I click the save button, it just gets stuck loading forever

[novel parcel]

yeah

[pine tulip]

yeah

[paper kestrel]

is there a limit to the amount of data you can upload to assistant CustomGPT?

[novel parcel]

in playground?

[paper kestrel]

Yes friend!
For making the API, I do not see anywhere about the limits and how much you are allowed to upload.

I also want to know if you have to pay each time you upload a new file?

[novel parcel]

you need to pay for the api

im not sure

[fringe ruin]

since the new update where gpt4 is supposed to have everything built in, mine seems to have lost browsing functionality and repeatedly says it can't browse. Has anyone else had the same issue?

[brisk path]

I haven't had this issue but quite a few have

[paper kestrel]

I got timed out for sending a "I saw what you deleted gif" XD

But thank you for the help ☘️ !

[shy mountain]

I think they gave alpha to ex premium user becouse i was plus member for 4 month and i got alpha even if i have a free plan now

[fringe ruin]

Glad to know im not the only one. It's not working at all

[fringe ruin]

Mine can't either

[elfin rain]

What are they bringing out new anyways?

[pine tulip]

customized gpts

and a full multimodal gpt4 turbo

And you can earn money by creating your own gpts

so openai is sharing their revenue if you create a nice one

[tardy totem]

I'm just happy if I get faster code generation

[cunning flame]

will chat gpt 4 from the site get higher context too?? or everything is just for developers

[upbeat pollen]

Anyone tested if new ChatGPT version has context window above 4k/8k yet?

[pine tulip]

no

[pine tulip]

it hasn't

[dawn flicker]

Are you able to send messages in a chat with one of the new GPTs? For me, the send button is enabled, but I can't actually send any message when I hit Enter or click the Send button. Tried multiple browsers and mobile too. Regular chats still work.

[worn geode]

Input yes, Output no

[upbeat pollen]

What do you think input limit is at the moment?

[sinful eagle]

Has everyone with chatgpt plus gotten the all tools feature yet? I thought sama said it would be released yesterday. I still don’t have it tho and I’m in the US

[upbeat pollen]

I got new gpts and stuff this morning in Serbia

[novel parcel]

I asked what is your cut off date and it says My knowledge cutoff date is April 2023. That means I'm informed about events, discoveries, and developments that occurred up until that time. If you have any questions about events or information that came after April 2023, I may not have the most current updates.

[worn geode]

Haven't tested it, but it's definitely more than 8K which was the previous limitation. Some say 32K

[upbeat pollen]

32k would be really nice

[sinful eagle]

Oh ok. I haven’t gotten GPTs either yet 😔

[dawn flicker]

are you able to actually send a message to one of the GPTs?

I see them, can open a chat, but cannot send a message

[red crag]

Huge amount of problems right now, literally unusable ChatGPT

[dawn flicker]

okay, so it's not just me

[novel parcel]

chat gpt is working

[dawn flicker]

what do you see?

[upbeat pollen]

yeah, works for me

[dawn flicker]

got it... that's good to know, idk what's up with mine then

[red crag]

Every possible thing, errors, chat literally breaking

[dawn flicker]

also regular chats? or with one of the new GPTs?

[urban river]

yeah I cant send any messages except to the base gpt

[red crag]

the new

[dawn flicker]

high five!! okay I'm glad at least someone can repro

[red crag]

even the regenerate button doesn't appear when the chat breaks

[dawn flicker]

alright, that's the main thing I wanted to know — it'll probably get fixed then... I mainly wanted to know if it's a "just me" issue

[red crag]

and if you modify a message and resubmit nothing happens

[novel parcel]

chat gpt works fine for me

[red crag]

not the new one

[novel parcel]

which one

[red crag]

was doing fine till 2-3 minutes ago, but now broke

[novel parcel]

https://status.openai.com/

[crisp briar]

Like if you have the new ChatGPT update. Want to see how many of you lucky people have it

[novel parcel]

it says the cut off is april 2023

[dawn flicker]

I have it but it doesn't work, does that count? 😛

[dawn flicker]

there are definitely issues... worrisome to see a green status board, maybe they don't know about them

[novel parcel]

mh

[dawn flicker]

not sure how to "officially" report

[crisp briar]

Haha it counts but that might be worse then not having it

[frosty valley]

What a wild day of development with OpenAI, the Assistants and Retreival tools seem very promising, can't wait for it to go live

[novel parcel]

up to the cut-off point of my knowledge in April 2023

[timber oyster]

Dall-E 3 currently not working for some time on multiple devices. fix this please

[sick bay]

Mine wasn't working then started again, now it reckons I've reached my daily limit. Not sure if that's a thing or if the bot is just wigging out

[untold seal]

guys how many pictures can the api handle in one go?

[novel parcel]

#dev-chat

[pine tulip]

Its 7am in SF. Can the devs please attend to work?? lol

[wanton lichen]

I still haven’t received the rollout lmao

[novel parcel]

im waiting too

[pine tulip]

You're not missing much

Looks cool, but nothing works

[prisma raptor]

Anyone know where I can see this "Keynote" everyone is talking about?

[pine tulip]

youtube

[untold seal]

Guys will the token limit for chatgpt-4-T be the same as the token limit for the normal chatgpt-4 we had until now?

[crisp cairn]

i think so

[wanton lichen]

I thought it was going up to 32k

[pine tulip]

this

[untold seal]

I honestly think that would be a huge upgrade if they would increase the token limit

[novel parcel]

As of my last update in April 2023, the context window for GPT-4, which I am based on, has a maximum of 4096 tokens
lol

[oblique thistle]

Hi there, since the new update I click on "Explore" and I can write prompts but for whatever reason I can't send the prompt. If I use the "normal" ChatGPT everything works as intended. I tried it in different browsers, still the same.. Any idea why?

[pine tulip]

it doesn't know technical details about itself

[novel parcel]

so dumb

[pine tulip]

there is no reason to ask it

[full veldt]

I actually think it's worse than that

[pine tulip]

wdym?

[full veldt]

If the cutoff is April 2023 then it will be partially trained on old articles and press releases about the original release of the model

If any of these contain technical details then it could be reciting this information which it thinks is accurate, but is actually six months out of date

[pine tulip]

yea right

so there is no reason to ask it

[jolly python]

Was browsing removed from the Android app? I know have the integrated models version (DALL-E+ default) but this version doesn't seem to have access to browsing

[west ridge]

I saw the dev speech says the chatgpt4 will integrate all into a single model today, but I still have many options under the gpt-4 button, anyone knows why?

[pine tulip]

@jolly python

[novel parcel]

its not rolled out fully

[normal quarry]

it's rolling out over the week, so sit tight, be patient and it will come

[west ridge]

Alright thanks

[pine tulip]

But Sam indeed said today lol

[novel parcel]

in the meanwhile you can use autgpt xd

[serene horizon]

😒

[lucid phoenix]

I can process up to 4096 tokens in the input, and I can generate up to 2048 tokens in the output. This allows for a substantial amount of information to be considered when providing a response and a detailed answer when necessary.

[lucid phoenix]

yea i already have it on web it’s pretty cool and convenient

[serene wren]

I am getting these "Error", "Network Error", "Error Generating Response" or "Error in message stream" messages while I am trying to get responses. Is this an issue on OpenAI's end or is this something I can solve? I don't know if this is maybe a cache issue, or other browser setting problems. As of my knowledge (and the fact I am able to send this) my network connection is fine. It seems as though the messages are timing out. But some of them are giving me the error after only generating a few seconds. I have been combatting the issue by having GPT pause after each step in my problems. However, it doesn't seem to be working. If anyone knows whether this is something I can solve on my end or not, please let me know :)

[lucid phoenix]

it could be your conversation is too long, happened to me i just had to make a new chat

[serene wren]

It has been a pretty short conversation. I assumed that as well. But it happened after the first request. The response appeared to have timed out perhaps. But even my way of combatting that isn't working

[strong breach]

I wonder if the new GPTs will have some sort of deletion safety mechanisms in place such as: "Are You sure you want to delete? Type the name of the GPT below to confirm...." Because I can see myself spending a lot of time training my custom gpt and then being silly and deleting it by mistake.

[west ridge]

wow that’s cool, good for you

[pine tulip]

🤝

[pure otter]

Is any of the new features working for anyone?

[serene wren]

New features?

Like the GPTs? Or just the overall new layout and integration?

• GPT's aren't out yet
• The combined integration is working, but I (personally) am getting alot of message errors

[hidden ibex]

or an "Undo" with a temporary 24 hour soft delete before it goes away forever

[pure otter]

I got this notification in the app but it doesn’t work. It won’t search the web.

ChatGPT can now browse the web, analyze data, and generate images.
These capabilities are now built into GPT-4, no need to select them. Available for Plus users.

[serene wren]

Also, idk about the rest of you, but my custom instructions don't seem to be working as consistently anymore.

[cunning flame]

We’ll begin rolling out new features to OpenAI customers starting at 1pm PT today.

that's what the post says

[dense halo]

All my saved conversations disappeared.
I already did clear browser cache, try to re-login from multiple browsers (Safari, Firefox, Brave) and tried incognito in all the different broswers on mobile and on desktop too, yet non of my conversations back. Anyone has any simillar issue llike this since last Friday (3-Nov-2023)?

[serene wren]

I have never tried the websearch function, so I do not know. Sorry, I cannot help much with that

[novel parcel]

I asked for an image but it says Let's go with an abstract image that represents the GPT-4 assistant as a sophisticated and interconnected network. I will now generate an image based on this concept. Please hold on for a moment.

[pure otter]

❤️

[elfin rain]

Does Chat GPT have a feature where you can name and reference conversations in the past?

[serene wren]

All I can think is either you deleted them, you're logged into the wrong account, or you may need to do a browser cache reset that is dated back even farther.

Usually if that problem happens on OpenAI's end, there is a message stating the issue happened on their end

[cunning flame]

Nope, unless you reference them in the same conversation

but you gotta give it context

[novel parcel]

you mean like a memory

[elfin rain]

Yes like a memory

[novel parcel]

you can use memgpt

[earnest sierra]

Does anyone else not have access to dalle3 or vision? It doesnt let me upload or use it in gpt 4 or any gpts

[novel parcel]

i have dalle 3 and internet and vision on my phone

[elfin rain]

I have to keep explaining my entire project in a nutshell everytime I want to consider some new functionaly.

[pure otter]

Nothing in the new single model is working for me

[earnest sierra]

phone isnt updated to new version yet

[serene wren]

You can always save the messages as scripts, or pdf's, or even screenshots, and provide them to your current conversation of GPT for them to understand the previous conversation.

That's what I do

[novel parcel]

i have it on iios and on my pc but voice only on my iphone

[cunning flame]

even the ones that are extremely long? because some projects are just extremely big, a bunch of text and context

what I do is just ask gpt at the end to do a summary of everything, like as a checkpoint, and go to a new chat for better performance

[elfin rain]

Oh, so to use memgpt I need the API. Oh so then people are using OPEN AI on the desktop this whole time and I'm the backwards one?

[serene wren]

It depends. You could summarize it. But I imagine if you save the conversations as PDF scripts, you will be fine

[novel parcel]

i will try it out right now

[solid dagger]

There is a chrome plug-in for that, gives you some more crazy features. I think it's called superpower GPT

[serene wren]

Didn't they make that a pay-to-use plugin though

I think I used to use that, but I started getting emails about paying for it, so I stopped

[cunning flame]

is this actually better than just reminding gpt with context? seems interesting

[earnest sierra]

chatgpt rn is actually worse than it was than on release date. this is terrible from openai

[novel parcel]

i think yes but you have to use it locally and im not sure how to configure

[cunning flame]

so you use it thru the command line instead of a gui?

[novel parcel]

yes

[cunning flame]

I see, thank you

[novel parcel]

np

there is also gpt-engineer and autgpt which is cool

[abstract nest]

GPT is weird today. It keeps deleting text from paragraphs when it is only supposed to be proofreading and fixing basic misspellings and grammar. I am having to tell it to specificly NOT delete information.. like it is trying to summarize everything...

my customiztion is not as effective either

[strong breach]

Am I the only one caught on this loop?

1. Cool feature! Used it for a month.
2. Nah 20.00 is too expensive
3. Cancel subscription
4. OpenAI releases new jaw dropping feature
5. I must have it!
6. Resubscribed

[high horizon]

so will the charge for chatgpt 4 turbo be a flat rate like it was before? Or will they now charge you per use?

[serene wren]

I've been having similar issues. It seems to be capacity problems with alot of people using it right now. It shouldn't be a long term issue

[abstract nest]

thanks

[novel parcel]

Yes similar

[strong breach]

🙌

[cunning flame]

for me it is too useful to cancel, it really saves a lot of time with my works

[novel parcel]

I mean I cancled it because it's to expensive but yeah

[serene wren]

Not caught in that loop yet. I use ChatGPT way to much lol

it is well worth the $20 for me

[cunning flame]

yeah, with a well paid remote job those $20 are nothing and it makes the work flow much faster and stress-free

[strong breach]

I wonder if they will release membership tiers eventually. $5.00, $10.00, $20.00

[novel parcel]

I think if the decrease the amount more people will pay and maybe they will get even more money for there servers

[serene wren]

I don't have that kind of job (as nice as it sounds), but it helps me rationalize, streamline, and create efficient solutions in my day to day life.

[cunning flame]

maybe and probably the difference would be the message limit cap

[serene wren]

Sir, they have gotten like 21 billion dollars from Microsoft funding alone (not counting other sponsors). I think they are good on money lol

[strong breach]

💡

[novel parcel]

haha

[tawny steppe]

Does anyone keep encounter the error today when using gpt4?

[sharp pilot]

Microsoft isn't really a "sponsor", I forget the exact details but they are the majority stakeholder for OAI but have some weird arrangement where they don't have to be paid back and it's guaranteed not to give them a controlling interest in the company.

[cunning flame]

same, though, I tend to not rely too much on it for my IRL decisions

[sharp pilot]

Microsoft is basically the forever largest external owner but also can't be a 51% type owner.

[serene wren]

The whole point of ChatGPT+ is to have a group of users they can use as data. The price is higher ($20) because that means they can expect the group of users to be a selection of more serious users that they can analyze rather that just a ton of highschoolers who want homework answers. They use this group of people to test settings, get diagnostic data, and analyze user activity.

[novel parcel]

will they drop the update today?

[sharp pilot]

OpenAI is weird, they announce new features, launch them, they aren't actually live, and everyone seems to think that's fine.
Granted they are probably doing a quick rollout it's just not really appropriate to do it that way.

[cunning flame]

exactly, in reality you don't need gpt-4 for common stuff like highschool homework

[serene wren]

I like to use it for IRL decisions by having it argue with me, until I feel like all options are understood and I can finally make the best decision

[novel parcel]

I only use it for new features but as it can only generate text it's not really helpfull in many ways

[serene wren]

Well...I mean...watching GPT 3.5 do math is a nightmare.

[novel parcel]

google will soon release gemini with imagen which is like gpt4

[hidden ibex]

https://openai.com/our-structure - their non-profit structure, charter, and overall setup is designed to maintain the integrity and control of OpenAI, they can pull the plug on Microsoft if they deem it necessary

it's designed not to work like most startups while having the advantages of rapid development and investing returns for partners like MS

[serene wren]

Microsoft donates and in return get help and options to integrate GPT into Windows and Bing

they are not a stakeholder though

[zenith falcon]

It accurately talked about something released in Sept 2022

[cunning flame]

there are probably better tools for that instead of gpt

[hidden ibex]

april 2023 is the latest greatest update fwict

[zenith falcon]

It's a random rollout

[brisk path]

But there aren't better tools that support natural language

[serene wren]

It would be a civil nightmare....but the second OpenAI allows investors and stakeholders to well...do just that, is the day ALOT of people become very very rich, and the overall quality and safety of AI becomes very very scary

[hidden ibex]

:aaaaaaa:

[cunning flame]

so right now for math with language interpretation gpt-4 would be the only option in that case?

[strong breach]

What if you can destroy SIRI and replace it with the Whisper API? Only then will there be peace in the world.

[brisk path]

Siri sucks

Literally the worst virtual assistant

[strong breach]

💯

[brisk path]

Also no one uses it anyways

Why is Siri even in the conversation

[normal quarry]

maybe stick to ChatGPT here, discuss other AI and assistants in #ai-discussions

[vital glen]

When the FOOK will I get access to the new gpt-4 turbo/new chatgpt all tools? Been paying monthly for 6+ months and nothing still for my account.

My own father already has access.

[polar ocean]

Anyone else notice that all GPT4 Beta features for the iOS app disappeared today?

[cobalt idol]

still dont have TTS/beta options in android app (I have everything else outside the app)

[novel parcel]

there are still here on my iphone

[brisk path]

Why is it all apple devices 😭

smh android doesn't get any love anymore

[novel parcel]

yeah because it actually bad

sry

[pine tulip]

they are out already, wdym

[brisk path]

[zinc sentinel]

I still haven't received the new chatgpt 😦

[brisk path]

okay but the android app has all the features

[strong breach]

Here is an idea: What if you can code your custom GPT, then deploy it as a month to month membership to your parents. So that they can get the help they need when it comes to asking tech questions and program their universal remote.

[brisk path]

[pine tulip]

good idea, lets scam our parents

[strong breach]

💯

[dense halo]

Why isn’t voice working for gpt4 still

[pine tulip]

it is

[novel parcel]

it is on iphone

[dense halo]

It better be fixed by tmrw

[dense halo]

Not mine

[cobalt idol]

not for everyone

[brisk path]

Because the iOS app is acting weird

Android seems to be a bit more consistent

[dense halo]

It tells me to switch from search to default as voice isn’t supported on search. But I’m on default

[cobalt idol]

I never had voice or beta options in android app
since the release

[brisk path]

Hell, I can even chat with custom GPTs there

Web is broken

[dense halo]

I’m gonna uninstall the app and reinstall it

[north osprey]

I was checking GPTs creation and turns out you can only upload up to 10 files in the Knowledge section, and also knowledge in GPTs seems to be only up to January of 2022. That’s not really that impressive. Am I doing something wrong, or is this feature that limited?

[dense halo]

Ok that didn’t fix the problem

[misty osprey]

Anyone else have access to gpt 5

[novel parcel]

what

[digital prism]

I've got gpt 6 you plebs

[hybrid summit]

What are the limits for GPT 4 messages? wasit updated?

[elfin birch]

imagine not having GPT-Toothbrushing

[digital prism]

what is this

[hybrid summit]

I mean on their site and not API

[vapid notch]

I'm having a hard time sending messages to the custom models

[misty osprey]

Truly a life changer

[novel parcel]

i think i got gpt4 turbo

[vapid notch]

they just wont go through

it actually wont even let me hit send on them for some reason

[elfin birch]

Not explicitly mentioned in the keynote but some testing showed a 100 message limit

[dense halo]

Someone needs to call Sam Altman and tell him all these problems

[elfin birch]

context?

[rancid otter]

another day and my chatgpt plus still doesnt have any update

[light widget]

Is there like a form or something where i can sign up for all tool modee?

[novel parcel]

na

[light widget]

😦 sad

[zenith falcon]

Only thing to can do is wait, as it's a random rollout

[unique loom]

how will you know when you get access to gpt-4 turbo? won't it have the same name at gpt 4?

[elfin birch]

It will have fast inference and the ability to evaluate large documents

[pine tulip]

It's at least more than 78

so 50 cap is gone

[knotty crypt]

does the chatgpt-4 also have an increase in the prompt token amount?

[tardy totem]

That gave me an idea: I should keep track of how many messages has been sent the last time-slice in my app.

gah, still getting network errors every message.

[novel parcel]

is this gpt4 ? I can try to provide information that's accurate and up-to-date to the best of my knowledge as of April 2023.

turbo

[elfin birch]

Can it read large documents greater than 32k tokens?

[fiery nexus]

no

[undone cairn]

chatgpt plus has been updated to gpt4 turbo?

[ripe marsh]

What is the token limit when using ChatGPT 4 at chat.openai.com (I have ChatGPT Plus)?

[fiery nexus]

8k tokens

[tardy totem]

so now 3.5 is slow as heck too. 😦

[ripe marsh]

Didn't it used to be 4k tokens for chat.openai.com (even ChatGPT 4)?

[novel parcel]

na

[fiery nexus]

No its always been 8k

[tardy totem]

For me it is. Whatever module I'm connected to is probably in the middle of updating

[fiery nexus]

In the beginning I believe it used to be more

[novel parcel]

slow internet?

[tardy totem]

Gigabit, 8ms to the test server 400km away.

[ripe marsh]

Is that per prompt?

[outer seal]

i see 4096 max for gpt4 vision/dalle, 8192 for ada/browsing/plugins/3.5turbo on /backend-api/models

[fiery nexus]

We're talking about ChatGPT, and it used to be 8191 tokens max for GPT-4 Code Interpreter... I am not sure with these new updates. I frankly don't like them

[outer seal]

yes that is the chatgpt backend api

[ripe marsh]

So if I am using ChatGPT 4 on chat.openai.com, is it 4k or 8k tokens?

[tawny steppe]

What happened to gpt4 today? It generate very slow and keeps encounter the network error

[slow marsh]

I'm confused. On one hand GPT-4 says its integrated with DallE, browsing and more, but on the other hand it tells me it can't generate images nor browse the web.
What's going on? can it or can it not?

[fiery nexus]

There's no way to choose code interp anymore 🥲

[slow marsh]

None of the tools work though

You can't code interpret, dalle or browse with GPT-4

its says all tools but none of the tools work

[modern dragon]

go to your settings

[past fiber]

dont ask chatgpt about himself he is kind of dumb about himself and is quessing or he is taking old data

[red mountain]

I think too many users are trying them out now

[limber scarab]

I haven't gotten any updates for some reason

[past fiber]

everything other he is smart

but not his specs

[slow marsh]

and what there?

I have everything enabled

[modern dragon]

make a new chat hover ur mouse over gpt 4 and click what you want to use

[fiery nexus]

I tried this but its the updated UI

How do you get access to the code interpreter specifically

[humble hamlet]

When will india get the Gpt 4 turbo update?

[slow marsh]

dude what are you talking about this was true before the update but now there's only gpt-3.5 and all-tools gpt-4, the latter of which can't use any tools

[elfin birch]

It's currently rolling out randomly

[humble hamlet]

You got it?

[elfin birch]

Mhm

[earnest sierra]

its real slow tho

[vague robin]

What do you think of a ChatGPT live stream. A GPT model that responds in real time with video, audio, and/or text.

[hidden ibex]

Make a GPT reaction stream of ChatGPT watching AI Seinfeld on Twitch

[unique loom]

will gpt 4 turbo go on the web version too?

[lost tundra]

HI all — apologies for the undoubted repetition, I'm a Plus subscriber in the USA and don't yet see all the new features from yesterday, is this just me or are others still waiting too?

[novel parcel]

same

[unique loom]

it's being rolled out slowly

[deft night]

This is a limitation I didn't think I would hit. GPT4-Turbo preview can do 128k tokens. But I can only send 10kTPM. 😞

[lost tundra]

Thanks @unique loom @novel parcel

[unique loom]

np

[novel parcel]

np

[lost tundra]

iOS app already has the new stuff

[brisk path]

Android app does too

[novel parcel]

I tried to make gpt assistant in playground but its not working good

with files

[paper kestrel]

What kind of files did you upload and why do you say it is not working good friend??

Is uploading files fine tuning it?

[novel parcel]

I uploaded json files in the assistant

[cyan sun]

Why does the iOS app have a mixed mode, but when I use it to generate an image, it can’t generate one successfully?

[paper kestrel]

What do you mean by not working good??

[novel parcel]

its not using the infos in the files

[paper kestrel]

Must it be formatted in a specific way for it to work properly?

[novel parcel]

im not sure i dont think so

[stuck holly]

im trying to ask chat gpt to optimize a prompt for itself and it keeps just returning the response for the prompt I want it to optimize. I've tried to prompt it letting it know I was going to send it a prompt, but then it forgets and sends me a response to the prompt and I tried just a sample "optimize this prompt for chat gpt 4" and it again just returns the answer for the prompt. Is there a way to get it to optimize a prompt I give it?

[novel parcel]

maybe there is nothing to optimize

maybe says improve, fix, adjust or change

[peak sequoia]

i can't use web browsing on the android app... it says that every modell is now in one modell, but that's not true...

[tardy totem]

So, does anyone know if chatgpt will be getting the improvements, or if they only go to the API stuff?

[elfin birch]

ChatGPT is getting a new UI, GPT-4-Turbo, All-Tools, and document reading

[tardy totem]

nice! Can't wait for turbo

Can't wait for gpt4 to work without crashing halfway through output as well. 😛

[compact bluff]

I'm assuming we won't get the 128k content

[elfin birch]

Nah, we got that

[eternal rock]

`` Server Tip: ``

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

[compact bluff]

Hmm last time I asked gpt it said it had only 8k input

[tardy totem]

I think the function call stuff is going to help a lot in not needing huge input arrays

[elfin birch]

It doesn't know about itself, it just came out yesterday

That is a hallucination

[compact bluff]

I'm taking non api

[tardy totem]

Europe will need a few weeks as usual, eh?

[elfin birch]

I'm talking about non-API as well

[elfin birch]

Didn't see anything about it 👀

Hope for the best, prepare for the worst

[tardy totem]

Well it's not here yet.

[outer seal]

my guess is 32k for plus / 128k for enterprise

[elfin birch]

Still rolling out like for everyone else

[tardy totem]

Right now I feel more eager for gpt4 to just work for me, than the new stuff. 😁

[outer seal]

the elusive alpha alltools version is 32k ctx

[elfin birch]

Not the case, queried a 95,475 word book of copyright entries and managed to give works of various genres across the entirety of those 95,475 words which is approximately 127,300 tokens

I have Plus, not Enterprise

[outer seal]

oh when it fully rolls out i mean

[elfin birch]

Why would it change? They said they are releasing GPT-4-Turbo to ChatGPT and I have it, they delivered

[patent mauve]

is it just me or this turbo is a bit slow, it outputs the text a bit more in a choppy manner sort of like a stutter

[cobalt idol]

You can now create custom versions of ChatGPT that combine instructions, extra knowledge, and any combination of skills.they still didnt learn to not say "now" when its not out yet for everyone? xD

[tranquil grove]

lmao

Coloring Book Hero
Got it! I'll create a coloring book page featuring two Space Marines on a simple battlefield without any enemy creatures. I'll keep it child-friendly and easy to color.

I'll get to work on it now and let you know once it's ready. If you have any other specific requests or preferences for the design, feel free to let me know.

User
You
I'm waiting!

GPT
Coloring Book Hero
I apologize for the delay. I'm currently working on creating the Space Marine battlefield coloring book page for you. I'll let you know as soon as it's ready, so please hang tight for a little while longer. Thank you for your patience!

[cobalt idol]

he will let you know with a ding 🔔

[tranquil grove]

lol

GPT wants me to click on a link, but I have to manifest it real first!!

User
You
So...??

GPT
Coloring Book Hero
I apologize for the delay. Here's the Space Marine battlefield coloring book page for you:

You can click on the link above to view and download the coloring page. I hope you and any young coloring enthusiasts enjoy coloring it! If you have any more requests or if there's anything else I can assist you with, please feel free to let me know.

[zenith falcon]

It searched the file, it didn't memorize the whole thing

[pine tulip]

No

Rolled out to me yesterday already

[tranquil grove]

^ (TR)

[tardy totem]

Where are you at?

@pine tulip

[pine tulip]

Spain

And germany

Dunno what location they saved

[tardy totem]

I'm only in Sweden.

[zenith falcon]

It's random who gets it

[tardy totem]

Makes sense I suppose.

[pine tulip]

Yea but worldwide

[pine tulip]

It wasn't in updates before

Us was first always

[tardy totem]

Over the last few hours I've tried "Regenerate" six times, and it fails partway through the code output.

It's getting slightly frustrating. At this rate I might actually have to turn on my brain again.

seven

[pine tulip]

eigth

[tardy totem]

At least it's not just me.

[pine tulip]

Nah i don't even use it rn

[tardy totem]

oh

[pine tulip]

oh

[tardy totem]

I'll see if 3.5 can teach me something about architecture meanwhile

[elfin birch]

What would it search for then? It would have to comb through the entries and whether it is of a certain genre
How else would you like me to test it?

[novel parcel]

When will it get rolled out?

[elfin birch]

It's getting rolled out right now

[tardy totem]

Is it possible GPT4 uses different algos for code generation and other stuff? When it generates code for me it is suuuuper slow, but not when I just ask normal questions.

[cursive path]

It seems code generation still uses OG GPT and it’ll use turbo for anything it can get away with

[hidden ibex]

the consensus in some communities is that gpt4 is a mixture of experts, so the delays are likely inference routing to different models

[cursive path]

I wonder if we will see less turbo usage when the inevitable articles come out with benchmarks showing Claude 2 being smarter than turbo 4

[brisk path]

Wdym

Turbo literally is supposed to be smarter

Probably just placebo since you'd think that the faster model is worse

[cursive path]

I don’t believe that for a second. Used it a bit already, like $10 in API fees already and it’s most definitely not smarter

[eternal rock]

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.

[brisk path]

Well...

Claude 2 is stupid anyways

There's no reason for a model that slow to be so dumb

[tardy totem]

Claude doesn't seem to have any european support.

[cursive path]

I don’t use Claude 2 so I can’t comment on it lol

Just have seen the benchmarks

[elfin birch]

Then how can you say GPT-4-Turbo isn't smarter than Claude-2?

[novel parcel]

I could use it one time with vpn i think

[maiden fractal]

For other ai I suggest to use the proper channel #ai-discussions . I am not a mod, but before getting a warning by a mod

[rancid otter]

I don't think OpenAI has a more advanced competitor tbh. Not even google can touch them

[novel parcel]

not yet

[tardy totem]

I've tried a few engines, and nothing competes with GPT4 so far.

[novel parcel]

they will release gemini

[zenith falcon]

Could give it raw, not a file

[tardy totem]

I think GPT4 is nicer to work with for coding than that thing Github has even

[brisk path]

It's still relevant to ChatGPT in a comparison sort of way

[elfin birch]

Can't paste in that much information

[brisk path]

Benchmarks don't mean anything

PaLM 2 does consistently better than GPT 3.5 in benchmarks and...

[rancid otter]

everything google publishes is on very nice papers, but not real apps. im not optimistic about gemini given their slow pace. openai crushes everyone every ~6 months

[novel parcel]

mhh

[maiden fractal]

Not really, since this is exclusive chatgpt channel. I refer to rule 8 #server-rules

[brisk path]

That's literally half a year

A pretty long time

GPT-5 could start and finish training in that amount of time

[tardy totem]

I just realised ChatGPT is better at architecture when I ask it without referencing programming.

[fleet pumice]

Is the log in feature down for any of you ?

[tardy totem]

So there's probably some truth in that it uses different models under the hood, depending on what you ask for.

[tardy totem]

It took me a few attempts

[fleet pumice]

they seem to be on a heavy load, i'll try a bit later

[rancid otter]

also its almost 1 year since ChatGPT* and google still cant compete

[tardy totem]

My guess is they are taking down clusters in order to prepare for putting up new ones.

[brisk path]

Ba

Or at least

Bard compared to free ChatGPT

Unless you're one of the lucky souls that have 4 all tools without paying a penny

[tardy totem]

I've tried bard a bit. It seemed to be even better at understanding what I meant, but worse in generating responses.

[brisk path]

Recently?

[tardy totem]

Not in a few months

[brisk path]

It's gotten way better

Or at least compared to it's previous state

[rancid otter]

it's still inferior to gpt4, but yeah it improved

at least there are less hallucinations

[fleet pumice]

Bard is worse than gpt4 for coding

When I tested it a few months ago

[brisk path]

Well duh

It isn't a big multimodal moe model

[tardy totem]

It still is kind of bad for it

GPT 4 rocks

[astral storm]

Are people stil lwaiting for multimodal all tools and gpt4.5 - or just me?

[brisk path]

Turbo isn't 4.5

Its more like 4.1

[astral storm]

oh okay - i mean turbo. Is it wideley rolled out now, or not so much?

[brisk path]

I think I have it

But not completely sure

I do have the April 2023 knowledge cutoff

[obtuse talon]

When are costum GPT's available for plus users?

[forest shore]

Been getting this message today "message in conversation not found" a lot. Every time I send a chat, this is the response.

[astral storm]

yep - i have that cutoff too. But assumed I hadn't got turbo yet, as I have not seen fireworks and confetti or anything - everything still looks the same

[feral zinc]

within this month

[pine tulip]

and if not, (maybe) this year

[feral zinc]

yep

[brisk path]

That's because it's literally GPT-4 but faster, cheaper, bigger context window, and slightly smarter

[novel parcel]

I got this

ChatGPT can now browse the web, analyze data, and generate images.
These capabilities are now built into GPT-4, no need to select them. Available for Plus users.

[tardy totem]

Will we get some notification when we are upgraded?

[rancid otter]

i only got a popup notifying about being able to upload other file formats

[serene wren]

Yall havin an error come up when providing images to GPT4?

[novel parcel]

same

I got new ui

[sharp mason]

Anyone else consistently getting errors with Dall-e 3 in the new ChatGPT?

[astral storm]

Just noticed that on my web UI I have these modes - Default / Browse with Bing / Advance Data Analysis / Plugins / Dalle 3. But in App I only have Default / Browse with Bing / Dalle 3. So is Plugins and Data Analaysis not available on ChatGPT App?

[serene wren]

Guess so

[tardy totem]

The list when you hover over gpt4 should be gone, right?

[strong breach]

EST, NEO. Just got my update but it says GPTs is not available yet but will come inthe "coming weeks" 😭😭😭

[novel parcel]

same

[tardy totem]

I've got the GPTs stuff in the API

but nothing new in chatgpt

[strong breach]

♥

[eternal tulip]

Just got the update. So, what's the new token max for the new model? 8k? 32k? 128k?

[static yarrow]

hello does the chat gpt turbo version is already out ?

[pine tulip]

yes

[eternal tulip]

It's accessible in the playground.

But, how many tokens is the "new" chat model where the plugins like data analysis and bing search and dalle 3 come automatically?

Really???

[wicked nova]

I really wish, they add more voices to ChatGPT, the current voices are somehow low pitch. It’s hard to understand low pitch voices. Wish they add a high pitch female voice!

[near summit]

Anyone know where chat history has gone?

[novel parcel]

is there

[near summit]

ohhhh the sidebar is almost invisible now, that's why I couldn't see it

[novel parcel]

on web

[eternal tulip]

128k context! https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo says 4096 output tokens, but the 128k awareness means it is able to remeber so much!

CanI get a link to verify that? Some people on reddit said they tried putting 120 k tokens in chat and that it woudn't "go through" ? Where they using the newest model? Just want some clarification.

[full dagger]

So, not everyone can access Custom GPT's yet correct?

[tardy totem]

128k tokens should be enough for a whole book, should it not?

roughly speaking

[fading cloud]

got it pdf reader but cant read excel file now?

[novel parcel]

it's always saying there is an error while generating an answer

[flint stirrup]

Got the web update for ChatGPT Plus. Is the web chatgpt using got4 turbo? With 128k context?

[eternal tulip]

That's what I am figuring out too.

[remote flame]

Do you have access from this very moment?

[tardy totem]

I'm slightly more than slightly annoyed at this stuff now

"There was an error generating a response"

[flint stirrup]

Yea, the web interface is updated but not sure is its 128k

[novel parcel]

same

[remote flame]

Ok, thanks!!

[novel parcel]

gpt creation will be available in the coming weeks xd

[full dagger]

thx

[sharp mason]

So weird. I can create dall-e 3 images in the Dalle GPT but not in the main ChatGPT interface.

[shut mauve]

i have problems with chatgpt where it keeps failing and saying "Hmm...something seems to have gone wrong." is anyone else experiencing this or is it something wrong on my end?

[tardy totem]

It keeps failing for me too.

[novel parcel]

bro what is this Message in conversation not found

[dreamy flare]

omg i just got access

niceeeee

[novel parcel]

can you create gpts

i can not i only see the openai

gpts

[hollow spade]

Anyone else getting this? Or just me — logged in, "Oops!
Our systems are a bit busy at the moment, please take a break and try again soon."

[novel parcel]

try again

[hollow spade]

@novel parcel did. cleared cache, differenet browsers. And the https://status.openai.com/ page shows no downtime

So I'm wondering if anyone else is affected similarly

[dreamy flare]

yea i can

[novel parcel]

it says coming soon