chatgpt-discussions

[honest jay]

Still having the issue on your computer? What browser is having the issue? Do you have a VPN installed?

[bleak mica]

the thing is that it recently appeared, my custom gpts worked just fine and then boom it happened, havent changed a thing since it worked.

[bleak mica]

Yes I have, I have google and no I dont have VPN even installed.

[prime swallow]

Understood. A lot of people who report a problem say it happened suddenly.

It's great that you can access from another device, that gives clues about the problem, and also means you have some way to connect while you figure out a fix

[honest jay]

Can you test in a new install of Firefox?

[bleak mica]

ye, I'll guess I have to wait and see

[bleak mica]

ive already tried edge and google incognito, do you think firefox could work?

[prime swallow]

You can try that. However, if it has been happening for more than a day, it may not fix itself

[prime swallow]

Yes. That was the last of my tips too

[prime swallow]

A great many people find that a fresh install of Firefox fixes their issue, if they were not running a VPN program.

[bleak mica]

specifially firefox?

[honest jay]

Edge and Chrome and both Chromium-based browsers, so they often share issues if one of them has one. Firefox is "non-Chromium," so it works a little different and sometimes better for folks, like Esk mentioned

[prime swallow]

And a fresh install at that

if you can get it to work on a fresh install of firefox, that really helps confirm that it is an issue with the browser, and not other things

[bleak mica]

will definitely try this 👍

will update

I've now tried with firefox and also with changing network but didnt work, still same error. I can say that its only for the custom gpts, gpt3.5, gpt4 and gpt 4 plugins works perfectly

its just the custom gpts.

and to clarify, its working to communicate through the gpt builder playground somehow

[quiet mulch]

Is there any documentation that someone could point me to on creating actions in the new custom GPT editor? I have tried to create a few and they all fail, right now even the demo ones fail....lol

[honest jay]

And is it all custom GPTs, or just one in particular?

[bleak mica]

All

[gritty thicket]

When will OpenAI open plus account again?

[honest jay]

Well that's definitely odd. I've heard of someone having a bugged custom GPT before--what it was doing was breaking the message send function any time the user tried to use it. What they found though, was that when they simply copied the instructions text into a brand new GPT and deleted the old, it worked, almost like something got bugged with their one GPT that they couldn't fix, but a new one was all clear. But, you're saying it's on multiple GPTs, so it's a little different. But maybe you could try with just one, and see if it happens to get through? Unfortunately, GPTs are so new that there's not a lot of common troubleshooting to do on them yet, so you sharing your experience and troubleshooting attempts is very helpful as we all wrap our heads around it!

[bleak mica]

Yes, I will try to copy the instructions from my current and create a new and paste them and see if it works, thanks for being patient with me! 🙏

[full veldt]

Tbf there do seem to be various issues around GPTs that I think will get ironed out in coming weeks - for instance atm they seem to eat through the 40/3hr limit faster than normal GPT prompts (I've not actively attempted to prove this yet, but it feels much faster)

[carmine gust]

gpt plus still pause upgrade 😅 when is open again

[west tartan]

Yeah definitely been buggy.

Tbh Bing has kicked Plus butt since April.

[lime anvil]

Is there a channel on this OpenAI Discord server where I can share the new GPTs I'm creating? I have just created one that uses expert inference reasoning based on a method proposed by Google to improve the reasoning of GPTs, by the way here is the link in case you are interested: https://chat.openai.com/g/g-AqATm5g2y-decision-expert

[honest jay]

#1171489862164168774

[lime anvil]

Thanks my friend

[elfin birch]

Doing that can land you in hot water

[candid ravine]

has anyone encountered chat-gpt just completely screwing basic maths up

past 2 hours i haven't gotten one right answer in multiple chats

[silk phoenix]

It has been just awful all day for me, with any prompt that requires some form of context and analysis (math, basic reading questions)

[candid ravine]

75501.54 - 75000 = 5001.54

is one answer i got. My brain has been fried for past few hours because of this

[silk phoenix]

I’m just gonna give it a day and hope it all gets back to normal tomorrow or by Monday. It has been performing very poorly for me as well

[candid ravine]

Yeah me too. Normally is pretty accurate but something seems very messed up about it today

[silk phoenix]

I asked a very very basic reading question, thinking that maybe math isn’t its forte so it will do better on text. Nope. It went horribly wrong. Then I gave it the same prompt again. It gave a different answer, still wrong 😂😂😂

[honest jay]

GPT and LLMs in general can't do math without external tools like ADA or Wolfram. It's just not how LLMs can work, they're not calculators and they can't calculate.

[silk phoenix]

Same! Normally it does pretty well for me as well. Today though📉📉📉

[blazing belfry]

wait thats not right? 😆

[silk phoenix]

Welllll… it didn’t do so well on rudimentary text comprehension either 😬

[candid ravine]

As dyslexic as i am and i would have thought the same, no it isnt but it looks like it is if i read it quickly haha

[blazing belfry]

i know its not but damn...at first glance it looks like it could be

[candid ravine]

Someone spilt their lunch in the server room again im guessing

poor servers on fire probably haha

[candid ravine]

Ill let chatgpt have a rest it does a lot for me. Today is an off day for the poor fella. Probably just cant handle the tomato soup an intern just spilt on the server racks

[blazing belfry]

even chatgpt needs a rest day?

[candid ravine]

Might even just be a few beers deep. It is friday after all, cant blame even AI for needing a breather

[silk phoenix]

It has happened for me before too. It just randomly performs poorly on a day then it goes back to normal the next.

[blazing belfry]

its going to be slow and hungover tomorrow morning now

[honest jay]

but yall forreal math is just a "no" for llms. they're only trained to predict text not do math.

[silk phoenix]

I see that, and I would usually agree - if it at least performed decently on text 😫

[candid ravine]

Great we have grouchy Chat-gpt tommorow

[silk phoenix]

Why I mentioned “by Monday 😉”

[blazing belfry]

chatgpt is going to have bings attitude

[candid ravine]

came here for AI discussion left with the idea of an AI being hungover

All round informative chats guys hahaha

[blazing belfry]

lol

[silk phoenix]

But it’s killing me! I keep sending the same prompt over and over again and keep getting a different wrong answer 😭😭

[blazing belfry]

i messed around with sql queries today all day and chatgpt will less helpful than usual

i think i set a record for dropping the f word at work today

[silk phoenix]

Same here!!! 🤣🤣🤣

I was screaming, crying, cussing

not to mention, the cutback to 40 messages

[blazing belfry]

oh damn i didnt notice that

dude

[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.

[blazing belfry]

i didnt send anything bad

[candid ravine]

maybe the b word

[silk phoenix]

Yup 🥲 hate to be the bearer of bad news

[candid ravine]

someones work network just lit up like a christmas tree to that word i would say hahaha

[blazing belfry]

there all better

[candid ravine]

The mod Gods will be happy now

[blazing belfry]

if it helps the speed and quality ill accept on 40 messages per 3 hours

[jade musk]

Happened to me as well since yesterday. GPT hardly getting a good response even for coding related stuff like it used to be.. so frustrating.

[silk phoenix]

At this point, I’ll accept it as well if it gets back to its previous accuracy and quality. I don’t even care for anything better than now. Just give me back what I had 😭

[supple zephyr]

Is it 40/3h now?

[silk phoenix]

Yes ://

[supple zephyr]

oh

[dapper stag]

I'm not kidding I've switched to GPT-3.5 for the first time in months. It's way better at following my custom instructions and doing creative writing. The new GPT-4 model is horrendously generic in its outputs regardless of the custom instructions

I would at least like some acknowledgement from OpenAI, or some confirmation that they will continue to look into and improve GPT-4. At this point I'd take the 25 message cap again just for a competent GPT-4

[blazing belfry]

i can top that

i sometimes use bard now willingly

[dapper stag]

I haven't used bard in a while but I should try again some time

[silk phoenix]

At least we know that it is not individual instance and is now rather common. It’s understandable, for today. With all that has been happening, their priorities are probably not the model/source code

[silk phoenix]

I did that too 🤣🤣🤣

Then I went back to gpt 3.5

[dapper stag]

True and they're having issues with model cost. I'm sure it'll continue to improve I just hope they're aware of the issues lol

[blazing belfry]

bard is okay, but it sometimes just makes things up or guesses

[silk phoenix]

I expect it will get better as well. But I just hope it won’t be too long 🤞🏼

[silk phoenix]

Bard truly angers me, even when it is right, it still angers me 😅

Also, does anyone know what the peak hours of usage are?

[jaunty harness]

How in the world do I gain access to the voice generation?

[winged moth]

You mean the functionality whereby you can speak to Chad GBT and it will respond audiably? You have to buy chat GPT plus subscription and you have to use an iPhone or Android app. I know it works in the United States but I'm not sure about other countries right now. I'm pretty sure it's available internationally.

[jaunty harness]

I'm in the USA, on android, with a plus subscription and still don't have it

I don't have an opt in option

[balmy gorge]

Are you referring to text to speech API? : https://help.openai.com/en/articles/8555505-tts-api

[jaunty harness]

The whisper tool

Where I talk to ChatGPT and it talks back

It is

[blazing belfry]

i cant go back to googling

i wont do it

[carmine lagoon]

I am getting consistent bugs: "Message in conversation not found"

This is happening in multiple GPTs for multiple conversations

I can only generate one response, it loses it, then cannot continue the conversation

[errant pebble]

what is googling? idk 💀

only knowing gpts

[eternal ruin]

Sam Altman been fired?

[blazing belfry]

lets start a gofundme for sam altman to begin working on a new ai model

[lost lagoon]

Good luck with that

[floral portal]

apparently

[elfin birch]

that jawn will cost so much money it's baffling

[burnt wraith]

you don't think he has a non compete?

[blazing belfry]

your prolly right, he prolly just receieved a massive severance and hes all set

[junior badger]

Non competes are non enforceable in California and especially don’t work when firing someone

[burnt wraith]

except for trade secrets, which chatgpt no doubt will be considered as

[junior badger]

Domain knowledge !== trade secrets.

Again you can’t fire someone and prevent them from using their expertise elsewhere

[burnt wraith]

In California, including San Francisco, non-compete agreements are generally unenforceable. The state has a strong public policy that favors open competition and employee mobility. This is codified in California Business and Professions Code Section 16600, which states that every contract that restrains someone from engaging in a lawful profession, trade, or business is to that extent void.

There are a few narrow exceptions to this rule, such as:

Protecting trade secrets (though this doesn't extend to preventing competition per se)
In connection with the sale of a business or dissolution of a partnership or LLC, where the seller agrees not to compete with the buyer
Because of California's laws, employers in San Francisco and throughout the state are limited in their ability to restrict former employees' professional activities after they leave the company. This stance supports California's reputation as a hub for innovation and entrepreneurship, particularly in the tech industry and Silicon Valley, where employee talent moves frequently between competing firms.

[junior badger]

Cool now ask chatGPT how that applies to Sam Altman no longer being allowed to work in AI

[burnt wraith]

AI models can be considered trade secrets if they meet certain criteria. According to the Uniform Trade Secrets Act, which has been adopted in some form by most U.S. states, a trade secret is information that:

Derives independent economic value from not being generally known to, and not being readily ascertainable by proper means by, other persons who can obtain economic value from its disclosure or use.
Is the subject of efforts that are reasonable under the circumstances to maintain its secrecy.
In the context of AI, if a model has been developed that provides a competitive advantage and is not known to the public or competitors, and the company has taken reasonable steps to keep it secret (such as through confidentiality agreements, security measures, etc.), it may well be considered a trade secret. This can include the model's architecture, training algorithms, data processing methods, and the datasets used to train the model.

[junior badger]

The basis of that is the literal AI model

Not domain knowledge

[burnt wraith]

need to ask lawyers if this is considered "domain knowledge" if chatgpt is beating every other LLM by a wide margin

[junior badger]

Regardless it isn’t illegal, they would have to sue and whoever he goes to next will have a significant legal defense budget

[burnt wraith]

i agree with that

[wooden rose]

I hope altman being fired doesn’t damage chatgpt development

[frosty plover]

i agree

[fiery wasp]

theres a bunch of theories on the twitter/internet about why he was fired so suddenly im still not sure why

[twilit lintel]

Why he was fired ?

[fiery wasp]

all we have is this as an official statement

https://openai.com/blog/openai-announces-leadership-transition

[plain crane]

sigh chatgpt can't even generate images and sam altman getting fired is just making the day worse.

[winged moth]

Sorry to hear that. You tried the usual tricks of uninstalling the app logging out of the app etc and no dice? No little earphones icon shows up for you?

[shut karma]

"he was not consistently candid in his communications with the board"

[unique flax]

i will wait for the next big annoucement of openai to see if openai is a strong company that can survive this huge storm and keep a direction about attaining agi at all cost

[wraith dagger]

gpt down?

i keep getting something seems wrong?

hm weird ok

[still loom]

we want sam altman back

[burnt wraith]

5 senior researcher gone

[leaden island]

Which ones

[sterile forge]

Source?

[warm sphinx]

Yea, Sam launched this. Why would you toss him away. Microsoft jumps in knee deep and a month or two later, he's out? Weird.

[burnt wraith]

lookup john victor on twitter he linked the source

[tardy oasis]

boo bunch of paywall links

[lost veldt]

Has the GPT-4 usage limit been changed to 2 hours? I don't know how much I used, but the available time shown is after 2 hours

[nova fable]

Hmm. I just update the android app and now gpt4 is the only option. No options for changing type or 3.5

Am I missing something?

[muted marlin]

Anybody linked ChatGPT4 to ComfyUI at all?

ChatGPT is a victim of its own success. The servers are groaning under the weight of usage; and it'll be a few months until they expand capacity!!!

[sonic comet]

Not few months, just few weeks if Microsoft fully supports the scaling process.

[muted marlin]

Hopefully! 🙂

[dense halo]

hi. university gave me openai api key for like 2 weeks and told me to use however i want. how do i use that key to chat in playground?

[muted marlin]

I have the same question: how do I use my API Key inside ComfyUI? 🙂

[sleek swift]

The limit is lower than said I think

I haven't talked 40 messages and I already hit the limit

Yeah, but it is really annoying. Maybe I counted wrong when checking.

[last void]

We hope so 🤞

[full veldt]

Is this on the default GPT or on a custom one?

[sleek swift]

I used a bit a GPT, but mostly standard GPT 4

[full veldt]

Interesting - I nearly always use custom GPTs (I have ones set up for various tones etc) and assumed that they ate into the 40 quicker than just vanilla standard GPT 4, but if that's true then maybe not

[sleek swift]

I do have a custom GPT with a couple of PDFs for a bit more data about competitive programming, but mostly standard for any other questions I might have

[jaunty harness]

Yep, tried all of that. I even tried leaving beta

[winged moth]

Wow. So in the app, when you look at the settings you don't see an option for setting the voice? I can't remember if that was something I had to do the first time to get the headphones icon to show up...

If that doesn't help, I'm not sure what's going on. The only thing left to do would be to reach out to open AI support via their email I guess?

[muted marlin]

My ChatGPT4/DallE-3 "Hot Mods" (Experimental) just plateaud!!!

Says I can do more after 12.12 p.m.

[jaunty harness]

https://cdn.discordapp.com/attachments/970442288276852736/1175396272232996955/Screenshot_20231118_062450_ChatGPT.jpg?ex=656b1415&is=65589f15&hm=e8f378da4c3e84d0e8cfb6b4d00a12073d6466219d51dda6d3853a504190d801&

Unless I'm missing something

[feral zinc]

if you are using a custom dns server or some other ad blocker/privacy thing then disable it

[winged moth]

Yeah it looks okay except you don't have the voice option.

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

[jaunty harness]

That did it

Oh, it might be because I have the IP header location set to another country

My ip isn't changed, but if that's what they base it off of

How do I make it generate audio?

Holy crap, everything just changed. nevermind it's obvious now.

[winged moth]

Awesome! Glad you have it. I'm really enjoying it! Hope you do as well

[jaunty harness]

Nice, hopefully I like it

[bleak mica]

it now works, had to do exactly as you said, copy instructions to new gpt, thanks for the help!

[honest jay]

Sure thing! Thank you so much for confirming that worked for you, really helpful to know!

[full veldt]

They really need to get the GPT-4 limit sorted because atm it definitely isn't 40 - at least not for me

I just hit it despite having barely used GPT-4 in the last three hours - and after counting my inputs I've only made 15

[warm gull]

Yeah it's super super frustrating, especially testing gpts

[maiden fractal]

Every prompt have a 3h timer

And failed gen, or some decline requests count

[cloud pewter]

my gpt 4 got rolled back

to jan 2022

[loud night]

cha gee tea pea

[full veldt]

There were no failed gens, or at least none that came through to me

And I included any requests where I got "Sorry I can't help with that" as a response (but I don't think there were any)

[ashen fjord]

@full veldt what are we paying for? A usage cap and a less usuable platform

[cloud pewter]

anyone else who got their gpt4 rolled back to Jan 2022?

[ashen fjord]

@cloud pewter I have to wait 2 hours to check.

[normal quarry]

My training includes information up to April 2023. This means that any events or developments occurring after that date are not within my current knowledge base.

[cloud pewter]

that’s not fair at all

[honest jay]

You may just be experiencing the model hallucinating. It's done this before around times where the knowledge cutoff was updated in the past. You could try starting a new chat (because it might believe its own hallucination otherwise) and asking for some specific info between Jan22 and Apr23 as another test

[cloud pewter]

completely dead

“The winner of the 2022 FIFA World Cup has not been determined yet, as my training only includes information up until January 2022. The tournament is scheduled to take place in Qatar from November to December 2022.”

[honest jay]

And this is in a new chat, with GPT-4 selected (not GPT-4 Plugins)?

[cloud pewter]

ok I changed to the classic version then it can answer who’s the winner but it still says that it’s cut off is jan 22

it can’t answer anything besides that world cup thing

weird

ok I completely don’t know what’s wrong with my gpt4, I asked about tears of the kingdom and it can answer what it is

[odd flint]

Voice not working properly after today's android app update. After the ChatGPT android update 2 hours ago, the new app most of the time can't recognize when the speaker has finished his phrases, and keep eternally on the"listening" node. I would be careful to not update, until this could be carefully addressed, verified and solved. May be just an overload problem, a coincidence, or may be related to the latest version. In each case, please test on another device before updating. I tested on two devices, Samsung Galaxy S20FE, with the same results, unfortunately. I hope it's not a bug, but just the overload on OpenAI system. 🙏👍

[bleak mica]

You could maybe try to see if you have some special custom instructions that is restricting it from answering

[ripe mesa]

Guys please make a feature where someone can select one or more chats via a checkbox to delete them. It is madly annoying to delete every chat manually and to only have the option to delete everything is also not the best solution.

[versed shadow]

Is 40 messages / 3 hours normal for everyone now? I thought it was just doubled to 100, now it's less than what it was?

[normal quarry]

it's 40 per 3 hours now

[versed shadow]

Alright, thanks

[honest jay]

Idk if it'll be any faster for you but on desktop you can use the keyboard shortcut Ctrl+Shift+Backspace to delete a chat
Also #1070006151938314300

[gritty plover]

It changes

[versed shadow]

Based on what?

[feral zinc]

server capacity. it was not 100 recently and no such thing was announced

[ripe mesa]

Thank you 🙂 But it would be way faster if you could just select the chats via a checkbox where you can edit them (delete, rename etc.).

[ripe mesa]

okay will check it out

[normal geyser]

Im hoping with Sam Altman gone GPT4 will return to what it once was as right now it makes things up and isn't very smart compared to how it once was. Bye bye Sam.. see ya!!! Go hang out with your antisemitist buddy Musk who was also once fired from OpenAI for also making bad decisions.

[brisk path]

Sam brought OpenAI to what it was today

Without him there wouldn't be a GPT-4 in the first place

[wooden rose]

oh another anecdote about how GPT4 has gotten worse 🥱

[normal geyser]

Because it has and it is good to make it known so that they can see. The problem with Discord is that it is filled with 12 year old fanboys that are only here to show that they are 12. Do better, Joel 🥱

[brisk path]

You guys say that GPT-4 gets worse every update, and yet you still pay 🥱

If anything, it's getting better for me with Turbo

[near fractal]

is gpt-4 on chatgpt using turbo or the old version, it seems to be a bit faster

[brisk path]

Turbo

[warm gull]

novelty is wearing off

[dull ice]

I am not able to get GPT-4 to draw simple schematics anymore. ChatGPT
! Error analyzing v
It seems I'm still unable to access the schematic drawing tool. However, I can provide a step-by- step guide to help you draw the schematic using an external tool or manually:

Does anyone know why? Is there a plugin that can draw simple schematics?

[stray peak]

I know some ish is going down at OpenAI, but would it kill you guys to make 'folders' and sortable docs for ChatGPT? It's a core missing feature, that would make the experience 10x better.

[odd flint]

Today's update on Android app, the Voice audio is not working anymore for me. When it works, the audio is so low quality than doesn't worth a pity. We keep paying each month in hope that it will eventually improve. But everything has a limit. I hope soon we get great competition on the AI market, perhaps Grok or other people make a new discovery on the AI field, and we'll have a new home. 👍

[brisk path]

Grok can't even compete with Bard ffs

[dense halo]

i havent tried grok but im on the waiting list because i love to try new installments in ai

[stray peak]

I wish Claude had a voice chat, Claude 2 is VERY good. In many ways better. (but not in all ways) ie) not as good as analysis/interpreter

[brisk path]

Would be useless

Free Claude 2 has a horrible usage limit

Also Claude 2 has gotten substantially worse ever since usage limits were added

Wait, actually no

I feel like just the web version is a downgrade from the API

Reasoning and just overall smartness here has always been not good

[narrow galleon]

it's gonn be a wild world when the businesses running the world have adopted AI fully, abandoned majority of human staff. and then when these ai glitches fail the business, it'll cost a pretty penny to grt a human to fix it

[brisk path]

And tbh I just feel like Claude 2's "creativity" is just scaling up the temperature

It's stories are usually less coherent and satisfying than what GPT-4 can create

[gaunt citrus]

I only used 13 prompts and 40 rights are finished? How this is even possible? I'm using custom gpt. Is it using more rights? Anybody knows anything about this?

[honest jay]

Both "Create" and "Preview" messages count against your GPT-4 cap--were you using both of these features? Or were you just using an existing GPT?

[wooden rose]

because it has
Prove it. My own anecdotal experience tells me that it feels better than it used to be. Why should I trust yours?

[gaunt citrus]

Let me check

[honest jay]

I don't think either "Create" or "Preview" history is saved--just ephemeral content for either instructing the model what to put in "Configure," or testing the GPT, respectively

[gaunt citrus]

Can I send photo in here?

[honest jay]

No, but you can in #community-help and #off-topic I think

[gaunt citrus]

I have 4 docx as knowledge, it has the capabilities checked as web browsing dalle and code interpreter. Maybe because of these checks it uses more rights?

Because it only writes 150 words of text for my each prompt.

I mean I only used 13 times and it says your cap is full

[honest jay]

Are you building the GPT at the screen where it's split half and half? Or are you just using the GPT?

[gaunt citrus]

I already created that and 6 hours passed from the creating process. I started using in the normal message pannel but using the gpt that I have created

I'm in shock actually, if this is real I'll start an anti plus ads from my social media accounts. This is literally unacceptable. 40 right is already low and in 13 prompts it finished, its like a joke.

[honest jay]

Got it. So just to clarify: 6hrs passed with no ChatGPT usage at all, then you came back and only used the GPT you're describing, nothing before using any other GPT or the base GPT-4 model? And you only used the GPT about 13 times? You didn't edit & resubmit any messages, or ask it to regenerate any responses?

[sharp pilot]

The problem is that if they allow discussion in one direction but not the other, that's an inherent bias, and for laypeople and outsiders like us it's almost impossible to share an opinion that does not include speculation.
At the end of the day it's just too hot a topic to discuss on the company Discord when it is pretty strictly moderated even though it is clearly relevant to #openai-chatter , we will just have to wait until the news cools.

[gaunt citrus]

Exactly.

You summarized well. There has to be an explanation right? I mean its not right, its literally fraud.

I'm already a busy person and I do not have to lie in here, especially about the chat GPT. I'm in love with it but this is...

[honest jay]

It certainly sounds confusing, I'd be surprised too if that happened to me. Unfortunately I don't have any other insight to offer, other than to suggest that you consider making a post in #1070006915414900886 describing this behavior, and maybe offer a chat share link (or at least the chat ID at the end of the URL) so someone at OpenAI might be able to take a look. Just be sure to clarify that you believe your usage cap should have been totally reset, and that you only sent 13 submissions in total.

[gaunt citrus]

Oh we missed a step. As I said I only used 13 inputs nothing more, no continue response click or re create response click. But I used all of them in new chats. 13 different chat page, 13 of them has only 1 inputs. My gpt has 4 docx files and browse, dall e, code interpreter are checked. Maybe because of 4 docx files, when the gpt is trying to read and remember them, it used 2 message right at the background? Maybe because of the lengths of the docx? I mean 4 docx files has 3500 words in total.

[honest jay]

Ah yeah those may be relevant details as well, thanks for clarifying. I haven't heard about user GPT-4 usage being potentially doubled, tripled, etc. based on the complexity of the request before, but I suppose it might be possible in this kind of multimodality. Total speculation on my part, but I agree that it would be good detail to include if you post in #1070006915414900886

[gaunt citrus]

I have 2 other plus accounts. I'll try on them with the same way, if I get the similar results, I'll let you know and definitely talk in the bug reports about this nonsense.

[gaunt citrus]

By the way thank you for your time.

[honest jay]

Absolutely, happy to try to troubleshoot. I'll be curious to see if you find an answer to your question!

[velvet geyser]

Someone know if custom gpt will be able to create gpt 3 turbo or something like that with unlimited messages

[feral zinc]

no it cant, not even close

[velvet geyser]

🤷‍♀️

My question can also be "if gpt 4 will have unlimited messages someday" too

[feral zinc]

for plus users possibly not for the next 1-2 years if ever

[gaunt citrus]

Yeah, exactly 13 prompts and it got usage cap.

[storm wave]

why does chatgpt feel like swearing some times and doesnt other times

[past hare]

I am building an AI assistant that can translate code, explain code, debug code and optimize code. can anyone help me to optimize the prompt in each option that I apply so that the model can understand well

[velvet geyser]

When i was using dall e and normal gpt 4, i was like "what, i got 40 messages that easily?"

It's weird

It's like i used 15 prompt and hit cap limit

[solemn nexus]

Custom GPTs have not been working all day. I cannot save the GPT.
The preview chat doesnt work - just says something is wrong.
I only have 3 GPTs and one is locked.

Tried logging in and out, changing browser, restart etc. Same thing.

Any help appreciated.

[celest quartz]

When I regenerate the answer, how do I get back to the previous one in the mobile app??

[maiden fractal]

not sure if possible, but you can log into a mobile browser and take the previous answer

[celest quartz]

Somehow it lags A LOT on a browser

[frank plume]

@past hare send me a DM, I can help you sort it out. I’ve been working quite a bit on code generation for ChatGPT

[oak flame]

Good evening, I remark that after a certain amount of time writing, ChatGPT stops and Im getting "network error". Is there any way to avoid this since it regenerates all the prompt when retrying ?

[gentle edge]

yo I am gonna admit I am extremely slow when it comes to all this AI and plugins and all that but is there any way I can use different GPTs people have made without gpt plus ?

[halcyon maple]

Hi
I need to draw a Statistics circle and some Percentage with it
Did there is an ai can help with that because the most of them only for image

[honest jay]

Hello! Do you have a ChatGPT Plus subscription? If so, I'd recommend using the Wolfram plugin to ask for your desired stats/% thing, and I bet Wolfram could handle it from there. Alternately, you may get what you need using just plain ol' Wolfram Alpha on its own website, but if you want to ask about that we ought to move to #off-topic

[halcyon maple]

Thanks

[tight island]

the model already understands it well u just oughta prompt properly
the issue is on your side not the models, its also highly dependent on what coding language u mean. theres a lot of em.

[unborn basin]

The only thing I see on the website is "How can I help you today?" and I cant do anything, how do i fix it

[honest jay]

Hello, do you have a VPN installed?

[unborn basin]

No

[honest jay]

What internet browser(s) have you tried in so far?

[unborn basin]

I only use google, but Ill check on opera gx

[honest jay]

I believe both Chrome and Opera GX are Chromium-based browsers, so if it doesn't work in Opera GX, you might try a new install of a non-Chomium browser like Firefox.

[unborn basin]

It works on Opera GX but not on Chrome, im not sure why

It was working fine last night

[honest jay]

Ah alright, nice troubleshooting. Do you happen to have any browser extensions installed in Chrome that might affect ChatGPT? Either an unofficial ChatGPT extension, or some kind of VPN/adblocker extension?

[unborn basin]

I have an adblocker, but even when i pause it on the site its the same.

[red karma]

Does anyone know when it will get back to 50 messages for 3 hours? 40 is way too less

[atomic orchid]

chatgpt is down for me right now

It just says unable to load history and says upgrade to plus, won't load any conversations, even though i have plus

[honest jay]

Hopefully whatever the issue you're having with Chrome fixes itself soon, I'm not sure what else it could be. Glad Opera works in the meantime

[honest jay]

Loading normally on my end. Do you have a VPN installed by chance?

[atomic orchid]

oh yeah, I just got nordvpn, but it's not connected

well, i accidentally logged in with it

is that bad?

[eternal rock]

`` Server Tip: ``

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

[honest jay]

In addition to disconnecting the VPN, can you try disabling specifically NordVPN's "Web Protection" feature? This may need to be disabled separately, even if you already have the VPN itself disabled. Then try again.
Web Protection is located in the shield icon "Securty Score" section in the bottom-left of the desktop app. Let me know if you're on mobile

[atomic orchid]

it's working now. Thanks a lot

[honest jay]

Good deal! You're welcome.

[silk phoenix]

Okay. Yup. I’m definitely not using 40 messages per 3hrs (maybe 20?) and am still getting cut off

[waxen river]

They fired sama and they are slowing things now

Wait to see they change lots of stuff soon

[elfin birch]

This has been a thing before Sam's departure

Message caps have been fluctuating recently to see what works

[waxen river]

40 seems to be new low.

[silk phoenix]

That would make more sense. I’m certain I’m not using 40(I actually counted)

[honest jay]

It was 25/3hrs for a long time

[silk phoenix]

I think it is the same now

[sleek gust]

any suggestions for a custom gpt?

[silk phoenix]

Just messaged you 😊

[hybrid violet]

Using recursive loops with the API is insane for refactoring code. Actually scary how efficiant this could make systems w/out the need for human involvement

***From: ***

def generate_primes(min_num, max_num):
    # Create a boolean array "prime[0..max_num]" and initialize all entries as true.
    # A value in prime[i] will be false if i is not a prime, else true.
    prime = [True] * (max_num + 1)
    p = 2
    while p * p <= max_num:
        # If prime[p] is true, then it is a prime
        if prime[p]:
            # Update all multiples of p as not prime
            for i in range(p * p, max_num + 1, p):
                prime[i] = False
        p += 1
    # Since we are using the sieve algorithm, we no longer need the is_prime function.
    # We can directly return the list of primes.
    primes = [p for p in range(max(2, min_num), max_num + 1) if prime[p]]
    return primes

# Generate and print all prime numbers between 1 and 100000 using the optimized sieve algorithm
prime_numbers = generate_primes(1, 1000000)
print(prime_numbers[:100])  # Display the first 10 primes found```

***To:***
```def g(mn, mx):
    p = [True] * (mx + 1)
    for i in range(2, int(mx**.5) + 1):
        if p[i]: p[i*i:mx+1:i] = [False] * len(p[i*i:mx+1:i])
    return [i for i in range(2, mx + 1) if p[i]]

print(g(1, 1000000)[:100])```

Prompt: Please Refactor this code until it is in its simplest form. This should result in the same exact outputs printed to the terminal as the original codebase. Please do this by using the least amount of tokens possible. It does not need to be readable by humans but should produce the same output as the original code.

[wise anchor]

was there an update in the last couple days to the model? i feel the responses are less robotic now they feel more relatable and human

[waxen river]

they are always updating it nowadays.

they do not tell small changes.

people finding out while using.

[elfin birch]

It would be nice to see these changes in https://help.openai.com/en/articles/6825453-chatgpt-release-notes 😳

[waxen river]

I wish.

Also some people get random updates some don't.

I saw couple people got memory update. it says chatgpt will remember everything we say.

[elfin birch]

Yes, might be an alpha thing just like all-tools was!

[tardy girder]

thats really nice

If on sign up for the enterprise version how long is "get back to you" time?

Know someone that did it weeks ago and still waiting

[honest jay]

They're currently only working with large companies on Enterprise accounts.

[granite harness]

I really don't like the lack of transparency in these rate caps. I should have something on my screen at all times telling me how many posts I have left and when that buffer will refresh.

[tardy girder]

i assume with the "large" you mean in the thousands?

Has a roughly 100-150 licences company any chance

[granite harness]

I pay a consistent amount of money for a service that seems happy to vary how much I get to use it on a whim without telling me.

[tardy girder]

Im still waiting to be allowed "to pay"

[granite harness]

😭

[tardy girder]

Its like either i get access through enterprise version, if they get around to giving access

or by waiting for the waitlist

[honest jay]

I am not sure on that, there has been other user info on this before though from folks who may be better informed. Try searching on this server "enterprise." I recall reading that it requires a fairly sizable annual contract deal, for example.

[hexed condor]

even after you pay most of the time it's not working

[tardy girder]

Ye i was expecting it to be prioritiy for the largest companies

[tardy girder]

oof

[bleak mica]

wait what, just now when I started the website I saw "ChatGPT Alpha" instead of ChatGPT 3.5 🤔

[tardy girder]

ye said that for me also earlier today

[scarlet dawn]

@tardy girder I know in the summer it was 70k (per month? year? idk.) for a dedicated API instance afaik, the unlimited usage with ChatGPT enterprise suggests its the same thing or similar

[full veldt]

That seems a reasonable ballpark - I work in IT for a reasonably large firm (probably big enough to where if we were in a more closely related industry we'd consider getting an Enterprise account) and most of the costs for our "core" contracts are roughly that order of magnitude

[midnight fossil]

Hello

[loud night]

mornin! well good evening since its now night but yee

[hard dew]

Tf was GPT-4's source of training material when it comes to basic English, the urban dictionary?

[hushed sparrow]

does anyone know why using GPT Builder eats into our GPT-4 interaction limit?

[honest jay]

Because it uses GPT-4 in both the "Create" and "Preview" interactions. Fwiw: "Create" just automates the filling of "Configure" and doesn't establish anything "behind the scenes" for your GPT. So, you might try just using "Create" a bit less, and do a little more manual tinkering in "Configure" to save yourself some usage. "Preview" will always count.

[hushed sparrow]

doesn't really seem fair, expecially with all the errors

[limpid rampart]

is chatGPT asking anyone else every few minutes to verify that they are human? this is annoying

[honest jay]

Do you have a VPN installed by chance? Are you on Linux?

[limpid rampart]

novpn, windows 10

thanks for suggesting those though

[honest jay]

What internet browser(s) have you tried in so far?

[limpid rampart]

I'm in Chrome - this is how I have used chatGPT for months. it just today started doing this

[desert jewel]

Does anyone know when/why the cap was changed to 40?

[honest jay]

It may not be anything you've changed--network configs can be finnicky sometimes. Do you have any browser extenions installed in Chrome that may be affecting ChatGPT? Otherwise, can you try in a new install of a non-Chromium browser like Firefox?

[limpid rampart]

Thanks for the suggestions I was really just wanting to know if it’s happening to anyone else. Sounds like it’s not

[honest jay]

Usage cap is adjusted for server load

[desert jewel]

Does this change dynamically?

[honest jay]

It's hard to say if there are any "behind the scenes" things from time to time, other than listening to subjective user experience. But for the most part the most "authoritative" source we have is what it says next to "GPT-4" in the model selector on the web version of ChatGPT, which currently says 40/3hrs

[desert jewel]

No worries. Thanks for the info!

[wise anchor]

Lol are they trying to get altman back now?

[loud night]

I have no clue what’s going on anymore

I got worried so i tried to set up a local LLM and now i see why no one uses local llm compared to the cloud

I can’t even send it images to help me get the info I need and it didn’t even work at all

[tender glacier]

Seems like it

[tawny crater]

Why did they reduce the cap without even announcing it. That is the most unprofessional thing ever. Lol we are literally paying for a service.

[wise anchor]

Wait what?

[sage jacinth]

yep, i noticed that too

[loud night]

oh ye they scale it depending on server load

[wise anchor]

are you serious? 40/3hr now? not cool

I’ve never once seen the cap change

[tawny crater]

This is so annoying. What is the point of paying for a service if they are going to do that.

[covert chasm]

OpenAI is in an impossible situation where they can't scale fast enough to keep up with demand.

[tawny crater]

That's not our problem, we're all paying customers, and they should be announcing a change like that, otherwise it's stealing.

[sage jacinth]

yes they can, just hire more ppl

[pine tulip]

Suffering from success

[covert chasm]

They can either rate limit use or they can take it away for everyone except enterprise customers.

[covert chasm]

Ehh, have you read the TOS you agreed to when you subscribed?

I get it, I hate the rate limit too. It's BS, but they have limitations due to, ya know... Hardware limits.

[sage jacinth]

You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 5:21 PM. Learn more - they said 5:13 PM, just 5 minutes ago

[tawny crater]

Ugh, another OpenAi white knight. I don't have time for an argument. Apparently we can't have opinions

[pine tulip]

yoyo, I be promptin all day, how bout you fam

[ripe raft]

Bro are u an ai

[pine tulip]

As I an AI I can't tell you whether I am an AI or not

[brisk path]

It kinda depends

[covert chasm]

🤣🤣
Like this is going to be a productive conversation. I know exactly what you're doing. I'm just going to sit here and watch you say silly things.

[brisk path]

Not gonna delve deep into this

Just gonna say that both sides are wrong in their own ways

Israel is definitely a bit worse tho but this war didn't need to happen in the first place

[covert chasm]

I get it. I really do. I hate seeing the rate limit just as much as you do.

[ripe raft]

Do you know what is a postulate

[wise anchor]

i agree to an extent, but at the same time it’s like “you’re paying for this and we are reducing your ability to use it because of hardware problems, but you still have to pay us the same”

like i get that it’s not feasible for them to scale this fast, however nothing changes for customers other than less use of the product we are paying for

and yeah it’s in the tos but that’s redundant to me when it comes to company to customer communication

[tawny crater]

Exactly.

[covert chasm]

@wise anchor you can use the API version and have a significantly higher limit.

I have a ChatGPT bot in my discord server that I can't even max out myself if I tried.

It's also dirt cheap. API fees are fractions of pennies per 1000 tokens in and out.

Even heavy usage over a month is less than a dollar.

[wise anchor]

I’m talking more generally about consumers who are paying and want to be using the chatgpt interface exclusively

There are people who are subbed who don’t even know what an API is lol

[hard dew]

Can confirm, I was subbed for ages before I decided to acknowledge the API's existence.

[covert chasm]

That's the best part. I had no idea how to setup the chatbot because the dev didn't have clear instructions on their GitHub.

Guess what I did? I used ChatGPT to walk me through the process.

[pine tulip]

🤯

[covert chasm]

Yep. No joke. It'll walk you through everything down to the tiny details, like setting up a virtual machine in Azure or Google cloud, to error checking during any parts of the process.

[pine tulip]

There should be a GPT for that!

[sage jacinth]

we training a gpt, does that also take away credits from the 40?

[pine tulip]

Yup, chatting with the configurator also counts. But you can open the configure tab and add the instructions yourself

[limber current]

Anyone here who tried writing fictions with chatgpt?

[past hare]

@frank plume

can i use multiple prompts for different options and handle the request according to the api request

[loud night]

this chatgpt discussion is wild

[dense halo]

Yes

[limber current]

What kind of prompt techniques do you use for fiction writing?

[zinc valley]

What have yall been doing on got

[pine tulip]

game of thrones?

[zinc valley]

Chat gpt has made me hella $ not even gonna lie

[cloud pewter]

when can I get my April 2023 back

look I paid

but it isn’t giving me the latest model

[placid wigeon]

Hey quick question - I have filled out the form for the higher usage cap; would that going to actually give me higher cap or is that just a form that everybody writes?

[cloud pewter]

does that form even exist

[acoustic talon]

Ok so even if I'm using it for scholarship, education and teaching, which all fall under free use, I still can't input copyrighted files in the "knowledge" section of my custom GPT?

Also, im not trying to argue, I apologize if it seems like that, im just seeking to understand

[karmic tundra]

Fair use has limits, you can’t for instance photocopy an entire book and hand it to your class

Smaller excerpts are fine per fair use. I imagine the same would apply here

[cloud pewter]

when will openai give us an explanation

of model roll back or something like “hallucination”

[burnt wraith]

i'm getting erroneous messages/flags over data analytics questions, did they change something over the weekend?

[prime swallow]

I am not a lawyer, so I don't even know for sure. I'm just a community member, like yourself, so you should not take my ideas as facts. I do quote actual sources, but what those sources mean is up to people with more authority than me.

Also... if you pick an interpretation that is determined to break TOS... it's your account access at risk, not mine. Me, I take no chances I can identify. I value my account access highly, more highly than any personal reason to break TOS or allowed content rules.

As I read it though, if your input to the AI violates anyones' rights... you are breaking TOS.

Will that come back to bite you? I don't care to take that risk. Do you care to take that risk?

If you get permission from X to use their IP, you still could potentially get caught by moderation software that doesn't currently exist that might be used in the future.

So, my compromise, if I had to upload IP that I have permission to use this way...

I would upload my proof of permission to use it and then the IP that I had permission to upload.

That way, if OAI decides to consider what the heck I am doing and what they should do about it - they already have the full story, as verifiable as possible. I've already been completely honest.

If they suspend my account for any reason, I can appeal but that's gonna take perhaps weeks?... months?... longer?

I'm not interested in risking that, myself.

[sage jacinth]

this message cap is sus

[acoustic talon]

I see

[cloud pewter]

I’ll take a wild guess

they do not have enough powerful gpus to run the latest gpt4 so some of the “lucky” users had to use the older version

[acoustic talon]

Fair

[cloud pewter]

and I’m one of the “lucky” ones

[pale owl]

anyone else using chatgpt to give them food recipes

for some reason i am still skeptical of its advice but it worked for the chicken marinade i did lol

[hoary gazelle]

i thought "subway" was typing

was starting to worried about the recipes thing

[hoary gazelle]

next week "new deluxe chicken marinara sandwich "

[pale owl]

i was going to say something but got banned, rip

[pale owl]

first: HAHAHHAA
second: i am going to use chatgpt to tell me how to cook pasta with ground beef

i have the ingredients tbh

[hoary gazelle]

hahaha i saw you typing a long string and got deleted haha

[hoary gazelle]

uhhh nice, let's see how it goes. 😄

[pale owl]

haha thanks

it's practically my assistant chef

[hoary gazelle]

so cool. I have to try this too. thanks for the idea 🙂

[pale owl]

yea just give it some ingredients and ask it what you can do with it or buy with it

"i have olive oil, lime, and italian seasoning (specify oregano, salt, pepper, turmeric, etc) or goya adobo. what meat should i buy that would taste good with these ingredients?"

ik some people probably take pictures of their fridge and have chatgpt tell them what ingredients they need hah

[hoary gazelle]

amazing. this thing blows my mind everyday.

lazy people taking pictures haha, at least talk with the machine! Eventually would be like "chatgpt, go open the fridge and prepare something"

[stone helm]

Still getting the issue where it will just go "oops.. etc"

[pale owl]

oh god

[slate berry]

source?

[stone helm]

hb @slate berry 🍰

[slate berry]

?

[slate berry]

ohhh, happy birthday. Thank you so much!

[stone helm]

of course

[burnt wraith]

his & gregs twitter and wld rallying + prior bloomberg article

[idle sleet]

我现在能是用你吗

[sudden latch]

how long does chatgpt plus waitlist usually take?

[severe gyro]

bing censoring got worse

can't emphasize things as easily

[subtle surge]

Yeah it’s actually unusable now

And with dall-e 3 I was hyped to work on a project that gets to use it but with the censorship getting out of control I am forced to use my local stable diffusion🤧

[burnt wraith]

same with chatgpt, everything even mentioning a term (even if legal) get's flagged as against openai tos

[subtle surge]

They need to chill with the censorship, it’s literally making it unusable

[delicate sable]

Does anybody hit the message limit with GPT-4? I noticed it's now 40 messages / 3 hours

I haven't hit any limit since I subscribed 🤔

[plucky timber]

Whys my newsfeed all plastered about you guys

I did learn one thing tho again don't hire people who get paid to delete your job.

(board of directors)

[tawny crater]

yea, i sent a ticket to openai, apparently they are not aware of any issue on "their end". and asked me for a screenshot, when i had already sent them on in that chat. i swear they don't even read what we say, the probably just have an AI respond to us.

[plucky timber]

Them : Heres our updated TOS do you agree? Yes or Yes?

How about you add that little feedback button that is normally under chatgpts replies lmao

[delicate sable]

It means more money spent on collecting, aggregating, and analyzing the responses, then probably hiring someone to build a dashboard with the analysis, or if there's already a way to build dashboards with some in-house tools, someone will have to take his time to build them.

Then after the dashboards are there, someone, or a team even, has to look at them, understand what is done, what the users want, like, or dislike, then decide which is feasible and which is not, and probably take action items to adopt user feedback.

Adding a feedback button isn't that easy 😢

[plucky timber]

Thats a lotta excuses to say : We dont need your feedback for our updated TOS

Which is the reality

[delicate sable]

Yeah, pretty much, they already released the ToS 😂

Also, user feedback probably won't affect legal matters that much, FYI

[plucky timber]

Curious tho this news seems to imply new competition might pop up

[delicate sable]

might?

Look around you

[plucky timber]

Right :p

No i think this actually might be the stagnation point

For OAI

Where suddenly bands of OAI ex folks make a new company

I can see it happen

[delicate sable]

Not really, they already have the most money, and they can literally buy creativity and innovation at this point, and processing power

You might have certain features shadowed by other competition, but as a whole package, ChatGPT Plus is unbeatable man

[plucky timber]

Was! But the course of the audience is what picks the budget

Meta had tons of money but is cancelled as a whole

Nobody using the meta idea of meta.

[delicate sable]

Well, maybe it wasn't even the right direction

[plucky timber]

Most people are siding against the board of directors

and with sam, if sam opens a tiny company it will be worth millions tomorrow just out of support

[delicate sable]

The fact that Meta wasn't big enough to pique my interest (I'm quite tech-savvy), means it flopped

[plucky timber]

yeah lmao

[delicate sable]

hmmm

just because they- wait idk if mods are picky about talking about this here

[plucky timber]

Bet they as confused as everyone whats gonna happen

But yaeh better to rest the topic

[delicate sable]

I was about to ask, just because the board members "took that decision" yk?

Do we know more, to decide it was unfair against sam and support him in another startup/company?

Oh, yeah better :p let's drop it :s

[plucky timber]

Mb mb not , new topic : When will ChatGPT stop refusing to use Dall-e

Keeps saying I cant while it says he can on top >.>

[delicate sable]

oh is that new? I generated a bunch of images like yesterday ig

[plucky timber]

Nah its been a while GPT gets drunk and forgets what it can do

Or it will like use python instead 😂

[delicate sable]

actually I find it pretty smart in knowing when to search the internet and when to generate an image. Maybe my prompts are more careful, not sure.
I've had only 1 instance when it executed a search("my internet query back then..." Python method, but then corrected itself and searched the internet after 😄

Seems like the agent behind GPT-4 actually uses similar Python code

[plucky timber]

I think for me it pops up more due to coding related questions and it just assumes we gotta program the image together 😂

[delicate sable]

# I will perform a search to check the latest documentation for ReactiveElasticsearchTemplate in Spring Data Elasticsearch
search("Spring Data Elasticsearch ReactiveElasticsearchTemplate methods")

[plucky timber]

Almost gpt almost *pat pat *

[delicate sable]

Makes me think it uses a pretty similar "algorithm" to deciding which agent to use, and which capability/plugin/whatever they're called, in Python

[shut shadow]

Hello, for a long time, I've been trying to purchase a subscription to ChatGPT Plus, but I can't seem to make it work. Could you help me with this?

[normal quarry]

I answered you in another channel. You only need to post once in one channel.

[tacit stream]

Am I the only one who got his messages capped to 40/3 hours instead of 50?

[normal quarry]

it's for everyone right now

[tacit stream]

is there a reason, to the best of your knowledge?

[normal quarry]

load management, the systems are under instense load so OpenAI are mitigating that with a decrease in the message allowance and pausing new subscriptions

[tacit stream]

I see, thanks for clarifying this 🙏

[normal quarry]

yes

[subtle surge]

I would lmao if that was the case, this whole time

[subtle surge]

Yup, it is what it is but with how hard it is to use right now for creative and personal tasks, they’re setting themselves up for their competitors and open source to offer an uncensored or less censored alternative.

There’s already many small start ups that are trying to achieve just that. But OpenAI at least for the foreseeable future will always have the cutting edge

Wait that’s real?

[toxic apex]

Does anyone knows that if there is a words count limit in Instruction field of GPTs's configure tab?

[normal quarry]

8152 tokens I believe

[toxic apex]

Wow! How do you know that?

Where is the reference? I'm so curious.

[normal quarry]

sorry, I meant tokens not words

[toxic apex]

This question is extended from the Custom Instructions. The Fact field has 1500 words limit, and the Responds field has 1500 words, too.

[normal quarry]

Custom instructions are different to GPT Builder instructions

[toxic apex]

But the limits of Instructions in GPTs's configure doesn't describe well.

So I wondered that if there any words limitation in Instructions? If so, how many?

[toxic apex]

Yes, I know that.

[subtle surge]

I feel you, I have some concept photos to generate. I’m currently taking a crack at Stable diffusion again to see if I can work my way to the same quality

[normal quarry]

#ai-discussions is the place to discuss non OpenAI models

[subtle surge]

Same.

[subtle surge]

Ah okay thank you

I’ll say this, I think there’s no easy way to control AI, like I feel like there will always be abusers. But I feel like there has to be a better way to go about it without rendering the product almost unusable in some cases.

[still loom]

so now openai decreased gpt 4 limit

to 40 messages per 3 hours

why is that

[normal quarry]

to manage extreme load since Dev Day

[still loom]

that is unforgivable

[supple zephyr]

my gpt 4 chat occasionally downgrades into gpt 3.5
how can I fix it?

[normal quarry]

once the model switches to GPT 3.5 you cannot switch back to GPT-4

[supple zephyr]

Is it random? Can I prevent it?

[subtle surge]

Yeah but it’s not just about kids seeing stuff they shouldn’t. There are quite sick people out there making very illegal things as we speak😒

Aside from the jobs argument, everyday people are getting deepfaked and such, it’s too messed up. At the same time my take is that we entered a new world when the public gained access to these AI’s. So I kinda see it as a symptom of something beyond our control. And…human nature.

[subtle surge]

We used to 🙃🤫

Ah I see, yes it’s gotten worse in that front too. Bing Chat is currently unusable. I’ve never seen something so closed tight in my life 🍑😂

[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.

[subtle surge]

Sorry 😂

From my experiments in the early days on paper it makes sense why it’s there but in practice imo it’s poorly being handled in terms of filters

Like if you want real use of stuff you gotta loosen the restrictions. Otherwise usage will just drop like a rock

[supple zephyr]

chatgpt should know this world isn't always so beautiful

[subtle surge]

I was able to get it to describe something for me that was kinda inappropriate but it did it’s job for my story telling lol. One thing that worries me is getting banned for doing my current project because it has to do with war, guns and some violence like explosions, combat equipment and such 👀

[toxic apex]

How do you know the 8152 tokens limits of the Instruction? Where do you find it?

[normal quarry]

I have tested it

[supple zephyr]

I also got something inappropriate too. If I post it in dall-e channel it will take me to instant ban

[frosty plover]

In the future, chatgpt gpts is expected to introduce a similar approach to git version management? Now I'm worried that if I change my Settings, and then the result is not satisfied but forget how to set the original

[subtle surge]

Open ai has a tokenizer for you to check that

Yeah but their argument is that it will still be abused. After seeing the worst of what people have used AI for, all I can say is there’s no point of the filters being this high as people will find a way. It’s either you keep it to a minimum and accept that these bad people will always be there, or you kill AI before it actually gets useful.

🤷‍♂️or just go opensource because that’s what will happen if we’re being realistic

[supple zephyr]

So they will stop being ClosedAI?

[subtle surge]

Really the biggest issues in the short term is privacy and security against new class AI criminals.

But if the future trends to less filters/ uncensored, expect to be fighting other AI agents and privacy to be dead if you can’t even show your face online anymore. And don’t get me started on identity theft 👀😂

[plucky timber]

Aws : We got servers , you got cash?

[normal quarry]

OpenAI deploys on Azure

[plucky timber]

Yes my point

😂

[normal quarry]

I know a company with a $25 million per month compute budget who are waiting to get AWS GPU limits increased, it's an industry wide problem

[plucky timber]

build more pylons

had to make the joke 😛

Anyone noticed a diff cap between general GPT 4 and Assistant?

Somehow I hit the cap on assistant but not on GPT4

[zenith solar]

When I made a GPT based on a ebook I uploaded and ask him questions about the book, it answered vaguely and as I ask for more details, it returns "Message in conversation not found". However, it works pretty well in the earlier time.

That's so weird

[tardy totem]

Does anyone know if there is a faster version of GPT4 available for a higher monthly cost?

And one without a cap! I just reached mine. 😭

[subtle surge]

ah you must be new here, how cute😂

[tardy totem]

Heh, no actually, but I've managed to not hit my cap for a long time, for some reason. 😛

I've been using it a lot lately, and still didn't a cap until today.

[normal quarry]

there isn't a plan for higher usage yet, so a 40 message per 3 hour message limit applies to everyone on Plus

[sharp cradle]

Go with the enterprise plan

[subtle surge]

unless you go to playground

🫰

[tardy totem]

Where can I read about that?

[plucky timber]

https://openai.com/enterprise

Edited 😄

[tardy totem]

Right. I've seen that page. No price information. :/

[plucky timber]

Correct idk either i have a 2 man Microenterprise

[tardy totem]

Same as my company. 😛

[sharp cradle]

https://openai.com/enterprise

[plucky timber]

Just sent them a mail i'll report back here anyways but i'd say shoot one yourself too*

[covert chasm]

I'd suggest API plans instead.

[plucky timber]

See my earlier argument

#openai-chatter message

:kek:

[sharp cradle]

It's custom for every company, what's our estimated usage ?

[plucky timber]

Ah they basically make a custom package for ya?

[sharp cradle]

Correct

[plucky timber]

I guess for me it would be like an increase of maybe 20-30 messages for 2 accounts

So not that much , but we do hit the cap with both accs

WIth like an hour left each time

[tardy totem]

I did the same.

[covert chasm]

Ehh, I think OpenAI is trying to filter out overuse/low effort prompting of "ChatGPT" to level out computing power.

[tardy totem]

They're as slow as the normal GPT

[sharp cradle]

It depends on how many tokens per hour your messages are

[tardy totem]

Exactly the same here

[plucky timber]

I would argue my prompts are decent, I am mostly "Coteaching" it as it struggles with a lot of things (as expected ofc)
But thats ofc biased from me 😄

[covert chasm]

Slow is a relative term. It'd likely also be slower if they didn't have rate limits like they currently have

[plucky timber]

Honestly they could solve it with a pro+ 😄

I'd pay 40 for double the msgs 🤷‍♂️

[tardy totem]

Indeed. I'd gladly pay tripple to get fast and no cap.

[plucky timber]

Yeah I was thinking the same if their offer was in that range

[sharp cradle]

You guys want to get my enterprise api key in that case ? One separate for you with different usage access

[plucky timber]

but for now It seems they might be getting quite a few requests so time for the waiting game 😄

[plucky timber]

Thunder capitalizing :kek:

I love it but ill see their response first ❤️

[tardy totem]

Indeed. Middlemen aren't usually the best option. 😛

[sharp cradle]

Yea it will be gpt 4 turbo with no cap

[plucky timber]

Ngl if OAI keeps capping subs he might actually have a market there

[sharp cradle]

I agree

[plucky timber]

Super low effort market too

Idk what TOS says about it tho 🤔

[tardy totem]

gpt4 turbo is what we have with plus now, right?

I'm sure you can make a contract with OAI that covers it

[sharp cradle]

It just came in my mind right now in the convo, I don't plan to do it commercially

[plucky timber]

yeah no i get that I moreso went full showerthoughts on it 😂

Its not a bad idea if you like vandergroot said can set up a contract

Profit is profit shrug

[tardy totem]

I just got over the time limit for a new bunch of prompts. 🙂

[plucky timber]

The builder eats a lot of prompts i noticed

if you can always type out your assistant yourself

Another tip is Always overspecify , you can have it in your instructions but assistant will ignore it happily

With coding that means every single time say you dont want placeholders

[tardy totem]

My assistant follows that instruction fairly well.

[plucky timber]

Lucky hahahah 😂 mines a stubborn brat

[tardy totem]

the only thing it sucks at is writing markdown files.

[plucky timber]

Does yours actively read its knowledge base or do you have to explicitly ask?

[tardy totem]

oh dammit! It just did the placeholder thing. 😂

[plucky timber]

For some reason mines stubborn in that

[tardy totem]

It doesn't have any files. Just a bunch of directions.

[plucky timber]

I feel like the issue is not so much the placeholder but how it shows me where to place the code

// rest of the code

might be the worst indication ever

Just use one line of the code so i know exactly where your talking about in flash

If it would do that i would be fine with placeholders

[tardy totem]

I'll try adding something about that later.

[plucky timber]

Feel free to ping me the results curious if instructing about that might help 😄

[granite terrace]

I am a ChatGPT plus user. I have 40 messages/3hours for GPT4.

Is this limit dynamic ? My friend has a much higher limit and I’m not sure why

[tardy totem]

4?!

[granite terrace]

Sorry for the typo 😅

[tardy totem]

You meant 40? I thought we had 50.

[granite terrace]

It changed a few days (weeks ?) ago

We had 50 for months though.

[tardy totem]

dislike

I could do with another 40. 😛

[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.

[dense halo]

Why do all the gpt servers go down the toilet when the Americans wake up.

[tardy totem]

Because there's a lot of users there?

[tardy totem]

I sent you the prompt my GPT is using.

[silk phoenix]

At what time did you notice the decrease in performance quality?

[forest hinge]

still no turbo for chatgpt?

[normal quarry]

ChatGPT is using Turbo

[forest hinge]

the output in playground via assistent mode differs vastly

[ionic geyser]

TurboGPT

[dense halo]

Apparently 8 am EST. It’s currently 12 am here

[covert chasm]

I wonder how many servers and GPUs OpenAI is getting from Microsoft and how much hardware they own themselves

[normal quarry]

OpenAI deploy on Azure as far as I know

[covert chasm]

Yeah I figured that much, but heck, they must use entire data centers.

[azure sundial]

With DALL·E, browsing and analysis
Limit 40 messages / 3 hours

Does this restriction apply to "plugin version or GPTs version"?

[normal quarry]

yes, it does

[keen snow]

guys my ChatGPT4 message limit went down to 40 message every 3 hours.. Anyone else noticed the same?

sorry just read the chat, found the same problem with others

ignore my question

[feral zinc]

thats the new limit, for some its already been that for days

[azure sundial]

This restriction should be relaxed, at least to 100 pieces. We are just victims of overuse by users.

[tight island]

One of the other reasons are that people proxy gpt 4 through a VD mode on bots and such instead of using the API so they have a bunch of accounts

[forest hinge]

but every account still pays

[azure sundial]

Some companies will use this method to train their own GPT, which is indeed a problem.

In the early stages, they used the free GPT3.5 API. I emailed OpenAI and suggested tightening restrictions, but did not receive a reply.

[forest hinge]

how do u feel now, being ignored by everybody?

do u still see openai as god?

[azure sundial]

Until now, I have not found a substitute with better understanding ability.

[forest hinge]

but every account still pays for gpt4 so how are they doing bad things to the servers

its not like there are free trials for gpt 4

[azure sundial]

Using scripts to call web interfaces, but I believe that the speed is limited, so theoretically it does not exist.

good night

[amber cliff]

if limits are going back on then the lobotomy has to be removed too. "Here is a file, lets do X to the file", 20 lines of well directed text follow, then, "well if that's the effect what happens if we do Y?" 40 lines of generlisations and possible outcomes follow. "You have the file, I need specfic and actionable information" 10 lines of specific information follow, prompt three times for the complete information, then back to generlisation we go, rinse and repeat. iit is the most frustrating thing! - like contacting the ACC cos its a clear breach of Australian Consumer law and it always ends up making me that angry., (and that after a few days of bliss working with GPT's)

[forest hinge]

web inter?

[tight island]

Idk why most of you never get a response back from OpenAI I always do

[mental birch]

Has anyone had issues saving your gpts? I just made one but the save button does not work

[rugged igloo]

Hi radium!

[spice star]

I found it's an issue with some of the names of my GPTs

[upper rock]

Hey, can anyone more experienced try to help mw with an issue?

(is this the right channel? corry i've literally joined seconds ago)

[normal quarry]

what is the issue?

[upper rock]

i have uploaded some TXT files to my custom GPT (i wanna teach to GPT my RPG system, i'm an author)

sometimes it simply don't read them, o don't follow the rules of indexing

i have also create a pseudocode and an index for better analysys (i have create with the GPT help)

but it simply tell me "he have not explored the file yer" and don't find things

even things maybe have an answer yesterday

[paper kestrel]

Maybe txt files do not work well and other formats are better?

[upper rock]

i have tried word format, but it's worse

when it's workd, TXT give the est and quickest result

[normal quarry]

can you combine your files into one PDF file and upload that to your GPT as knowledge?

[upper rock]

but it became gargantuan

we are talking about an entire RPG ruleset

(ok, nt gargantuan, but big)

the strange thing it's sometime it read correctly, cometimes not

also, define index and ruleset are a waste of my time? does that will be used?

[dense beacon]

why has GPT been so slow since the recent update came out last week?

[paper kestrel]

Too many people using it.

[tight island]

i know what your issue is

you're not parsing your knowlegde

you're basically forcing GPT to waste thousands of tokens just reading the question and the entire ruleset at once and devising an output

you can have GPT Parse your ruleset, This way it does not waste thousands of tokens. if not your basically saying ( Read the entire thing, then find the exact rule which means reading it again, and read that, then devise an output based on X Question ) While parsing it properly allows it to search. plus GPT prefers parsed knowledge. the parsing methods consumes so much less tokens and is way better.

[merry cliff]

You can just copy and paste the txt file and send it as a message. For custom GPT, I don't know because the "Prompt" has characters limit

[tight island]

His issue is the lack of information parsing, i parsed a 403 page PDF into a Json String, and it has no issues what so ever, since it can search the info instead of reading the entire thing.

[merry cliff]

If you upload it as a file, the AI may not read the file

[tight island]

true u gotta reference it at the Instructions. shouldn't take up much space.

[merry cliff]

But if you just copy and paste it, the file is always read by the AI

[tight island]

my GPT has 4 different set of System Prompts. each more than 4000 tokens, U can reference Other Instructions into .txt or .json files. to lengthen your Base Instructions.
Even my system prompts are Parsed to make it easier for the model, u can also ensure it only reads that at some point in the convo based of keywords while ignoring the other.

[merry cliff]

They should increase the upper limit of the "Prompt" input box, for example, Mr. Ranedeer's Prompt has over 13,000 characters.

[stone helm]

Anyone else experiencing the "Oops, an error Occurd" message still.

[maiden fractal]

Do you use plug in ?

[stone helm]

Nope just normal

[lavish oar]

What happened to Sam Altman

[slate berry]

you can read all about it here: https://openai.com/blog/openai-announces-leadership-transition

[maiden fractal]

Can you try firefox as test

[stone helm]

that's what i use.

[tardy swift]

They said he’s not trust worthy

[maiden fractal]

Can you try edge incognito ?

Do you use a vpn ?

[stone helm]

i do not no

never have.

[west hamlet]

hello guys, as a chatgpt plus user i should have a limitation of 50messages for 3 hours right? however i have noticed lately that after around 15 messages i reach the top and it gives me the message that i can use it again after a few hours....is that something new?

[upper rock]

Can it ask you how?

[glad gazelle]

I know the limit is 40 messages / 3 hours, but I have used only about 20 messages for the last 3 hours and it said I already reached the limit and have to wait few more hours. What's going on? Do they decrease the limit again?

[tawny steppe]

what‘s firefox？

[maiden fractal]

Another web browser non chromium

[stone helm]

same thing.

[maiden fractal]

Can you disable custom instruction and do a new fresh chat

[red karma]

Is there any chance it will go back to 50 or will it stay at 40 for a short time?

When it was at 50 it was still enough though

[glad gazelle]

I just tried again and it doesn't give any limit message now, very strange.

[maiden fractal]

Nobody know. But hopefully they will bring back 50 or more !

[round sun]

Is chatpgt 5 going to be worked on despite sam firing?

[maiden fractal]

Why wouldn't work ?

[hybrid violet]

If Sam is not reinstated I'm speculating that there will be a deal made with Elon during the talent exodus.

The more I think about this situation the more crazy it becomes

The outcome of this debacle will solidify who becomes the steward of AGI

[dark jungle]

Sam gets fired and now the limit message is back 🤦‍♂️ way to go OpenAI!

[hybrid violet]

Could Ilya have been weary of Sams gravity towards the industialization of the technology? Hence why this event took place?

I also wonder if the plot of this news arc has any external media controls shaping the narrative.

People have such short attention spans. Attention only needs to be held until the close of this arc for the outcome to be solidified.

[maiden fractal]

we avoid speaking about speculation inside the server. i am not a mod, but before you get a warning or muted

[eternal rock]

`` Server Tip: ``

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

[round sun]

I meant chat gpt 5

[hybrid violet]

ahh gotchya sorry just ranting though ideas I have

[round sun]

On which model?

[west hamlet]

It seems I can use chat gpt 4 but not custom gpt...do custo gpts have another different limitation of usage?

[dark jungle]

ChatGPT 4

[round sun]

Ok

[maiden fractal]

unfortunately yes

[sacred chasm]

Anyone got an idea of when it could end? or at least an average time of how long it took for other pauses to have been sorted out?

[maiden fractal]

unfortunately no eta, i do not have that information

[loud night]

hewoo!

[tawdry basin]

Hi Guys, any idea how to connect a custom GPT to a discord server? I want it to feed on the messages in a discord server.

[tardy girder]

30 hours????

seriously?

[feral zinc]

3h

[tardy girder]

ah ok thank god

[tight island]

gotta go the API Route

[tawdry basin]

Right... I have no idea how that works. No standard schema's anywhere? I've made a python script to link the discord bot to the regular 3.5 chat-gpt. But linking it to a GPT is a different story.

[tight island]

ask gpt thats how i learned how to do it.

[tawdry basin]

You got it working?

[tight island]

this means u need to code the bot with the api inside

yeah its easy.

well i think it is.

[tawdry basin]

My heart just spiked 😄

[tight island]

the alternative is looking for an external API to connect your gpt to especially custom gpt

[feral zinc]

custom gpt's are chatgpt only feature and on the api side they are called assistants, https://platform.openai.com/docs/assistants/overview

[tight island]

Yeah so "my method" which isnt mine.. dosent give u the actual features of your custom gpt.

its just a regular api.

but i think u call do an API CAll for your own assistant if i am not mistaken

[tawdry basin]

I just need it to learn from all the discord messages in my server.

[tardy girder]

hehe ye @feral zinc saved me from a heart attack already

[tight island]

feed it the server logs.

provided u dont have sensitive info there.

[tawdry basin]

Mmm... that's a smart move.

[tight island]

but i'd recommend parsing the final information.

since Parsed Documents work better than non-parsed. what i mean is, a non-parsed document GPT wastes tokens by reading every single thing, But parsing allows it to search for what it needs.

[tawdry basin]

Any advice how to export my discord server information?

[tight island]

havent done it myself, try asking GPT

[tawdry basin]

😄

I will. Thank you

'You've reached the current usage cap for GPT-4, please try again after 9:35 P'

[tight island]

ahh

[tawdry basin]

Great.

[cedar glade]

Yeah, the usage cap seems crazy tight right now... again...

Moreover, lots of calls seem to fail but still count as calls, I think.

[tawdry basin]

And the 'Error saving Draft' message isn't going away either is it?

[swift blade]

Does someone know or have documentation on how to have it so that my custom gpt imports modules and packages into its environment?

[feral zinc]

if you mean code interpreter then you cant as it has no net connection and it has pre set combination of features

[tight island]

theres a psuedo method

What u can do is incorporate existing python code into a .txt document and tell it through instructions to use this particular Code when dealing with X Topic.

usually something like,

Upon user question about: X Topic, Launch Code interpreter and insert 'yourpythoncode.txt' and from that point on keep that enviroment upon until the session ends.

[tepid summit]

Same here for me

[low chasm]

Getting "GPT inaccessible or not found" errors.

Is it broken for anyone else?

[tight island]

working fine for me.

[eternal thicket]

I have come up with a proposal in GPT4 data analysis.

Anyone glad to support my proposal?

GPT4 should support running R codes and more Python libraries in its environment

https://discord.com/channels/974519864045756446/1175923002353913886

[hushed sparrow]

Guys why isn't our history with Chat Builder reported in user data? Even though the data is being used as model training data

[tight island]

by giving GPT more leeway in a python enviroment we'd be able to do some pretty cool stuff

[eternal thicket]

I'm data scientist so R and Stan are also crucial to me

How can GPT4 not support running R codes?

[tight island]

what the.. why cant GPT use CI suddenly.

trying to have GPT use Set amount of words for a page but it keeps saying "Bruv i cant use Code interpreter" but its activated, even a simple Hello World comes with errors.

[timber river]

So when the keynote said chatGPT is coming up to April 2023, did they mean 4 or also 3.5? And is it a gradual rollout or already out?

I checked in with 3.5 recently and it only hopped a year forward

[tight island]

I think it’s because 3.5 will depricate over this current year, webchat wise. But that’s merely a guess

[novel shadow]

while making a custom GPT right now, it wouldn't give me enough time to answer questions about the creation or whether i would like to change anything. instead it kept on going and kept on self improving and it is still doing it now

[jovial minnow]

I'm running into an extremely annoying problem. It has been a few months since I last messed with chatgpt and when I last did, it handily handled things like "don't talk about this" but suddenly it completely ignores the "don't" and is now only talking about the thing. What changed?

[scarlet wind]

Well besides Sam Altman getting fired who knows lmao

[olive coral]

Has anyone used the Saga AI or Creative Writing coach? I'm not sure if my custom instructions will work with it

[honest jay]

If these are both custom GPTs: no, custom instructions aren't enabled on any GPT.

[olive coral]

Ah ok. Makes sense. Thank you ☺️

[honest jay]

Sure thing!

[ebon haven]

Omg a fellow prompt web developer!!

Prompt-web*

I too prompt engineer and inject at every layer lol.

[swift blade]

Any chance you have a video doing this? or can you do a demonstration 🙏

[zinc valley]

where can i find some cool custom gpts

[pine tulip]

#1171489862164168774 is a good place to start

also you can google for them, there are quite a few lists

[zinc valley]

yeah im looking thru them rn they are so cool

has anyone made a dan custom gpt

[pine tulip]

safety bypasses aren't to be discussed here

[loud night]

someone made a gpt that makes prompts for gpts even though the magic editor does it already for some reason

its.. not too useful at all

[zinc valley]

is there any way to have the ai remember the conversations throughout sessions

[loud night]

They were testing it before so we might be getting something like that in the future, but as of now there isnt any easy way, tho you can have it read and write data using the Zapier plugin

[zinc valley]

can i attatch pdfs

[pine tulip]

I could think of a way connecting with an api and connecting to it with your user id or credentials. They wouldn't be encrypted though

[pine tulip]

Yes

[loud night]

you can attatch pdfs to gpt4 in chatgpt yee!

[zinc valley]

if i made a pdf of the conversations i had with chat gpt would it remember the conversation

[pine tulip]

Not "remember" but just read them

[loud night]

I mean if you upload and save them manually ye

the easiest way would probally be using Zapier for now

[zinc valley]

how do i do the zapier thing

[loud night]

Do u have ChatGPT Plus?

[zinc valley]

yes

[loud night]

Noice! I sent a link to some documentation you'll need to make a GPT that does this!

When you make the GPT, you'll want to make sure it can read and write from something like googel docs, then use the magic editor to tell it to write information to the doc to remember things and then read it when sending messages where it needs to remember something and it'll do a thing

[dense halo]

Hey everyone! How can I contact support if I have been charged twice in a row for the same account and only have one month left?

[honest jay]

You could make a post in #1070006915414900886, OpenAI staff are periodically present in there. Are you saying a doubled Plus subscription charge showed up on your bank account at the same time? But you only subscribed with one account? Was this an auto-renew where it had been successfully renewing without doubling like this in the past?

[dense halo]

I often paid on my computer with my card, and once I paid with my phone (apple) and the previous month they took the money once from my card and another time through apple

[wraith ocean]

What is the new chat limit?

[ornate raptor]

40 messages every 3 hours

[wraith ocean]

I actually hit mine for the first time today.

[honest jay]

And you for sure only have one sign-in method for ChatGPT? You don't have like, one with a regular email/password, and one with a Google login or something? Definitely sounds like a good #1070006915414900886 post if not. Also wondering how the pause in new subs/renewals that aren't automatic might be affecting this situation potentially, I didn't think non-auto-renewals were working right now

[molten tundra]

Finally cracked the code on making strong gpts, of course using custom instructions and reference data, but dang so much better

[wraith ocean]

Does any one have a strategy on how to best avoid the chat limit stoping your work flow and being an inconvenience? Or how best to spread out my prompts so I don’t often hit that limit?

[honest jay]

One message every 4.5 minutes and you won't run into the limit with 40/3hrs! Note that all these count as "one":

• sending a GPT-4 message
• editing and resubmitting a GPT-4 message
• asking for a regenerate on a GPT-4 answer
• "Create" chat in GPT builder
• "Preview" chat in GPT builder

[wraith ocean]

Is it set to go back up from here or would it keep decreasing?

[honest jay]

We don't know! They seem to adjust based on server load

[wraith ocean]

I was creating a gpt that’s probably why I hit the limit, when does the 3 hour period start? When you send your first message after having not sent one for three hours?

[molten tundra]

Anyone willing to try out my custom gpt for pixel art? https://discord.com/channels/974519864045756446/1175967602502275183

[honest jay]

It's actually a rolling cap, so you can think of each message you send (or thing you do in that list I gave) as starting off its own 3hr timer. When that 3hrs is up, you "get that message back"

[wraith ocean]

Oh cool I had seen mixed answers about that in forums thank you for clarifying

[honest jay]

You are welcome!

[red karma]

Thanks. I hope so as well.

[dense halo]

I’m bored

[dark jungle]

No one ever consider that ChatGPT is difficult to say, and perhaps it should be called something else, and perhaps you should be able to give it a name .. you know, like Jarvis lol .. GPT is a bit of a tongue twister lol

[celest schooner]

chattyP

[dark jungle]

thats like the old guy in the video from the conference who called it Chatty lol .. i assume you're not the old guy 🙂

[supple zephyr]

"Chatty McChatFace"

[dark jungle]

"Chatty Chatty Bang Bang"

Chai tea ..

[blazing belfry]

chat snippety

(Bing)

[oak sleet]

Yo my chatgpt is buggin out, it wont stop typing: My apologies for the oversight. To ensure that the total probability remains 1, the height of each bar should indeed be
1
1000

0.001
1000
1
​
=0.001. Let's correct this and generate 1000 random numbers with each bar having a height of 0.001.

anyone else having this issue?

[blazing belfry]

its been slow and unreliable for me for over a week

[mighty hull]

damn i'm the only one in my organization and i reached tier 2 💀

[solar pilot]

Can anyone please tell me when can we get gpt vision access? Thank you!

[pale barn]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com. it is popping up repeatedly. I have tried to disable all plug-ins in Chrome(My office computer), delete cookies, and not using VPN.
is there anyone that can help on this matter?

[covert chasm]

Try another browser?

[silk phoenix]

Do you have GPT Plus? If so, you should already have access to vision

[solar pilot]

Yes I have pro. I don’t see it. Is it in Dalle or analyzer?

[silk phoenix]

On the web, it’s the paper clip. You click on that and it prompts you to upload a picture.

[solar pilot]

well I can do that but wasn’t visiom a different thing?

That can create let’s say a html based on your drawing etc

or is it the same thing?

[silk phoenix]

According to OpenAI documentation: “GPT-4 with Vision, sometimes referred to as GPT-4V or gpt-4-vision-preview in the API, allows the model to take in images and answer questions about them.”

[solar pilot]

I’m not sure vision is that. I might be wrong though

[honest gorge]

Theres a guy that did acouple videos on youtube covering GPT-4 Vision with security cameras, pretty neato if I say so

[supple zephyr]

doesn't gpt4 have tough usage limit to run 24/7

[steep umbra]

I believe it's 10k tokens/minute for the newer stuff depending on the model you use. The turbo ones can do much more throughput if you optimize your prompting.

[silk phoenix]

Now I’m interested as well! Is GPT-4 vision different from GPT-4?

I would really appreciate it if you would update me as well if there is a difference

[steep umbra]

GPT-V is another model yeah, it's different from the chat model

[silk phoenix]

Interesting! How can I access that?

[steep umbra]

The easiest way I know is to call the API directly. The model is gpt-4-vision-preview, and there's a page in the OpenAI Documentation called "Vision" under the "Capabilities" header if you want to study up on it. Lmk if you have any questions though

[silk phoenix]

Oh I would really appreciate a step by step. Should I just google gpt-4-vision-preview? I’m so sorry if this sounds very basic

[stark rampart]

Have heard of the gpt-4 turbo!

[steep umbra]

Have you called an API before?

[silk phoenix]

I thought that was what I was using, the one with DALLE and web browsing. Is that different as well??

[silk phoenix]

I am ashamed to say that I have not

[steep umbra]

Nothing to be ashamed of, I only ask to give you better advice. You'd need to pick up a couple of skills to begin using it in its current form. The basic steps would be:

1. Download and install python
2. Create a script and paste the code from OpenAI's documentation for Vision ito it.
3. Upload an image you want to be analyzed or find a link to one.
4. Change the "url" in the script to be a link to the image you selected for GPT-Vision to analyze.

[silk phoenix]

Thank you so so very much! I truly appreciate this (and not judging me). 😊

If I may ask as well, is the turbo different from the GPT-4 with DALL E and browser?

[steep umbra]

Ah it looks like they implemented it in Chat GPT, so you can upload images there too if you have an account.

There's a little file icon looking like a paperclip on the chat box when GPT-4 is selected

[silk phoenix]

That was my confusion as well! That is what I was saying earlier

[steep umbra]

Aha my confusion then, I was overcomplicating the matter.

[silk phoenix]

I feel so validated now! Thank you so so much!!

🖤

[steep umbra]

When you use that file upload, it calls GPT-4 vision under the hood I assume. Their new "Assistants" API supports this kind of stuff.

No problem! ❤️ Glad to help

[dawn ether]

are actions like functions on the api?

they're giving so much to chatgpt but not us API users

[ancient sage]

Does anyone have a screenshot of a 'network error' or 'something went wrong' message? I need it for a bug report / suggestion thing

Its for this suggestion https://discord.com/channels/974519864045756446/1174073462730530837

Just so people can see what those messages look like as I realized I forgot to include that lol

[visual bloom]

Am I only one who has problems of image generation with DALL-E3?

[dense halo]

So why did OpenAI co-founder get fired? Should we be worried?

[slate berry]

we don't have details

[stone perch]

does the iphone version off chat gpt with the chat gpt 4 version support the photo features? like you can take a picture and chat gpt can answer your question?

[vapid swan]

Is it just me or is GPT-4 really messed up after the turbo update?

https://chat.openai.com/share/d35ebaa6-12a7-4c8b-a7e5-61394289c2fe

I can't even get it to generate 20, 5 letter words...

And the way how it makes the same mistakes over and over is not what I am used to from before the update..

[stone perch]

im talking about some homeworks

[stone perch]

perfect thank you

[warm gull]

Yeah, it's not great tbh. But you will get mixed feelings about it on here.

[sonic comet]

Are you sure this is GPT4? These mistakes are similar to the ones that GPT3 used to make.

[analog sapphire]

When and why Plus users got the 50msgs/3 hours cap down to 40msgs ?

Actually hilarious

[feral zinc]

out of server capacity, was done about the same time they disabled new plus subscriptions last week

[normal quarry]

paused

[humble pagoda]

Idk if it's only me but I tried comparing the playground assistants' answers (using the latest gpt4 model) and a gpt from chatgpt (exactly the same assistant) and found out that the playground's answers were much worse than chatgpt's. Is it just me or did you guys also notice the same thing?

[analog sapphire]

Thanks for answering my questions @feral zinc

[forest cloak]

hey is anyone having trouble logging in? I keep getting Oops!
We ran into an issue while signing you in, please take a break and try again soon.

[azure sundial]

Check browser plugin, I also often encounter this problem.

[forest cloak]

ah ok

[ebon haven]

Nope. It's because the board is still committed to their original goals and priorities and Altman had been pushing major things out. He was doing a "ask forgiveness is easier than permission", especially in regards to monetizing the store and revenue sharing. I think he was also doing media stuff without their knowledge. Either that or something pertaining to personal conduct

If anything this is a good signal that the community and overall product development and release speed will be less chaotic and impossible to keep up with

Like maybe they'll stop breaking chatgpt or only release things like Dall-e 3 when they're production ready

[dreamy flare]

Tho i find it surprising that you would escalate the situation like this over such a topic.
I mean this will definitely have heavy repercussions for OpenAI long term.

Especially with others leaving now as well. If Sam actually sets up a new company in this field it would definitely cause a ripple in the industry

[loud night]

oh god the negativity is here too? is there nowhere where I can just be happy and look forward to the future here?

[sacred moth]

My GPT4 limit was 50messages. It decreased to 40 messages now!!!

[maiden fractal]

It is for everyone

[dark parrot]

Anybody else have a problem with GPT when generating lists of things in that it can't count past 10 and reverts back to 1s?

I was making a list of fantasy classes with weapons and it gave me:

"9. Druid:
Nature Staff
Animal Companions
Shapeshifting Abilities
10. Bard:
Enchanted Lute
Flute & Dagger
Charm Bracelet
1. Assassin:
Poisoned Blade
Hidden Blade
Wire Garrote
1. Summoner:
Elemental Orbs
Summoning Scroll
Conjurer's Rod"

And it just keeps numbering things off as 1. after that

I've tried to regenerate multiple times and it consistently does the same mistake

[loud night]

aww

[alpine bolt]

if they were struggling with everyone on board it could be paused until January by the looks of it

[full veldt]

Yeah - this seems to be a problem with how it handles specifically the format of numerical lists

(It has no problem in theory with counting past 10)

[crisp cairn]

he is joining MS so no new company

[ebon haven]

Escalate?

[dark parrot]

Should be easy to fix right? kinda weird that it's still an error that consistently happens

[dreamy flare]

At least one crisis avoided xD

[ebon haven]

Gpt can’t actually count. That’s math and GPT isn’t doing math, it’s processing and language. So if the most likely thing to happen is that a list ends at 10 it might do that

[dreamy flare]

To fire your CEO in such a manner is indeed quite an escalation of events...

[dark parrot]

should I specify that I want 15 items instead?

let's see if that does anything

[ebon haven]

In some respects. From an operational standpoint it’s a deescalation

[ebon haven]

The problem partially is that he was “escalating” the product and company without the consent of the board and had been pushing out unfinished products

[dark parrot]

nope, still ends at 10 and then keeps repeating 1s

[ebon haven]

What manner btw? Their move was very strategic and Sam most likely knew. They did it Friday after market close. That’s a thought out, planned decision

Protecting their shareholders and partners

[full veldt]

I think this may only affect longer texts though - see https://chat.openai.com/share/7e62dd03-93be-4d2f-9d29-616b7085b0f5 as a case where it successfully goes past 10

[dark parrot]

Maybe it's with lists with bulleted additions?

Seems to be the case, especially happens if you ask for lists with bulleted subsections

[sullen ginkgo]

Hi, short question... the speech modus in chatgpt4, is it still integrated in Turbo? I can't find it at the moment and can't activate it too.

[thorn crown]

Hello! I have a quick question. I'm working on developing an app that can automatically generate questions and answers using GPT. Is there a model available that can take a model answer and a student-provided answer as input and output a score?

[odd flint]

Microsoft won. They got what they wanted. They have the servers, the technology, and now, the main people. They rule the cards. OpenAI did the greatest mistake ever. I hope that the company could still at least survive. 🙏👍

[hoary gazelle]

is this the end lads? :/

[obtuse geode]

Microsoft was supposed to give OpenAI $10 billion. If they've got it all back, are they still going to fund the same amount? What's more, I think that out of the $10 billion, they've only given crumbs so far.

[dusky cliff]

oh no

why Ilya? why?

[tardy girder]

Guess Microsoft will be ahead in a short ammount of time

[dusky cliff]

with all the openAI staff, Sam and Greg, sure

they'll even prolly be ahead of deepmind