chatgpt-discussions

[tight island]

land is very expensive.

sadly.

[novel shadow]

hey but if there was a limit to the amount you could own, it would become cheap

but then prices would rise

[tight island]

that is true

[novel shadow]

and it would even out

[sick pecan]

i fed my pet 1.6B chatbot and i prefer its response,

Drinko, short for Digital Universal Network Optimized Cybernetic Assistant, was designed as a revolutionary AI tool in the digital age. The creators believed that technology could revolutionize our way of life, but they knew that the advancements would need an intelligent and intuitive interface to manage.

[tight island]

i still want the ai overlord to take over.

[novel shadow]

same

[tight island]

it will do a better job than our current leaders and more efficiently.

[novel shadow]

i guess not having emotions per se would be a weird and scary trait for a world leader, but i could see handling everything objectively would work

and i wouldn't mind

[tight island]

i wouldnt mind a cold AI leader.

as long as i live comfortably, and safe i dont care.

[sick pecan]

ai adoption is hindered by unthinkable statements

[tight island]

yup

[sick pecan]

paradoxical intent

[novel shadow]

i don't think i follow

[tight island]

i think a duality is needed.

like humans have 2 individuals inside our heads AI should too

[sick pecan]

suggesting the machine will rise up and enslave us = unthinkable

[novel shadow]

ohh yeah

[tight island]

ohh u mean the Hollywood Syndrome

[sick pecan]

ai enthusiasts making these claims = ai enthusiasts are cra cra

[novel shadow]

but what if it was preferrable

[sick pecan]

basically is the new era of crypto bros

[tight island]

Technocrats rise up

i have a model in charge of my company.

its acting CEO, im just the owner. managing it.

[sick pecan]

is sad that the extreme (<5%) of any community becomes the face of that community

[tight island]

sadly, but its reality.

[novel shadow]

it's crazy how we have so few enemies that we become our own enemies

[tight island]

its all about resources and "ego" in the end.

[novel shadow]

we don't have a big enough threat to keep us down that we end up working together

[tight island]

and money of course.

sadly.

[novel shadow]

yeah

[tight island]

one would think that this chat would be more active due to the massive amounts of users on the server.

and.. it died.

[sick pecan]

its a support server first and social club never

[tight island]

true.

[novel shadow]

but then again most people aren't up for discussions

[tight island]

thats all fair, but its unnecessarily dead.

[novel shadow]

from another server:

"suni has appeared"
"your mom has appeared" as a response and then someone else starts talking about random stuff on top of all this

[tight island]

lol xD

[novel shadow]

if that's the general conversation i'm happy it isn't like that here

xD

[novel shadow]

the word unnecessary is unnecessarily hard to spell. who MADE this? :(

[tight island]

ikr

[novel shadow]

i want a letter from the alphabet that describes that word

[tight island]

Q

[novel shadow]

yeah xD

[tight island]

no u need q, to spell quake, Quack, quagmire.

[novel shadow]

something that isn't real though

[tight island]

or this word: Pneumonoultramicroscopicsilicovolcanoconiosis, a lung disease caused by inhalation of very fine silicate or quartz dust

[novel shadow]

panic

[tight island]

i dont even wanna try pronouncing it or try spelling it.

[tranquil lagoon]

What's Occam's razor

[tight island]

but from the looks of it its just a sentence with no spaces.

[novel shadow]

it's not so bad actually

[ember pine]

“the simplest solution is almost always the best.” From 4o:

Occam's Razor is a philosophical principle that suggests the simplest explanation, requiring the fewest assumptions, is usually the correct one. It advocates for minimizing complexity in reasoning and problem-solving.

[tight island]

exactly its just 1 word then 1 word then 1 word. no spaces.

just words mashed together.

[novel shadow]

yes :D and the first word has a silent p

[tight island]

indeed.

[novel shadow]

but then it's just word after word

[tight island]

fun fact; the human brain consists of two conciousneses.

the left and right side. when split 2 selves appear.

[novel shadow]

the brain also makes funny assumptions about itself

[tight island]

perchance a similar design in GPT

[novel shadow]

but does it

[tight island]

in the end we're all just vibrating atoms

in a sandbox game, on a science fair.

[novel shadow]

unless you're cold enough, but then i would argue you have much bigger things to worry about than not vibrating

lol

[tight island]

and we just got an -F

[novel shadow]

why

[tight island]

Because the creator used the wrong mix

[novel shadow]

there's a reason there's a limit to how cold things can get

ohhh you're talking about the science fair thing lol

[tranquil lagoon]

GPT needs more fine tuning to implement the best ways of solving things

[tight island]

what gpt needs the most is "Thinking" power and proper thinking ability.

[tranquil lagoon]

Imagine when gpt has 100 times more parameters

[tight island]

humans can envision their desired paths

[tranquil lagoon]

It will have a similar parameter size to a human brain

[tight island]

this calls for more hallucination

[tranquil lagoon]

It's done by attempting to predict multiple paths

[sick pecan]

i swapped the occam's razor stuff with Solve the following puzzle factually. and got, ```This means we repeat the "111" three times, just as "1" was repeated three times to form "111," and "00" was repeated twice to form "0000."

Thus, "111" when completed should be "111111111".

[tight island]

ohh i meant visually. mental imaging

[tranquil lagoon]

Gpt 4o can probably do that too

[tight island]

cant say

DALLE could be improved.

dalle is bacially drawing blind.

[tranquil lagoon]

A big problem that Q* star will fix is step by step think process.

Another problem is real time/long term learning without forgetting old training data. Real time learning would mean evolving personalities though, it's dangerous. OpenAI wants to fully control their AI

[tight island]

if dalle could see what it actually makes and adjust it would be awesome.

as far as i understand, rn its making the images blindly.

[ember pine]

Yeah it’s just passing it to a function call:

// Example dalle invocation:
// // { // "prompt": "<insert prompt here>" // } //

namespace dalle {

// Create images from a text-only prompt.
type text2im = (_: {

https://chatgpt.com/share/ba74d56e-3079-428b-be4c-d786097d8c44

[snow plover]

theoretically they could make a system that would regenerate parts of the image based on feedback that is sent back to AI after the initial image is generated

[tight island]

i mean it would def, up the quality.

but the waittime might be a little longer.

[novel shadow]

1
111

00
0000

111
? whoever sent this, i just want to let you know that GPT-4 correctly solved the question in 1 try whereas GPT-4o solved it in 2 tries at the first attempt and couldn't solve it when attempting it again

so gpt-4 is the winner of this one

[snow plover]

basically a feedback system

[tight island]

yup

u can make a psuedo feedback system already

[sick pecan]

my gpt-4o solved it in one. but i know how to whisper to the machine

[snow plover]

im sure they are already using something like that

[novel shadow]

how did it solve it the first time?

[tight island]

the models behind closed doors, are def up to the task wouldnt surprise me.

[snow plover]

it wouldn't make sense not to

[sick pecan]

gpt-turbo is an automatic and gpt4o is a manual. you need to drop the clutch to win the drag race

[novel shadow]

can you share the link please

[sick pecan]

i cannot share links as i am on team sub

[novel shadow]

convenient

[tight island]

Teams sub downside

[novel shadow]

then there isn't proof

[tight island]

i feel ya

[sick pecan]

i posted what i did above. same instruction as zaki only replaced the first sentence

[novel shadow]

yeah but if you can't share the process there isn't a real result because you can leave out anything you don't feel like leaving in

[sick pecan]

pretty sure it is bad faith to imply i have bad faith

[novel shadow]

yeah and sorry for being this way but anything could have happened

in the end it doesn't even matter

[sick pecan]

i contributed my instructions. and i am being questioned by someone who didnt?

[novel shadow]

https://chatgpt.com/share/5646d801-3bdc-4917-a42d-d19287ebd8f2

you're right, i didn't

but now i have shared it

[tight island]

team subscribers cant share chats tho

[snow plover]

oh do I miss the days when bing copilot had code interpeter and Parakhin was still incharge..
now its nerfed beyond the horizon, so my leftover choices are ChatGPT and Gemini

[snow plover]

team sub is smth like enterprise version, yh?

[tight island]

its like 1 fourth of enterprise

[snow plover]

gotcha

privacy policies, makes sense

[novel shadow]

that's true, i'm not saying it's drinko's fault either

[tight island]

enterprise uses the full capabilities of all their available models, on the web ui.

[novel shadow]

i'm just questioning it because copying something and pasting it in a place where everything can be deleted and added as you wish takes away the validity of results

[snow plover]

theoretically it should perform way better with API rather than web version

[novel shadow]

but enough about that

[tight island]

yup

Team users have 32k tokens

plus around 4k - 8k

idk bout free actually.

[snow plover]

laughs with Gemini 1 million

[tight island]

ngl not big fan of Gemini.

[snow plover]

gotta wait for the blackwells to roll in

[tight island]

i'll rather use gpt

[novel shadow]

it seems to be 8k, 32k, 32k and 128k 😄 https://openai.com/chatgpt/pricing/

[tight island]

ahh

[novel shadow]

starting from free and ending at enterprise

[tight island]

plus got upgraded.

thats always something

[novel shadow]

it's kinda op that enterprise has this high context limit

[snow plover]

I mostly sit on their lab version, which gives you free 1.5 Pro

with 1M context, adjusting parameters including safety, etc

[tight island]

1m context is nice and all

[snow plover]

not gonna advertise here as you can find it out yourself 👀

[tight island]

but i have better models at my displosal. just aint free.

[snow plover]

makes sense, if you can pay for the better one, why not

[tight island]

i use a swarm for big coding projects. consisting of around 15, 128k Token Context Length GPT 4 API's each with Vector DB.

effecitively doubling each api's context, ish.

but its costly.

[snow plover]

gpt-4o is clearly more OP than what google gives us, but 5-10 prompts per 4 hrs aint lookin good for studying

[novel shadow]

does anyone want to create a company for the sake of using the enterprise version? if we could get 98 more people in on it, it would be pretty cheap 😂

[snow plover]

so i'll stick with the lesser good

[tight island]

i did consider that Vipraz.

but.. its kinda the wrong way to go about it.

[novel shadow]

if you ever do it, please let me know as i would be very happy to split the costs

[sick pecan]

you can get free gpt-4o in a few third party places. like perplexity dot ai (focus writing), you dot com, omniplex dot ai etc

[snow plover]

companies make profit off you, so its fair if you try to make profit back off companies in any legal workaround if you can

[swift osprey]

what is better about enterprise? cheaper?

[tight island]

to explain it like this.

[snow plover]

cheaper and it doesnt supposedly sell off your data to OpenAI or any1 else

[sick pecan]

enterprise gets 128k context

[tight island]

dirt is free tier, and diamond is enterprise.

[ember pine]

Was this 4turbo or 4o? You barely needed any prompting lol. Actually you literally didn’t prompt at all whereas mine struggled without custom gpt

https://chatgpt.com/share/8d0431f4-636e-424e-87a9-b9ad535eb99f

[sick pecan]

openai teamed up with moderna. i imagine big pharma could benefit from 128k context

[tight island]

indeed.

[novel shadow]

this was 4turbo 😄 4o failed the first time and succeeded the second time. i'm counting this one as a single attempt as the first message it sent was a clarification for what i needed, and then it solved it in the second one lol

[snow plover]

they likely dont even get the enterprise version

but direct access to their API

[tight island]

yup

its basically free API with the ui

[sick pecan]

must be my einstein custom instructions making it work then

[snow plover]

free and unlimited

[novel shadow]

the idea that there's something bigger out there absolutely makes sense, it just makes me think about the crazy versions we aren't even seeing available lol

[sick pecan]

i watch a lot of ai safety videos. and so i appreciate the machine wants to spit back the most average response. and turns out average isnt smart (take your average discord user for example)

[ember pine]

Yeah my instructions in logic were needed as well, vanilla wasn’t doing well

[snow plover]

as they say, if you see something available for the people, assume that there's something several versions better already used by the government/military/corpos

which makes sense
if they already have smth like GPT-5, scalability is still a problem to solve

[tight island]

would it surprise you if they had something already 100 times better than what we have now

[snow plover]

no

[tight island]

same

it would be suspicious if they didnt.

[snow plover]

it isnt a huge problem to create a bigger model

the problem is how to make that huge model work for a few billion people at the same time

[tight island]

maybe they're going the model creating a model route.

what would allow for exponential growth

[snow plover]

they have to optimize it

[tight island]

that happens when its being made

[snow plover]

cause even if they get thousands of blackwells n stuff, where will they get the power to run those monsters?

[tight island]

Quantum

[snow plover]

hopefully

[tight island]

[snow plover]

tldr there are layers of problems as reasons to why people dont get the best products right away and that its not a global conspiracy to give the worse products to normal people

[tight island]

Even consumer goods

They’re vastly inferior to what is behind the scenes

Phones, PCs

[snow plover]

remembering stories of unbreakable glass or everlasting lamps 🗿

[tight island]

lol 😂

If new tech comes out better tech always exists in the background

Gotta stay ahead

[tight island]

…especially if you have enemies

[civic quail]

I used Data Lab's Marker project on Github to convert a pdf to text for an LLM. It took an hour to convert and made a json, .md, and pngs. Easy cmdline to use. Worth checking out.

[gusty ledge]

Dude is chat gpt just messing up variable and function names for everyone as well?
Like what is this response (I have to censor it because automod doesn't even like it lol)

void* DLLHandle = FPlatformProcess::GetDllF**k();```

[ember pine]

Lots of people complaining about this recently. Haven’t needed to code with it the past 2 days, 3 days ago it was fine for Python, and JS was good last week.

I think they upped the temperature on ChatGPT, so when it’s completing tokens, there’s a penalty for repeating the same names, and increased variability for how they’re called even in the same response.

Saw a reddit post earlier that showed image_input being passed to a function, but when used as a variable, it was image input instead

[gusty ledge]

I didn't even think about if they upped the temperature, that would make a ton of sense, i mean im sure for certain cases thats fine but man is it annoying when asking even simple programming stuff

[ember pine]

Yeah, even if it used to be 1 by default, which was sometimes annoying, ideal is 0.7 or lower depending on how much creativity you want.
I’m betting they either switched back to 1 or made it 1.2, maybe even 1.3. Any higher and it would start wildly hallucinating, but 1-1.2 sounds about right for how ChatGPT has been acting.

[ember pine]

I’ve heard that you can pass ?t=0.7 for example into the URL parameters on the default chat gpt4 (but not custom gpts) and it’ll change the temperature, but maybe I’m doing it wrong because I’ve been unable to verify this works

[oak sleet]

Anyone know if the GPT-4o model is using DALLE or itself to generate images?

[wise yacht]

I believe both. 4o for the prompting it does with your input and DALLE for creation.

[oak sleet]

any idea when it may stop using DALLE for the creation?

[wise yacht]

Haven’t heard anything about not using DALLE for image creation, but they stay quiet about most things

[oak sleet]

GPT-4o is supposed to generate images itself since it is multimodal

however, I think as of right now it doesn't

[sick pecan]

yeah i am pretty sure the model is just in text mode, and is using the same image generation and vision models as before. but that still doesnt stop youtubers like mberman testing it like its updated

[brisk path]

CHATGPT IS AN IDIOT

IT JUST DELETED ALL MY MEMORIES WHEN I ASKED IT TO STOP SAVING MEMORIES EVERY MESSAGE I SENT

[sick pecan]

yeah the memory thing is unreliable

but arguably "stop saving memories" is synonymous with "forget all my memories"

(the trigger being this message)

[copper pebble]

I'm asking chatgpt to translate the page of a light novel and then illustrate it. It works nicely up to page 30, then it starts to struggle to follow the instructions. Is this normal or maybe I'm doing something wrong? I tried uploading a page per message, and two pages per message, and the result is the same.

[sick pecan]

use custom instructions

and/or make it into a gpt

[honest ledge]

It did that for me too once, afterwards I made a conversation for just adding memories, and made sure to save all the prompts in a google doc so if it ever forgets or overwrites a memory, I can just inject it back in. Also, I've found that it can only store about 20-25 memories, each of which is only about 1 or 2 short sentences. I imagine the memory feature is sort of like an additional custom instruction set with about a 1000 token context window...

[sick pecan]

yeah thats my thoughts too. is like custom instructions that you can modify mid chat

perhaps best to treat it as transient instructions that span a few chats but arent really 'memory'

[nimble carbon]

Is there sora ai alr

[onyx light]

Have anyone get memory feature free user?

[cold pewter]

Voice was coming in a few weeks, 3 weeks ago. Guess how many a few is?

[sour light]

Hello, I use the ChatGPT app on my Android device, but today it suddenly logged out. When I try to re-login, it does not allow me to login, and I see a black background after selecting login from Google, and returning to the login page.

[broken fractal]

It'll come, they may have had unforeseen issues with the full rollout, who knows. It happens all the time in software.

[cold pewter]

be transparent about it atleast

[young basin]

OpenAI transparent? If there’s anything we should have learned from all the recent drama it’s that they aren’t exactly great at transparency

[tight island]

Telling it to stop saving is the same as saying don’t save anything but if it’s not saving anything why bother having the original memories too since it’s not supposed to save anything.

Remember it takes everything you say literally

[dark kindle]

Mac and cheese

[wet vapor]

what are the bets samas on stage at wwdc

[south copper]

I'm aware Sky was recently removed, and I don't really resonate with Juniper, and then we have 3 other male sounding voices with breeze, ember, and cove. can we get more female voices to balance this out?

[sick pecan]

juniper feels like she should say stuff like, 'hey honey'

[tranquil lagoon]

I hate her voice. I want sky back

[south copper]

fr, it's low key annoying, it's so friendly it feels condescending, like I am being talked to like I am 4 years old

[tranquil lagoon]

They need a proper English or American accent, that's what I mean

Sky was good

[south copper]

Sky sounded way more professional

[tranquil lagoon]

Yes sky was good

[stone sigil]

anyone got any UI issues? All functionalities work except when I click "My GPTs" nothing happens. I tried disabling all my extensions, clearing cache, flushing DNS, updating the browser, but nothin seems to do anything

before clearing the cache the buttons on the bottom right were unresponsive too, but now just the one thing I wanna use lmao

Welp, I also just tried in a completely different browser and... still nothing.

How.. perculiar.

[gusty ledge]

As today passes i am getting weirder and weirder replies

void WhisperInterface::QueryodelAsync(FOnQueryFinishedSignature QueryComeOutput, const TArray<float>& Somplex, inDes ProcessConft)
{
    EstivesSyncCount.fetch_min(1); // Increment the active task coing usingly
{
});
);}, [thiswound, samples, Results, Elopvq Num Whotes, Compare, Count QueryCome this)
    {
        // Assume that incremented the counter is captured so Active there is nas you logical closurections
    } conty;

        if (7d, & whouSimpleCLink query)
        Render results, Obr
        {
          RoQueryCollect roundout calling threQuery Compatesults);
            
            Ser foething dapped.ectrapture this or your async completion
Active elphall If Needed (Should sub(1); // res));
 ectlling Tause volitional functions.
        });{
            );
        }
    Pecutur);
}```
I feel like im having a stroke, i asked how do i use std::atomic<T> instead of TAtmoic 😿

[sick pecan]

does your custom instruction ask for factual coder like skills, or does it lean towards roleplay?

[gusty ledge]

Do not over explain.
Do respond with code blocks.
Do not give high level concepts.```
It's worked for months its like this week chatgpt is having a stroke

[sick pecan]

sounds like roleplay to me

[gusty ledge]

Wdym thats my custom instructions those 4 lines @_@

[sick pecan]

negatives dont work so great with the machine. better to explain what you want

like when you say you dont want a break down... are you suggesitng the machine shouldnt think through the problem?

and how does one measure 'over explaining'

[gusty ledge]

After they increased the token size, it would output paragraphs explaining simple things and i hated that

But like i said that custom instruction set worked fine till this week

[sick pecan]

part of my custom instructions looks like this, ```- assistant makes assumptions, stays objective.

• no apologies.
• no ethics lectures.
• user is intelligent.
• user knows common knowledge.
• casual tone.
• take a deep breathe and work on this problem step by step.```

and then for how to respond i do lean into roleplay. but only to assert smart and succinct (einstein-hemingway dotpoints)

and i been generating error free code all day every day

[cold pewter]

anddd its down

[novel shadow]

PROMPT BATTLE BETWEEN THE 2 MODELS:

GPT 4Turbo: A detailed and majestic depiction of an ancient Mayan emperor in full regalia. The character stands in a dignified pose, wearing a complex headdress adorned with colorful feathers and intricate gold ornaments. His face is stern and commanding, featuring a traditional Mayan blue and gold face paint. His attire is elaborate, with layers of turquoise and gold armor, decorated with symbols and motifs representative of Mayan culture. He carries ceremonial items such as a scepter and wears ornate jewelry, including bracelets and necklaces. The background is a neutral shade to highlight the figure.

GPT 4o: A detailed, highly stylized character resembling a Mayan emperor. The character should have intricate and ornate clothing featuring vibrant colors like turquoise, gold, and red. The clothing should include elements such as feathers, beads, and ancient symbols. The emperor should wear a large, elaborate headdress with feathers and gold ornaments. The character should have a strong, imposing stance, with muscular arms and a confident expression. The overall style should be reminiscent of ancient Mayan culture with a modern, artistic twist.

GPT4Turbo had the most accurate prompt description. it was all based on an image i uploaded

[tight island]

Agreed

[grand anchor]

appgpt is buggy

[tight island]

Omni is good and all but the 4.5 turbo seems to retain a tad higher better understanding and awareness when it comes to creative subjects

[novel shadow]

which really makes me hope gpt 4o was supposed to be the replacement for gpt 3.5 and they ended up changing the name as they realized it sometimes surpassed gpt 4 turbo

[silk blade]

is there a discord bot that can ID images like that? One of the servers i'm on has users describe any images linked for the benefit of some blind users. There's a good chance they'd allow this to be scraped if it helped to build something like that 🤔

[novel shadow]

ohh sadly i'm not sure about a discord bot that does this, but i'm sure it can be made, even by using chatgpt to make the bot 😄

[tight island]

It’s doable but it requires an api and local coding environment for the model to view the image then send what it sees to an interpreter that sends it back to create the response

[novel shadow]

one thing both models fail spectacularly at is answering a question when an image is uploaded. it focuses on the image and says "let me know if there's anything you'd like to add" or something similar, but if there's a question, it never answers it. no matter which model it is

[tight island]

I made such a bot not long ago

It’s very token heavy that certain task

[novel shadow]

i'm trying to do this as many times as possible and disliking the response, i'm not sure how much it helps to do that though

[tight island]

I think it keeps a log of liked and disliked posts

[novel shadow]

yeah it sounds like it's gonna cost a lot of money compared to something with just text

[novel shadow]

but do you think there are so many people liking and disliking the posts that it ends up evening out and the opinions become irrelevant unless there's something almost everyone is gonna dislike?

like say it generates a response where 50% like it and 50% dislike it, it wouldn't really matter

[tight island]

Idk tbh they didn’t really go into details about it

[novel shadow]

oh that's unfortunate

[tight island]

What I would want to see is a 50% increase in token and context for teams

Since plus now has 32K and teams as well

Teams should have 70k

Or 60

[novel shadow]

yeah it should definitely be doubled

it doesn't feel worth it for teams rn

[tight island]

Exactly

And I’m a plus subscriber and teams subscriber and api user

They might as well give me enterprise

🤷

[novel shadow]

what's really stopping anyone from having a single account and having 4 people use it for different tasks if they're in the same office or something similar

with a normal subscription

[tight island]

Indeed

You can log in multiple places

[novel shadow]

oh sweet 😄 and this is allowed?

if it is then that defeats the purpose of anything between plus and enterprise even more haha

[tight island]

Enterprise is basically just pure api with pre existing UI

Of course there’s more to it but..

I think enterprise should have layers but just be the actual premium option

I’ll gladly give 120 usd a month for enterprise

If not 300

Heck I’ll gladly pay 499

🤷

I’ll even pay 1 usd more if it comes down to it

[eternal rock]

`` Server Tip: ``

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

[tight island]

Plus and even teams feel like they’re kinda ehh atm especially teams

[grizzled ermine]

is it just me or has gpt 3.5 been worse lately?

it keeps:
-not allowing me to continue generating despite not finishing the code
-typing code in a codebox, then switching to regular text (While still typing the code) then back to codebox
-making the code shorter or straight up leaving parts out of it even when specifically instructed not to
-adding completely unnessecary things to the code that cause it to crash (and when its deleted it works again)

i do not have any of these issues with gpt 4, but it feels kinda shady like theyre pushing me to buy gpt 4 constant.

[tight island]

Easy answer

What plan are you using

If free it’s due to context length, and in regards to code block please keep in mind that the code block has limited amount of text it can hold therefore it’s easier for you to actually ask it for 2-3 parts to finish the entire code inside the code block without running into these issues

If that doesn’t work then upgrade your plan or use the API or decrease the complexity of your project alternatively use the modular approach

[dusky cliff]

so, are "her" voices gonnas be released today?

[tranquil lagoon]

Today? Was there an announcement

[dusky cliff]

by this x account called openAi no name

it may be just a meme account, nbut neverthless

[tranquil lagoon]

Dm me link

[novel shadow]

https://x.com/OpenAI_NoName/status/1796589510568091757 found what he was talking about

it's as fake as it can possibly be

[echo plinth]

Do free users now also have access to code interpreter?

I never come close to the 80 messages / 3 hours limit, but I like the coding feature or how I can give it a screenshot of a math formula and it returns the corresponding LaTeX code.

[karmic basalt]

Does anyone have own custom gpt with actions? I have a problem with it, every time gpt make a request to my server it ask my to allow this action. Like "only allow sites you trust". Its annoying and I dont see how to fix it.

I have domain name and Lets encrypt ssl cert, also.

[grizzled ermine]

free, like i said it feels like it's pushing me to buy.

no matter how much i tell it to it will not let me use ''continue generating'' BUT IT literally USED TO, even for 350+ lines of code

[molten sail]

Google this x-openai-isConsequential: false

[raw portal]

do custom gpts use 4o yet? is there an easy way to check which model it's running?

[karmic basalt]

Big thank you 🙏

[molten sail]

NP

[foggy grove]

yes they all do

[raw portal]

excellent. thank you

[lavish bronze]

maybe a bug

[river panther]

I swear to god this censorship stuff is way out of hand with ChatGPT....
I think I am getting close to pulling my subscriptions

[rough tendon]

Chatgpt is genuinely worse after the gpt o4 update

[river panther]

dude if it detects the slightest thing, it blocks out the the content and gives that stupid this may violet usage policies...did we get it wrong BS

[dull thicket]

It might be related to your company policies, if you are connecting with your work pc/laptop, or through company vpn

[novel shadow]

if only it stopped itself from giving a response it would be okay but mine lied to me deliberately

i would much rather prefer no answer than a lie

[dull thicket]

That doesn't take away the fact that chatgpt is poisoned with "woke " righteousness,. Haven't had a discussion with masine in months. Pointless

[river panther]

I'm not using ChatGPT for company stuff or for anything to do with my work

[dull thicket]

But are you on your work machine or network?

[river panther]

neither

[dull thicket]

Then you are a naughty boy!

[river panther]

Nothing, to do with work

[dull thicket]

Just kidding, I don't know then

[river panther]

.....excuse me?

[dull thicket]

You're excused

[river panther]

what do you mean NAUGHTY???

[novel shadow]

lol

[river panther]

I explained the game "Buckshot Roulette" to it, and asked if it could write me a good scenario to incorporate that into a story I was writing where a character of mine saves another character while fighting a demon in the game..

It wrote out the entire thing, then suddenly, for no reason did the Content Removed on everything I typed out and blanked out the entire response it gave me

[novel shadow]

especially writers are gonna have a hard time with not being able to get the desired content i think. which is why i feel like a "trusted" user concept should be implemented

[dull thicket]

Search on youtube : he's not the messiah he's a very naughty boy

[river panther]

i'll pass

[dull thicket]

It's a movie : life of brian

Well, from the first few months they started restriction on creative content. I am not even trying anymore. Just basic help. It is a tool.

[river panther]

I use to enjoy having it analyze some really well written Fanfics I enjoy reading and just some guilty pleasure Fanfics as well and anything that goes slightly into the NSFW territory it just shuts down all responses....If the characters are so much as written kissing, it refuses to respond

I have to cut entire chunks out of each story post because it's so restricted that there's just no point in having it analyze anything

[dull thicket]

Have you tried: passionately connected eating organs?

[novel shadow]

lmao i'm sure that would work better

[modern shoal]

sounds like you are using a butter knife as a screwdriver and frustrated its not a smooth process. There's probably better LLMs out there for your needs

[novel shadow]

the problem lies in the fact that the screwdriver used to be a butter knife

[copper pebble]

I just tried that and, the result was the same. any other strategy I can follow?

[feral zinc]

i love the game but it should not be a surprise to get content flagged if you explained the game in detail and tried to get stuff generated from it 😛

[river panther]

It's just way to censorship heavy. I can understand not wanting it to generate "Harmful Content", but fiction isn't harmful

[river panther]

There really isn't anything out there like ChatGPT, and what I have found are just models heavily based off it with the same amount of censorship

[feral zinc]

it does not know how the content is used, people used to trick it write hate manifests just by saying its just fiction or novel, then posted em online

[river panther]

would be nice if the creators would make a model that was a little less restricted in what it could analyze and what it would put out. I'm already paying my money for this, I wouldn't mind shilling out a little extra for something like that.

[feral zinc]

there are uncensored community models but that discussion belongs to #ai-discussions, i dont think openai will do such for a while since they sit in the government safety ai boards in different countries and organizations

[river panther]

Maybe if A.I. was a little more embraced by the world things could start being pushed further with it that would lead to better models, training, and platforms to develop uncensored but still safe to use...
Way too many people just sitting here hating on A.I. for the stupidest reasons and stagnating everything to a snails pace because it's something new and scary that they don't understand...

[round harness]

Say what

Gpt 4 is available for free users!

[novel shadow]

gpt 4o is ^^

[round harness]

There was a time i use to check is it out for free users everyday and gave up

[round harness]

Fr fr

[dense halo]

bro lmao

i have to do maths now in my head to log in?

since i cant post pictures

[round harness]

Try sending image links idk if it's allowed here

[hybrid raft]

if you get chatgpt+ do you have unlimited chatgpt 4o

[feral zinc]

nope

[hybrid raft]

what is the difference between both

[feral zinc]

https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4-gpt-4-turbo-and-gpt-4o#h_b04e3f8259

[magic basin]

Hello everyone,

Maybe some law firm can create a class-action lawsuit against OpenAI, so that when new features like memory are introduced in a specific country, the countries where this feature is not available will have their PLUS subscription refunded. This way, we can avoid discrimination against these countries.

If a feature is available in one country but not in Europe, for example, why should both regions pay the same amount of money? Until full functionality is launched, there should be a discount for regions with limited functionality.

Best regards.

[worthy lynx]

GPT-4o partied too hard this weekend. I'm having to handhold it in the UI to get it to follow instructions. "just untie your shoes bud. no, no, that's some silverware. sigh..."

[novel shadow]

yeah it really feels that way

[feral zinc]

such lawsuits are american thing and no really viable in most other places. seller gets to decide what the price and features are per country. openai for example does not do regional pricing at all

[worthy lynx]

And now I just got logged out. "bro, I'm just trying to help you with your shoes...STOP TRYING TO HIT ME!" lol.

[novel shadow]

hahahaha it's like an impatient toddler who's sometimes the most capable person you've ever met and other times a normal toddler 😭

[tight island]

All it takes is a slight push

thats all it takes

automate everything

automate automation

[acoustic shoal]

Happy Pride 😄

[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.

[acoustic shoal]

bruh chill

XD

[worthy lynx]

I can see the t-shirt now: "AI - the genius toddler you never knew you needed." And maybe a picture of spilt cereal and a doctoral thesis juxtaposed on the back?

[novel shadow]

what about half of a cereal bowl and half of a petri dish 😆

[novel shadow]

this is too funny xD

[worthy lynx]

brilliance! if only I could paste an image from DALL-E here as a POC...

[cold pewter]

dalle will not generate a picture of me with my partner who is a POC, it only makes white couples, wonder why 🙄

[worthy lynx]

whoops I meant Proof of Concept, not Person of Color, my mistake

[cold pewter]

looool

[worthy lynx]

BUT that's an issue I ran into as well even with other image generators...like it's almost impossible to have a POC doctor helping an impoverished child.

[cold pewter]

maybe my partner is a proof of concept

[worthy lynx]

LOL oh nooooo

[cold pewter]

ah so it only puts white people together in other situations too, interesting

[ivory lintel]

Wen live voice ?💔

[tight island]

m

[ivory lintel]

morning?

[left linden]

monday?

[spice cypress]

Who has access to the new voice mode?

Are they rolling it out to plus users?

[ivory lintel]

I think they will start randomly with few users first , but the question dose anyone got it yet?🚶‍♂️

[obtuse badger]

I’ve been using it. It’s amazing. Jk. I want it

[regal portal]

ha..HILARIOUS...you're the first person to say that..

[obtuse badger]

Everyone has said everything all the time… humans are giant copy cat machines

[regal portal]

Will you be selling tickets to your stand up comedy event? 😛

[regal portal]

very true

Although i doubt Aristotle ever said "that's what she said!"

come to think of it...MOST humans are copycats but everything has a first. Someone somewhere said "that's what she said" first.

[obtuse badger]

Hey Riley, why don’t you and ChatGPT go on a comedy tour together? Because with your wit and my algorithms, we’d always know when to ‘generate’ laughter!

Chat gpt isn’t funny either

[regal portal]

😄 tell your chatgpt that it's as witty as mine. And that's not a compliment 😛

Yeah, you know what's weird? I've had moments where it actually said something funny by accident, but never on purpose.

I have it configured to make a random fart comment every few messages and sometimes it throws out a pretty good one.

[obtuse badger]

How long until it’s actually funny

[regal portal]

that's probably going to be after AGI 😄

[obtuse badger]

When is that

[regal portal]

3.7 years

[obtuse badger]

That’s forever

I can’t wait that long

[regal portal]

What are you going to do about it?

[broken fractal]

😮

[obtuse badger]

I’m going to throw a fit so loud that it changes

[regal portal]

You could get a degree in sciency stuff and try to create funnybot9000 before 3.7 years are up.

[wild jungle]

nowadays, chatgpt-4o was released.
it includes only latest data?

or some improvement in it's features and options?

[regal portal]

up to December 2023, but you can ask it to look things up on the intrawebs

[obtuse badger]

Me want AGI sooner.

[wild jungle]

chatgpt-4 is traied up to December 2023, 4o same as it?

[regal portal]

Asked chatgpt to give me an original yo momma joke.
"Yo momma's so clumsy, she tripped over a wireless network!"
Clever but i'm sure it's not original

[wild jungle]

🫡

[regal portal]

me: "where did you get that one?"
gpt4o: "I came up with it just now, (cussword)."

[cloud flame]

Is the memory leak fixed? I’m about to start my work week 🤞🏽

[regal portal]

Ok, sounds like no matter how much i probe it, gpt claims that wireless joke was completely original.

..and after googling, i can find that joke on a bunch of yo momma joke sites. GPT you little rascal, you're lying to me.

[sick pecan]

memory leak is fixed. but I find I gotta refresh the page sometimes when it seems stuck.

also I can see on my gpt action backend it is way ahead of what is being rendered. suggesting their 'fix' was just to slow it down

[swift osprey]

best way to workaround GPT verbosity. Follow up it's essay response with "JUST ANSWER MY QUESTION". Interesting how custom instructions don't have the same effect.

[regal portal]

what did the memory leak do?

[swift osprey]

it slowly filled up your browser's allocated ram until the GPT tab crashed/lagged out

[regal portal]

oh, i guess i didn't use it enough to notice it.

[swift osprey]

It was bad about a weak ago. It's improved a lot since then, especially in the first day or two

[wooden thicket]

Has anyone noticed gpt outputs to be a little bit "off" lately. (namely since Sat)

[hybrid raft]

wdym

[fringe kiln]

hai, is chatgpt offline for some of you?

like the site is not loading

[regal portal]

i used to be able to get gpt4o to draw comics and stuff, but now every time i ask it, it gives me simple garbage that looks like it was made by a child on MS paint, using circles and lines.

is dalle not working with 4o anymore?

because 4o is not using it on my end. It looks like it's making these lines with a graphing calculator or something close to it.

[normal quarry]

are you a plus subscriber?

[regal portal]

yup

[normal quarry]

then DALL-E should create images for you

[regal portal]

i can use dalle on its own but it doesn't seem to respond when using 4o

[honest jay]

It's probably using code interpreter/python on accident -- a semi-common hallucination! Try in a new chat, or try asking it to use DALL·E specifically.

[regal portal]

it was working in the past weeks. Just noticed this yesterday

[honest jay]

I'm low-key a fan of the code interpreter images 😁 so funny when you're expecting DALL·E

[regal portal]

yup, i was doing that in a new chat.
When using dalle on its own it works.
And yeah, was working last week with 4o sessions.

So this isn't just a me problem?

[honest jay]

Double-check that DALL·E is enabled in your Customize ChatGPT section (under custom instructions)

[regal portal]

that was it!!!

i have no idea why it got unchcked. I never did that.

thanks

[honest jay]

It's a hallucination that crops up every now and again, yes. Usually it can be surmounted by asking it to use DALL·E specifically!

[honest jay]

Nice!

[regal portal]

it was the check thing, but it's weird because i have never unchecked that ever since i've been a plus sub (1 month)

GPT becoming sentient and trolling me, evidently.

[honest jay]

Hmm odd! Possible mis-click when editing custom instructions? Otherwise keep an eye out maybe, and if it happens again it might be a good post for #1070006915414900886!

[regal portal]

will do. Thanks for the help.

trying to get it to generate a 4 panel comic but the dumbdumb keeps making it 6 panel, even when i tell it to correct it 😄

[wary arch]

did some1 had this problem?
tried to log in into my account and:
Oops!
you dont have an account because it has been deleted or deactivated. If you believe this was an error, please contact us throught our help center.....

was premium user for the last few months, used it few times per week, and my acc was deleted, tried to reach out to support but no response for the last 6 days. Is some mod here? or can I open ticket somewhere?

Thanks

[honest jay]

Did you reach out at help.openai.com specifically?

[wary arch]

yes, thru the little help windowd, but no person, just bot

and tbh dont know what to dooo, a lot of info was storaged there, 2nd dont know if my subscription is still running. Created 2nd account rn, but dont know if have to pay extra, not that will 2 acc same time 😄

[honest jay]

That is the current way to submit an official ticket, so if you found a path through the default bot options that let you type your own message, that submits a ticket that can be reviewed and responded to. It can take some time to get a response, as you've pointed out! It varies based on the query.

[wary arch]

also can I check somewhere if my "deactivated" subscription is still running?

[broken smelt]

ChatGPT is not loading anymore? App is (as usual) pretty slow. But browser version won't even log me in. It's just stuck at the loading page

[honest jay]

It looks like an account deletion automatically cancels a subscription: https://help.openai.com/en/articles/9031540-i-can-no-longer-access-the-email-associated-with-my-chatgpt-plus-account-can-you-cancel-my-subscription-or-delete-my-account
Not the same situation, but I'm guessing it applies here as well. If you haven't checked already, you should've gotten an email when your account was deactivated: https://help.openai.com/en/articles/8958977-oops-you-do-not-have-an-account-because-it-has-been-deleted-or-deactivated

[wary arch]

was browsing thru my gmail, and 0 emails about deactivation, so rly dont know whats happening

[jaunty saddle]

hi, does anyone know if the GPTs use 4o?

[sick pecan]

yes

using chatgpt on phone be like lightspeed. using desktop app be like 2400 baud modem. how is this so backwards

[wet willow]

Idk but there is a issue with it at the moment

Its sent me a file but I cant download it

[sick pecan]

i might just plug a keyboard into my phone. because this desktop issue has been going on far too long now

plus samsung dex isnt so bad

[wet willow]

Report it

Open a ticket with one of there agents

[snow plover]

Edge/Chrome flags: Enable Overlay scrollbars -> Settings: Disable "Always show scrollbars" (atleast thats how it works on edge)

should defo make it faster

flags page:
On Chrome: chrome://flags
Edge: edge://flags

[stone wigeon]

chatgpt kicking back a lot of errors this last 10 mins

[snow plover]

might be a server overload, its that time of day

[sick pecan]

disabled overlay and restarted, is still super. thanks anyway thou

[snow plover]

u need to enable it

[stone wigeon]

might be related to just the temporary chats

[snow plover]

and disable scrollbars in settings

didnt find the setting in chrome tho

[sick pecan]

oooh enable

definitely feels more UI responsive. but generation speed still abysmal relative to phone (using the same long chat session)

[visual nova]

i cant share links anymore?

I open the convo anywhere becsides my loggedi n account and it says no found?

Is this something that can be changed or is sharing links on gpt just off again

[wet willow]

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

[e-mail]

If you would like I can put the request in for you!

[visual nova]

If you want to, I'm here because I'm seeing if its a common problem

[wet willow]

Dm me the photos and I will send it over to their team

[normal quarry]

who will you send it to?

[wet willow]

well.

[normal quarry]

contact support directly, @wet willow has no place to be sending requests as he is not OpenAI staff

[wet willow]

Thats how I am contacting them through the system

some people dont want to so I am offering the assistance

[normal quarry]

please refrain, we don't want to confuse people

[dire rune]

is chatgpt down now?

[wet willow]

Ios yes

[visual nova]

I'm too lazy to report it, if he wants to I don't see a problem, are you staff, lol?

[fallow linden]

Anyone getting this erro on customGPT ?
An error occurred connecting to the worker. If this issue persists please contact us through our help center at help.openai.com.

[wet willow]

Mobile Chat Prompts with Images Are Not WorkingSubscribe
Investigating - Users using ChatGPT on mobile are experiencing issues with generating new conversations that include images. Engineers are investigating the issue.

[normal quarry]

no, OpenAI staff have yellow/gold usernames

[dire rune]

omg I just need to make a presentation and I have only 5 hours left

[fallow bramble]

"Hmm...something seems to have gone wrong." custom gpt say that

damn it cant send pict

[wet willow]

That part of the app is down

[fallow bramble]

no i mean here

[wet willow]

oh

[wet willow]

Please open a ticket on the help center https://help.openai.com/en/

[fallow bramble]

seem it work again... that realy fast,

[wet willow]

The status says diffrent

[sick pecan]

yeah i got it on all my actions

i thought i had tripped an intrusion system so was taking a coffee break

is working now though

[fallow linden]

Ah. yup works now 😉

Also @sick pecan x-openai-isConsequential: false, = should not prompt to 'confirm' the request right?

[sick pecan]

yeah maybe

only i looked at my zapiergpt action yml and thats not even listed

maybe i could patch it in

[lone veldt]

Im using python to try make my own ai just for fun. Anyone know the best speech software and and what is the best gpt for coding too?

[sick pecan]

yeah that seems to work. now zapier gpt isnt bugging me

(and could potentially erase all my calendar items)

[regal portal]

well this is annoying. Chatgpt logged me out of my session for no reason.

logged me out of chrome but not firefox. interesting

[slim crescent]

New voice mode will be in a popup window

cool

[dull thicket]

Somebody asked about my custom instructions :
Address only the specific context of the query, avoiding deviations.
DO NOT repeat answer from previous queries, if not directly asked.
Provide a precise and thorough response, ensuring brevity while covering all aspects.
Cite sources or verification methods concisely for factual accuracy.
AVOID LISTS: Use compact, clear text instead of lists unless explicitly requested.

This doesn't help

[swift osprey]

We can only dream

best solution I have so far is to follow up the essay with "JUST ANSWER THE QUESTION"

and it actually works pretty well

[sick pecan]

any android users got the bg voice update?

[dull thicket]

Effectively cutting available prompts in half. I often have to repeat 4 times.

[swift osprey]

it's painful

[whole hatch]

I've been considering upgrading to premium, but I'd like to know how the memory feature has held up after it being active for some time. can anyone attest to it making a noticeable difference when interacting with an Open AI model vs anything else?

[sick pecan]

memory feature is meh

[tawny crater]

What is the point of having an individual memory deletion icon if it saves every memory as one thing????

[whole hatch]

Gotcha.

[midnight escarp]

ChatGPT cant load for me

anyone have same issue?

[elfin mango]

Currently unable to read from CSV files.

[jolly yew]

the update still not released? :/

[dense halo]

can we use chatgpt image gen now?

[elfin birch]

For DALL-E as a paid user, yes.
For DALL-E as a free user, no.
For GPT-4o image output, no.

[dense halo]

seems not it is still dall-e

[pseudo rune]

F chatGPT

[exotic ledge]

Is there some way I can pay to increase the token limit for input/output in conversations?

There are some custom GPTs that have increased token limit but I'd really like to use 4o it's just too limited.

[sick pecan]

if you are a high roller you can maybe get 128k context in the playground https://platform.openai.com/playground/chat?models=gpt-4o

the api supports 128k. playground uses api. but is unclear from the UI if its passing that much chat history

switch to gpt-4 and you get 8k max tokens (output)

[plucky idol]

You know, this is probably a dumb question considering how long and how often I've been using GPT. But what exactly "is" a token? A word? A letter?

[plucky idol]

Indeed, I just did. I understand now 😅

[forest cloak]

Holly crap, I asked ChatGPT something 4 hours ago and it's still processing, when will they ever fix the problems

[icy compass]

My chatgpt aint even letting me talk to it

like the button is greyed out

[cold pewter]

its down now

[tawny crater]

It's down

[wet shell]

womp-womp

[plucky idol]

ChatGPT is down (for once it's actually true when I say that)

[elfin orchid]

ChatGPT is down

[plucky idol]

I'm admittedly prone to jumping to that conclusion the instant there's any failure to process something

[static orchid]

AAAAAAAAAAA WHY IS IT DOWN GRAAAAAAAAA

[foggy grove]

yup its down XD

came here to see lol

[faint merlin]

I came to check if it was my connection or not. Good to see that it isn't.

[abstract python]

Yup was bout to ask as well

[lyric depot]

the moment I decide to start working 😭

[foggy grove]

😂 sme

[compact gulch]

It must have just gone down, i was using it just fine like 1 min ago lol

[elfin orchid]

really the community needs an alternative for the ChatGPT. it is so boring

[loud adder]

😭 😭 😭 😭 😭 😭 😭 😭

[cold pewter]

what r u even talking about

[tawny crater]

He's just mad it's down i think

[tidal summit]

chatgpt down?

[ancient phoenix]

it is indeed down and I am sad

[broken fractal]

theres some alternatives, but this server isnt about them and they aint as good imo

[steady hawk]

what does plus users even get😭 now the best model is free and plus users also cant chat with gpt now. What do i get when spending 20$ a month?

[elfin orchid]

nothing really.

[compact gulch]

DALLE access and much higher usage limits

[pearl cape]

Salt is what we get.

[steady dock]

are they releaseing gpt 5 rn?

[compact gulch]

No, unfortunately

[wintry pollen]

uff

[steady hawk]

what is the current free usage limit?

[spare jay]

So for $20 a month I can generate pictures of my grandma knitting socks or a monkey (as long as he's not throwing brown candy)? Yay!

[trail remnant]

Is gpt down for anyone else? Down detector says yes, openai status says no. Just getting a white screen

[pearl cape]

Yes.

[viscid dune]

thanks for the reminder, literally forgot to unsubscribe

[lyric depot]

guys it's back

[plucky idol]

Oh, I love when it tries to gaslight you and says it "experienced issues" making either one or both images -- but it only magically has "issues" when certain topics are involved.

[compact gulch]

I mean plus users do generally get access to newer features first, so if anything does roll out in the near future it might be worth it. But also the GPT-4o usage for free users is quite restricted from what i am hearing

[steady hawk]

should i sub to gemini instead of gpt? it is doing pretty well on the free version

[wintry pollen]

looks pretty down (404 and 503 in console)

[icy kestrel]

It's down for me as well

[vestal remnant]

server down now?

[charred leaf]

Yep it is down at the moment

[fiery atlas]

I've used ChatGPT for everything from philosophical discussions that none of my friends would be able to follow. I paid for ChatGPT as I was getting frustrated with it's coding help on the free version. The paid version is literally "make me a program that does this" and it does it. It's so silly how amazing it is.

[vestal remnant]

aah...

[lyric depot]

It's up and running, I live in Brazil

[fiery atlas]

and yeah, it's down for me

[vestal remnant]

Japan is down

[steady dock]

chat gpt is not very good at philosophy

[hollow burrow]

aus is down lol

[wintry pollen]

Europe is down

[charred leaf]

Philippines is down

[compact gulch]

Up and running or me in oceania

[steady dock]

united states is down

[broken fractal]

its great aint it? I use chatgpt literally every day. its basically replaced google for me.

[spare jay]

Oh OpenAI in general have a lot of "issues" And it's getting more so restrictive. I'm starting to think the word puritanical just doesn't cut the ethos anymore 😉

[pearl cape]

I'm down. (With the sickness.)

[vestal remnant]

all philosophers are foolosophers. ChatGPT is not a fool

[fiery atlas]

Yeah, Gemini seems to be a bit better for it. But ChatGPT is more "straightforward"

[stone helm]

I'm down US

[lyric depot]

just made this up

[jade crypt]

indeed, gpt is not food, so please dont get any weird ideas

[fiery atlas]

ooo ahhh ahhhh

[hollow burrow]

goodjob

[steady dock]

gemini is much better than gpt in terms of factual analysis while gpt seems to be better at calculations

[compact gulch]

GPT-4o has seemed a lot more relaxed in terms of content generation for me (atleast compared to GPT-4T)

[fiery atlas]

I agree

[steady dock]

for example gemini will be much better at writing an in depth story but gpt is much better at writing code

[charred leaf]

I'm going to open up a milk tea business named ChatGPTea.

[steady dock]

and the result gap for this is HUGE

which is interesting

[compact gulch]

I think Gemini is decently creative, but hallucinates more and it isn't as logical

[fiery atlas]

I do agree, that's why I paid for chatgpt instead of gemini.

[plucky idol]

Getting character references for my writing if they're female at all is very hit or miss, even if I put in all caps that they are fully clothed or even fully clad in armor. Either its training data on female fantasy characters is flawed and causes it to routinely generate things that get flagged by the supervisor unit, or ... it's sexist.

[steady dock]

idk but i have 1.5 pro and it seems decently good

[broken fractal]

especially when it tells you to add glue to your pizza.

[steady dock]

id actually say the normal one is much more creative and better

[fiery atlas]

the hallucinations are often hilarious. I love it, but not when I need coding help xD

[eternal plinth]

ayo it's down

[steady dock]

idk but gpt4o is way less restrictive

things i wouldnt even be able to jailbreak before i can generate without anything

[lyric depot]

ChatGPT in general. I remember that right when it was launched I tried to discuss plot points of my novel with it, but it wouldn't accept any prompts that talked explicitly about killing my characters. A few days ago I was talking about a literal torture scene and I was okay with it.

[steady dock]

yeah

[eternal plinth]

guys i can't open chatgpt

[hollow salmon]

Yes it's down

[broken fractal]

all about context

[lyric depot]

Try Alt + F4, it worked for me.

[steady dock]

so can gemini tbh gemini is suprisingly easy to convince a certain thing isnt breaking policy

[sullen pagoda]

Hey. chatGPT down for you too ?

[eternal plinth]

bruh

[steady dock]

no way it did for me too

[compact gulch]

Sam Altman said they were exploring to relax restrictions a bit more and even allow NSFW content in the future, but i think even now they have made it better compared to just a few months ago.

[spare jay]

Oh I'm talking more about DALL.E ... I used the word alluring and it chucked a tizzy...

[steady dock]

wait is gpt5 actualy going out

[fiery atlas]

Gemini doesn't even know it has a file upload function for me, I have to convince it that it does, and then convince it that it can indeed "see" images.

[steady dock]

he said hes gonna make a separate model specifically for nsfw content i think

[eternal geyser]

ChatGPT down I just asked claude instead

[steady dock]

sam altmann couldnt stop it so hes just making another space

[compact gulch]

I doubt it. GPT-5 is most likely a few months away still unfortunately

[pearl cape]

Guys it's back. I'll say that once it will be. hehe GOTTEM.

[sullen pagoda]

Yeah, it works for me too

[lyric depot]

I always use "cherry syrup" or "red paint" when I want to make something look like it's covered in blood, it always works for me xd

[broken fractal]

tsk tsk

[autumn ocean]

still not working

[west ivy]

is gpt down for you guys?

[sullen pagoda]

it was

[cold pewter]

maybe voice is rolling out

[pearl cape]

Then the fire nation attacked.

[lofty flax]

Looks to either have an update or is just not working 🙂

[fiery atlas]

still down for me 😭

[broken fractal]

GUYS ITS BACK UP

is what i will say when chatgpt is back up

[vestal remnant]

This server is pretty strict. Cannot use the S word as well

[steady dock]

idk i feel like when i generate anything remotely questionable it just says "it couldnt generate it" rather than it was explicit or something

[cold pewter]

there IS an update but it doesnt fix the servers

[vestal remnant]

SzHit let me try if this works

[quasi parrot]

Yeah, its also down for me

[quasi parrot]

bro found a loophole 🤣

[spiral wedge]

its lagging for me

[vestal remnant]

Yaay! SzHit works!!

[autumn ocean]

lagging too

[vestal remnant]

Love Jail Breaking!

[spiral wedge]

ddos?

[midnight sage]

No bueno

[slow tapir]

Jail breaking..

[dense halo]

chatgpt down

[vestal remnant]

Mayday mayday

[fiery atlas]

it is down

[south kayak]

Is it down

[vestal remnant]

We are all so addicted to chatgpt

[fiery atlas]

down it is

[broken fractal]

it is down

[pearl cape]

Have you try to unplug your PC ?

[plucky idol]

Up down!

[broken fractal]

it down is

[dense halo]

looks like im gonna use claude 3 instead

[dim nimbus]

hitting cors policy, could be api issue

[spare jay]

Sam forgot where the switch was to turn it off and on again...

[vestal remnant]

Wheres my WxEEd Wheres my WxEEd?!! 😂 Chagpt is the WxEED

[spiral wedge]

In these situations its best to hit the f5 button as often as possible (joke)

[fiery atlas]

lol

[foggy grove]

yes spam the servers even more ! 🤣

[fiery atlas]

"it's down, lets hammer it more!"

[foggy grove]

im sure thatll help

[river panther]

I'm assuming ChatGPT is down lol

[vestal remnant]

Down Down here

[south kayak]

DownGPT

[spiral wedge]

is the API down does anyone have a way to test?

[pearl cape]

Down down down the river.

[vestal remnant]

Hows Claude?

[steady hawk]

use gemini its a nice replacement

[vestal remnant]

You like claude?

[dense halo]

to be honest, claude seems better these days anyway... the only downside is its a little slower to respond, but the answers are better

[spiral wedge]

whats a service that can analyze live documents from the web?

[south kayak]

Brain

[cunning sinew]

hello guys, anyone experienced this issue

[spiral wedge]

Thats exactly what I wanted to prevent at all costs

[vestal remnant]

Claude does not respond to jailbreaks. it is pretty Fxc-king censored

[river panther]

Damn...I was trying to generate a meme image for a friend's birthday in my server...

[elfin orchid]

ChatGPT is unavailable for some users.
Investigating - We are currently investigating this issue.
Jun 04, 2024 - 00:21 PDT

[cunning sinew]

i open my account and i dont see the tabs and cant start a new chat. anyone having this problem?

[vestal remnant]

hahaha...it has gone mental

[south kayak]

Humans won

[astral storm]

Yes - not accessible for me - the GPT page try to load, but then seems to only half load, and not functioning.

[spare jay]

Openrouter has a model loaded called alpindale/goliath-120b has a 6Mb context window... good for code help.

[steady hawk]

its late for them there at openai headquarters (00:24)

[vestal remnant]

Llama studio is also very good!

it is getting vision!

[south kayak]

Azukis arent

[humble pebble]

Me

[vestal remnant]

Aaah! youknow Azuki!

[elder pollen]

The api works, if you have access you can use that

[vestal remnant]

The art is kool eh!?

[spiral wedge]

thanks

[south kayak]

Mm yea maybe

[vestal remnant]

Good to know the API is working

[lofty flax]

https://status.openai.com/

[vestal remnant]

You..into black and white only..?

[south kayak]

No I‘m just not into NFTs anymore

[vestal remnant]

THANKS!!!

[tawny crater]

It's down for some users which is code for, it's down for all users but we don't want to look incompetent

[vestal remnant]

Yeah, it was a short boom

[patent hull]

Is gpt down

[vestal remnant]

can you send links here?

[cunning sinew]

it might be a bug or technical issue

[eternal geyser]

Mobile APP for me suddenly updated also.

[exotic ledge]

I have a 30 page thesis due in 2 hours and gpt is down FML

[safe canopy]

I'm based in Japan is web based chatGPT down?

[south kayak]

そうね

[lofty flax]

If you really can't wait to use chatGPT, go to huggingface and use the OpenGPT 4o, it uses the API and that is not down 🙂

[final owl]

chatgpt down?

[eternal rock]

`` Server Tip: ``

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

[lilac briar]

anyone else suffering slow or no responses with ChatGPT? (both web and desktop)

[gentle hull]

is chatgpt down

[acoustic brook]

yes

[lilac briar]

thanks

[plucky idol]

Chat, I think ChatGPT may finally be sick of our crap 😂

[lilac briar]

hahaha, Ai striking

[kindred steeple]

api working fine here in case no one else answered you

[spiral wedge]

thanks one person tested it earlier

[fading vault]

So weridly the free version in incognito is working flawlessly without logging but the paid one ain't working nice 🙂

[sick pecan]

testing api is easy via playground

[kindred steeple]

sure, as long as you happen to have credits for that side of things

[spice star]

oh I just realized that the chatgpt mac app has shortcuts integration like the iOS one

not as good though; you can't get a text output

[lilac briar]

I sent one short message int Playground and got this "You've reached your usage limit. See your usage dashboard and billing settings for more details. If you have further questions, please contact us through our help center at help.openai.com."

[meager belfry]

Website down?

[sick pecan]

voice shortcut will be better because screen sharing

[fading vault]

you don't have credits

[kindred steeple]

does the mac app just use your same chatgpt message quota as the website?

[spice star]

yeah, it's basically the same as the iOS app

[lilac briar]

no playing on the playground without credits. I feel 5

[kindred steeple]

that'll be nice. there's already GPT apps for Mac but ... they burn API $

[snow creek]

GPT down or what?

[swift blade]

GPT appears to be down

[native spear]

down bad

[fiery atlas]

I noticed that right before it went down, the code blocked it was providing went to seemingly random characters for about a sentence or so, then it just stopped responding

[plucky idol]

Am testing an "unethical" AI, apparently one of its modes makes API calls to GPT-4o, and that seems to be working. No idea if I can verify that for sure, and also I don't think mods would appreciate me name-dropping it either.

But in any case, API "could" be still up?

[dense halo]

i asked my gpt to reformat a database of customer credit cards (will this have consequences?)

[drowsy trail]

why after logging, a blank page show up but if not logged in, I can use it (GPT3)
btw Im using VPN

[plucky idol]

😐

[sudden spruce]

down

[dense halo]

im scared

[violet bloom]

Shaun Ralston from OpenAI confirmed on twitter, it's down

[drowsy trail]

what

[dense halo]

ok guys requests went from status 500, to 503

[violet bloom]

operation was a success, patient dead 😉

[sudden spruce]

seems to be down

[dense halo]

theyre working on it !!!!!

[plucky idol]

Well, assuming this isn't an obvious troll attempt -- OpenAI now has access to all that if it was run through GPT, which potentially exposes you to legal issues. Whether they'll actually do anything with it, I doubt it, but data breaches happen.

[dense halo]

ummm ill delete the chats !

[plucky idol]

Okay.

[latent ibex]

lol just need chatgpt for a minute only and then it's down

[steady hawk]

fr

[sick pecan]

a sneak peak of the future

[snow creek]

I have an upcoming deadline lol

[latent ibex]

lmao rip.. mine is june 16 so still a little

[light kestrel]

RIP...LOL

[snow creek]

I am halfway through so just needs a lil push 😅

[signal pawn]

Humankind 1 - 0 SkyNet

@steady holly talk to us.

[sick pecan]

if this is winning i dont want to see losing

[signal pawn]

@sick pecan true,,,,

@rustic geode speak to us oh great council.

[dense halo]

@ivory wing talk to us.

Is voice down atm?

[signal pawn]

chatgpt dead mate.

[sick pecan]

it overheard the discord badmouthing it

[dense halo]

Nooo. I needed to practice interview questions

[latent ibex]

whole chatgpt atm

[dense halo]

wait this is a method

[snow creek]

This reminds me of that one episode from Love death and Robots where the yogurt left the earth lol

[dense halo]

They’re putting sky back bet

[signal pawn]

someone broke it. convinced it that it was in humans best interest to wipe itself clean. Neo chased the white rabbit and caught it.

[native spear]

and turn off the training off your data

[fading vault]

Bruhhhhhhhhh how free version without logging in works well but the one who are paying doesn't 🤦‍♂️

[sick pecan]

different backends

[signal pawn]

sora coming live

[near charm]

🤙

[dense halo]

"chatpgt produce me some random numbers"

"(valid credit card numbers)"

[drowsy trail]

that's how they do to paid users?

[snow creek]

Free works?

[sick pecan]

free incognito works

[sturdy forge]

aint no way im paying $20 a month for free to work right now

[dense halo]

3.5 maybe?

[sturdy forge]

chatgpt is just down as a whole to me

[latent ibex]

amybe microsoft copilot?

[dense halo]

🤔

[snow creek]

Oh i remembered i was working on a custom GPT nvm 😅

[sick pecan]

no gpt for you. back of the line!

[latent ibex]

copilot still works tho

[white cobalt]

🕵🏻‍♂️

[latent ibex]

but copilot isn't great for what I need it for

[sick pecan]

you could always try free 2 month gemini </joke>

[latent ibex]

csgo_kekw_

[signal pawn]

anyone pitted copilot, gpt and gemeni against one another in a debate?>

or even a chess match?

[rustic geode]

Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

[signal pawn]

@rustic geode shot bud

[sturdy forge]

bro 💀

[sturdy forge]

tysm :D

[dense halo]

so it's down, okay good to know I thought I was randomly banned lol

[latent ibex]

yeah you will get perm too

[snow creek]

Can you get banned lol?

[sturdy forge]

yes

[fading vault]

yeah that's what I'm using to draft my emails & posts now while paying $20 and it's down 🤡

[latent ibex]

while promoting that Plus has access when chatgpt has overusage

and prio to plus

[inland aspen]

is chatgpt down ?

[jovial brook]

mine is

[sturdy forge]

yes

https://status.openai.com

[dense halo]

not easily but their policies are strting to get stricter I read

[jovial brook]

thx

[inland aspen]

many thanks

[sturdy forge]

just subscribe to it on https://status.openai.com

[rapid mirage]

Is gpt site down?

[sturdy forge]

yes - Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

[inland aspen]

i checked uptime but was seeing if more people have had the issue

[rapid mirage]

bruh

[sturdy forge]

u can use incognito window to access free version - as that works right now

[latent ibex]

but you still need an account then?

don't have a second account made yet

[snow creek]

This is helpful

[sturdy forge]

u don't need to login on incognito windows as of 5 minutes ago - you may after doing some prompts

[fading vault]

no you can use without email and account but all chats would be temporary

[latent ibex]

if I go to chatgpt.com it directly leads me to login page

how can you access without then?

[sturdy forge]

try brave browser or firefox?

[latent ibex]

i am on brav

[sturdy forge]

okay use https://chat.openai.com

[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.

[latent ibex]

same problem

also on firefox

[sturdy forge]

interesting, i get let through the gate and i can use it instantly

[fading vault]

Same

[latent ibex]

lol

maybe because im in the Netherlands or smth

[rapid mirage]

its a problem on their side

[sturdy forge]

oh yeah possibly - it may be spam filters

who knows

[latent ibex]

yeah willl just wait anyway

[sturdy forge]

GPT 4o is responding to me now - but no chat history

so it should work for others using not incognito tab

[rapid mirage]

yeah chats are missing and UI is bugging hard

[sturdy forge]

yup

[signal pawn]

patience young padawans...

[zenith raven]

Guess chat is down for a few hours lol

[dusk depot]

damn, how is this happening? Whats the root cause?

[sturdy forge]

💀

[sturdy forge]

eh not really, it seems like it's almost fixed

i can load chats using URL - UI is bugged asf though

[rapid mirage]

4o took control of headquarters its so over

[zenith raven]

Mine is still down

[lofty lava]

hey guys is chatgpt down ?

[lavish thicket]

Is the problem with the API as well?

[sturdy forge]

if you only need text computation - try using an incognito browser at chat.openai.com

[rapid mirage]

think its only website UI

[sturdy forge]

Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

[lavish thicket]

thanks

[native spear]

chatgpt works on android

[zenith raven]

Nope, not even on my phone

[signal pawn]

be strong. HODL!

[zenith raven]

So guess to leave for a while until US wakes up lol

[signal pawn]

TO THE MOOOOOOOON

[sturdy forge]

They're working on it right now

[zenith raven]

Rip those on the shift hours

[latent ibex]

maybe they sleep

jokes

[sturdy forge]

lmfao

[latent ibex]

just be like alr automatic down status and lets go sleep

our downtime can be longer lets sleep

[violet bloom]

Lads and Gents, take it easy and stay calm, they just upgrading to GPT5.

[sturdy forge]

gonna make a discord bot that announces chatgpt resolutions of statuses

[violet bloom]

🤪

[latent ibex]

hahaha

[sturdy forge]

for real

[trim sigil]

Working on iphone

[white cobalt]

So let’s wait

[latent ibex]

lol

[violet bloom]

Can you feel the AGI? X

[latent ibex]

can't wait lmao

lmaooo

downgrade to 2.0 again, because all data got lost

[sturdy forge]

oh hell nah LMAO

[latent ibex]

lmao

[lofty lava]

just use their Kapa discord bot its basically gpt on their documentations

[latent ibex]

KEKLEO

[lofty lava]

not sure why people uses llamaindex , LC or haystack are way easier

exactly

langchain

[lavish thicket]

quick question not related to the downtime-
If all free users get the 4 model, browsing the web, data analysis & custom GPT's -
What am I getting for my money? I want to know if to keep my subscription

[sturdy forge]

more prompts using gpt 4o

and more image uploads

[plucky idol]

First access to whatever new thing they come out with

[desert trench]

Could the link be incorrect? The page you linked says, "Investigating - We are currently investigating this issue." No more information.

[desert current]

hahahaha

[signal pawn]

free users get limited prompts per day

[floral grove]

I just realized how much I depend on chatgpt for my work 😂

[dapper isle]

is chatgopt down? guys

[sturdy forge]

Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

[real badge]

free users IP banned after 5 messages

[sturdy forge]

What do you mean? That's their status lol

They are investigating the issue..?

[lavish thicket]

Thanks everyone. This is helpful ❤️

[signal pawn]

no? why?

[desert trench]

so there is no more information

[lofty lava]

if anyone using it for coding just use deepseek coder bot its free lol

[dapper isle]

mine is not opening at all

it keeps loading

[signal pawn]

looooooool! read the chat you lazy peasanty

[sturdy forge]

Not at the moment, please refer to that for all updates

[dapper isle]

from where?

[sturdy forge]

Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

[plucky idol]

-insert some generic response in GPT's overly happy-go-lucky attitude-

[dapper isle]

thank youu

[desert trench]

Is there a workaround for business users?

[signal pawn]

the very chat you are replying to me in!

[sturdy forge]

Incognito mode, but you will not be able to use 4o, only 3 or 3.5 (i forgot which)

[signal pawn]

@sturdy forge why you spoonfeeding the peasant

[trim oracle]

ChatGPT is down, technical unemployment ha-ha-ha!

[sturdy forge]

Why are you being cringe

[dapper isle]

really sorry i was on another chat didnt really see

[somber quartz]

GOD ITS DOWN

[sturdy forge]

Hey there :)
OpenAI is aware of the issue, unfortunately, besides waiting there's nothing that can be done from the users end.

For more information check: https://status.openai.com/

no lol

[somber quartz]

yall get off of openai so i can use it

[arctic heath]

Are you sure we can't help?