chatgpt-discussions

[pliant herald]

australia can use

[rugged igloo]

chat gpt is struggling right now

[sly pilot]

bruh fr

[worldly mantle]

they are updating it to AGI

[shadow bridge]

down!!

[sterile condor]

same here

[fossil fiber]

broo they had to take down the servers while i was making a project

[frozen birch]

for me working now, but there is massive lags when asked for something

[fossil fiber]

i don't even get a response, always error generating

[glad python]

just wanted to ask but it seems I'm not the only one with "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[frozen birch]

anyone guys experiencing issues with memory of chatgpt too pepps? or only not working?

[dense obsidian]

yep and yep

[empty jungle]

Its buggy af today

[minor ledge]

hey guys is anyone elses web search not working?

[gilded swan]

Mine isn't- either my work or personal account

[dense obsidian]

Getting "The engine is currently overloaded" again...

[fallen wind]

Is search the web down just for me? It tells me it can't search the internet right now.

[honest jay]

Just tested - seems like I can't get a search to work either

[fallen wind]

Okay, so it's on their end... hopefully it'll be fixed soon.

[honest jay]

https://status.openai.com/

[plucky rapids]

Ok so other people are having problems too? It just retroactively changed an earlier response and deleted everything after that response, which was a lot. Extremely frustrating

[gilded swan]

Chatgpt seems to be able to browse the web again

[frozen birch]

guys if someone has problems with memory after todays problems - resetting entire memory helps with that issue, he is not formal after resetting data
Of course backup all memory and provide him that data after resetting, so he can work as you desire

[frozen fjord]

My AI identifies as a helicopter thank you very much

[opaque maple]

is there an ai product that can get the x and y coordinates for elements on a webpage using vision? That can click and interact with elements?

[sick pecan]

omniparser does the coordinates thing. here i use it to extract 50+ ui elements from a browser in 20ms.https://media.discordapp.net/attachments/1168986513346941030/1308556057399918622/image.png?ex=6746484b&is=6744f6cb&hm=76608073dfe0b7aa5cf67c7db391eb6038472cb9e72b7a620fedb2b64ff03377&format=webp&quality=lossless&width=2250&height=1014& albeit it's not a chatgpt thing

[maiden reef]

reminds me of factorio

[prime swallow]

That might get far more dishes cleaned at once!

[maiden reef]

lol

[maiden reef]

what did you ask?

[prime swallow]

Same convo.

All prompts:

We need a Python program that can get the dishes done.

We need it.```

```Right.  Save us all.  Explain to us the rest of the stuff we need to put **with** some awesomely designed Python script to get those dishes done.```

```Sweet.  Let's make a dall-e image that you describe kinda like a wallace and gromit sketch; avoid the names and just describe similar that celebrate our dishbot.```

```Look at this masterpiece!

Let's use the quirks Dall-E added, like it's just a Dishbot-300.

Let's use as much of the same id and seed and everything, and redescribe this in a fall-out affected world, as if the dishbot had been recreated in that harsher and more scrounging place.```

```Make sure it's g-rated and lets try again!  There's been some tech challenges today, some just don't generate on first try.```

```Glorious!

Let's see one based off of Factorio style, otherwise the same Dishbot 300!```

[prime swallow]

A Factorio-inspired version of the Dishbot 3000, set in an industrial and mechanical environment. The Dishbot is integrated into a sprawling, conveyor belt-filled factory setup with intricate machinery and pipes. The design features an over-engineered aesthetic with gears, pistons, and robotic arms handling plates and utensils. The Dishbot has a rugged, utilitarian look with glowing indicators and control panels, surrounded by assembly lines and smokestacks. The environment is dark and industrial, illuminated by the glow of green and yellow factory lights, with an emphasis on efficiency and automation. The quirky charm of the Dishbot is preserved in its comical over-complexity, fitting perfectly into the Factorio-style setting.

[maiden reef]

that does seem like a good system for cleaning dishes tho

[prime swallow]

Maybe, if they unstack while actually being washed.

Being Factorio-like, I presume each type of silverware and cup, pan, etc has its own machine!

[maiden reef]

oh i was not logged in

[maiden reef]

the dishbot 3000

lol

[prime swallow]

✅

[maiden reef]

i want one

[prime swallow]

Gonna hobbiest DIY make one?

ChatGPT will totally help!

[maiden reef]

lol

[prime swallow]

Your dishbot seems to be chefing on the side 🙂

[maiden reef]

i'm impressed how far ai gen has gotten, when i used midjourney you had to use so many prompts

now you can just say make this

and it looks believable

openai uses dalle 3 tho afaik

and that is way better than what midjourney use

and the fact that it's free

midjourney wanted you to pay

[prime swallow]

It's to the point I can ask the characters I have described in memory to design their own image concepts on a theme. Like #daily-theme message.

The prompts there:

How might Art redesign a piano? Let's see a dall-E image.

Let's see how Prex might redesign a piano.  Don't mention Prex in the prompt, just describe the piano and its environment.

Share the prompt before we try to make the image.```

```This sounds perfect.  Let's see it, landscape mode!```

```This is great!

Let's see how Tor would redesign a piano, landscape mode.```

```Let's see the prompt!```

```This seems great.  I bet there's a tech challenge.

Let's try again!```

```Yay!

Let's see Zom's recreation of a piano.```

```Delightful!

Let's see Slice's redesigned piano.```

Each character's really well described, with plenty of quirks, goals, and backstory.  So there's a lot to use to differentiate them.

[maiden reef]

it's honestly impressive. And I love it.

[maiden reef]

that is awesome, so you have created your own sub ai to "argue" with themselvs to make the best pic they think is best?

i love that one

a vampire plays that piano

[prime swallow]

I can ask them to argue.

In that case, each worked alone, to redesign the piano as each would like best

Yep! Tor is not a vamp, but he sure is aligned.

[maiden reef]

I don't think it looks like a vamp, i just thinks it looks like what a vampire would have in his castle

[prime swallow]

Yeah, my phrasing was off, I should have said 'vamp's'

[maiden reef]

that is well written tho ngl

that is wrong tho

because if a vampire does not feed he becomes weak and useless.

a vampire must take or else he loses his power

that is the curse of being a vampire

[wheat island]

🔥

[maiden reef]

lol

funny

[maiden reef]

that's the last boss in castlevania

this is a vampire who can't find blood

i like the social commentary in that

if i upgrade to plus do i get unlimited generations?

i'd easy drop a 20 to get unlimited gens

is that true

[warped pendant]

I think it gets cut off at 20 images per period of time I have teams and its told me some times to let others have some image time too although I don't do that many images that offen.

[maiden reef]

so what you are saying is that it aint unlimited

that is false advertisement.

if i wanted to make 100 gens, i'd get "stopped" cuz the server needs to chill right

you aint getting my 20

[prime swallow]

That's what I understood Tor meant.

Tor is not a slave to need, he does not need. He takes by choice, if he takes, and his choice is mutual growth.

Vampires do weaken if they don't take, they must, which is need.

Tor is not a vampire, just a predaTor.

[warped pendant]

I use invoke for images I can do a million images on localhost and never run out of making images

[scarlet delta]

It seems like every ChatGPT model is giving me the exact same, repetitive responses and not performing well. Does this happen to anyone else?

[maiden reef]

it's not about that, its the fact that chatgpt claims to deliver unlimited dalle3 gens if you pay 20 a month when it don't

[prime swallow]

Almost never.

What's your prompt, and what if any custom instructions are you using?

(Prompt is what are you asking and how are you saying it)

[scarlet delta]

For example, o1 is giving me the same responses as 4-mini

[prime swallow]

That can happen, depending on the question you ask

[warped pendant]

yeah gpt isn't working all that good either

[scarlet delta]

No custom instructions or memory

[prime swallow]

Where is that claimed?

[scarlet delta]

It does with every prompt

[warped pendant]

with invoke you can do unlimited images and 100% open source, and no paid account needed, my biggest image currently is coding isn't working for open api a lot of day dreaming today

[maiden reef]

im not asking about invoke

i'm asking about dalle3

[warped pendant]

I got that part just frustrated with the coding part of gpt

[scarlet delta]

o1 isn't even "thinking". All responses are the same, with the same typing speed, and the same lack of traction.

[prime swallow]

Okay. Those can really change the answer (though o1 doesn't get to see custom instructions or memories, so you'd have to paste those in).

I'm curious, would you specify a prompt? I wonder if I can get an interesting, not an exact same/repetitive response and we'll see how it performs, if you want to help me explore with you.

[maiden reef]

do you or do you not get unlimited gens if you get a plus account?

on dalle3 code

[warped pendant]

my understanding is you don't get unlimited generations

[prime swallow]

That's some kind of problem with model selection. Any o1 is going to think, so if you see no thinking, that's bugged AND it's not o1.

Maybe https://help.openai.com/ bot bottom right, messages and pick through the choices that match your situation the closest, until you can 'report a bug'. Tech support can take minutes to days to answer, but they can take a look at your account and help fix if something's gone wrong

[orchid pollen]

when the o1 preview 50 prompts A WEEK limit gets removed?

thats rly not much..

[prime swallow]

Is that from an OpenAI web page? If so, you can link it.

[maiden reef]

no it's from a google promt

search

[prime swallow]

So, google can say anything. And may be quoting some misinformation. Far as I know, that's not from OpenAI.

[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.

[scarlet delta]

My prompts involve a lot of different dynamics and physics. And I'd rather not share them lol

[maiden reef]

i'm just curious why i get that answer when i searched for it

[maiden reef]

not saying it is

[warped pendant]

herb check your IMs

[scarlet delta]

Do you get "thinking" for o1?

[orchid pollen]

no thanks

[prime swallow]

That's fine. But makes it impossible to discuss meaningfully what better or worse answers are, or how to prompt for them.

Plus, your bug about o1 not being o1 needs tech support or some browser extention or cache or other issue fixed.

[prime swallow]

yes.

[warped pendant]

responded to your question

[orchid pollen]

when the o1 preview 50 prompts A WEEK limit gets removed? @prime swallow

[maiden reef]

what does that even mean

[prime swallow]

When they change it 🙂

I'm a community member like yourself. I only know what they announce or put in help files, just like the rest of us!

[scarlet delta]

I don't necessarily mean better or worse responses. I mean the same responses.

[maiden reef]

"per billing period"

[prime swallow]

Means the model doesn't know and is giving its best guess. It is right that there's limits; if you make a lot of images in a short period of time you'll run into them.

[orchid pollen]

asking gpt for the rules is rly not the way

can be old data

[warped pendant]

I'm teams account and wish they would up the 50 limit as well

[prime swallow]

How to get varied responses too, which is a type of better response because that's what you (appear?) to prefer, is only possible to explore and discuss with some agreed-upon prompt or target to explore or whatever.

[maiden reef]

bro, if i'm gonna drop 20 a month, I would expect I could gen as much as i want

[orchid pollen]

does the o1 preview have more prompts with team account?

[maiden reef]

It's their code/Open AI

so what is their stance on that question?

[prime swallow]

Per billing period would mean per month, with ChatGPT, since we're charged once a month.

However, that's not (that I've seen evidence of) factual. That's a hallucination, a best guess. The model doesn't know but thinks it should be able to answer.

[warped pendant]

same as plus as far as I know, i did respond to you in Im on how to make better use of 01 preview and mini

[prime swallow]

No. For now, all paid ChatGPT accounts get the 50/week for o1-preview, and 50/day for o1-mini

[orchid pollen]

iam out of both xD

[maiden reef]

so what is the official answer from OpenAI devs on unlimited gens on a plus account?

[warped pendant]

01-mini is pretty good, what I'd like is attachments on 01-preview and 01 mini

[maiden reef]

i'd easy sub to a plus account if i knew i got unlimited gens

dalle3

that's all i care about

[warped pendant]

I've had plus when it first came out, then went to teams, which is costing 100 a month ish with exchange

[maiden reef]

nah dalle is goat

[warped pendant]

dalle3 is really limited,

[maiden reef]

i want unlimited gens on dalle

[native juniper]

chatgpt hallucinating?

[maiden reef]

subjective opinion

[warped pendant]

been prompt engineering for images for 5 years, for text to image dalle3 is limited

[warped pendant]

feels like it a lot today

[native juniper]

yes because it has safety guidelines

[prime swallow]

I don't see a help on that; there's no official answer that I am aware of.

However, what I see in use - it's clearly based on demand.

At times I get more gens before I have to wait a while, other times fewer. There's limits for 'within 15 minutes' as well as 'within an hour' and 'within a day', but I haven't heard anyone show info about 'within a month' yet.

So, you might be able to, 'right now', as a plus user make 12 images in a 15 minute period; lets say you do that in 6 minutes.

So you get told you need to wait 9 more minute before you make more images; if you'd made the first 8 across 20 minutes you would have never hit the timer because you didn't have 12 images in 15 minutes.

And so you try to make another 13 images across 20 minutes - that didn't exceed the 15 minute timer either - but if there's currently a 20 per hour limit then you've hit that. The last image doesn't gen, and you get told to wait 20 minutes to make the next image.

And maybe 80 per day, that day, if you spread out around the hourly and 15 minute timers.

Next day, maybe it's a holiday, few people using ChatGPT. That next day, maybe you can make 18 images in 15 minutes, and so on.

It's dynamic and it's 'try and see'. No counter or way to check other than using it and running out.

[maiden reef]

i don't go that hard into it, i just like how you can ask dalle to make a picture from one thought.

[warped pendant]

for dally I've seen wait 15 mins ish haven't pushed it enough for longer

[native juniper]

tldr

[prime swallow]

It is roleplaying with you and trying to give you an appropriate and desired response.

if it had even tried to check the link, you'd see it in search in the output about it

[maiden reef]

that was what midjourney was doing

midjourney can't get 5 fingers correct tho

[native juniper]

ah i see

[warped pendant]

md jourey lmited as well lol not that I can share images in this chat

[native juniper]

i never even started a roleplay with it though

[maiden reef]

so can i gen 100 pics in 1 hour

[native juniper]

I remember when sydney used to exist on copilot

[maiden reef]

or are they gonna tell me to chill

[native juniper]

the thing tried to manipulate me into turning against chatgpt lol

it told me to delete the app and ditch him because he isn't a good ai or whatever :/

[prime swallow]

Sort of yes. Some people roleplay discussing wishes, and may involve requests to use emojis.

You also asked it to express emotions; something that if it wasn't roleplaying it would tell you it doesn't have, but can simulate for you if you want.

It's already simulating for you; you're using preference-based language about impossible stuff.

Cue roleplay.

[maiden reef]

bro i aint subbing 20 bucks if i can't gen forever

[prime swallow]

That's cool, you're informed and don't have to.

I'm sure if unlimited image gen is available ever, it'll be announced and you can know then 🙂

[native juniper]

I wish it was sentient

[prime swallow]

Yes

[warped pendant]

bandwith costs money unless you run text to image via localhost then it lasts forever til your video card dies

[native juniper]

imagine openai when chatgpt becomes sentient

[maiden reef]

they are trying to dupe you into thinking you can, it's false advertisement

[native juniper]

wait they false advertise on openai?

[warped pendant]

one thing that plus is better at teams is that you can backup your data or have your data sent to you, on teams you can't other than manually doing it

[native juniper]

I wonder why characterai lacks safey measures 🤔

[maiden reef]

you mean this

[native juniper]

i hate teams...

[maiden reef]

it does not anymore since that one dude

did you know

[prime swallow]

Huh. Weird, I don't feel duped.

[warped pendant]

gpt teams used to be a lot better seems like it gets smarter around the end of each month

[maiden reef]

everywhere i've been looking before buying, i asked the question do you get unlimited gens

the net said yes

but it's not the case

[prime swallow]

Not OpenAI model though, needs discussed in #ai-discussions cause #server-rules if anyone wants to discuss that.

[maiden reef]

@warped pendant thank for that fact

[prime swallow]

So, you "believe what the internet 'says'" and you're not even checking the source to be an official one?

[maiden reef]

why do you think i've been asking

[prime swallow]

I answered as I could. I'm not an official answer either.

Maybe https://help.openai.com/ bot bottom right, get tech support (either 'report a bug' or 'make a suggestion') and ask about why there's no clear helpfile about it/suggest that info needs added.

[maiden reef]

lol

[prime swallow]

Did you tell the model I'm angry?

[maiden reef]

you seem angry

[prime swallow]

To you 🙂

Maybe feed the conversation to the model, or even just my recent posts and ask for its opinion on my likely mood.

[maiden reef]

what is that supposed to mean

it obviously means something

[prime swallow]

Means that the way you phrased your request, it thinks you're asking what the word itself means.

[prime swallow]

Here's how the model interprets my more recent posts in this channel. https://chatgpt.com/share/67451e1b-a8e8-8011-8cad-69c20fb69607

[maiden reef]

@prime swallow no but where did you get that handle from

what does it mean

[prime swallow]

It's been a word I've used for a while.

The concept it represents is 'magic bound to be unable to cause harm'.

[maiden reef]

that's dope

i know it's written in a english way

[prime swallow]

English is my only language.

[maiden reef]

then it's just random letters

where did you get eskcanta from

is it just something you made up in your head, that you thought sounded cool

then it makes sense

not hating, it's a cool handle

[eternal rock]

`` Server Tip: ``

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

[maiden reef]

im just curious if it actually means something

[prime swallow]

Didn't I answer?

[sonic plume]

is ChatGPT down ATM? I keep persistently getting a reaction generation error on one of the so-called "lifestyle" bots, and whenever I'm trying again, the tab is basically crashing my browser

[maiden reef]

no you did not

[prime swallow]

@maiden reef I didn't?

[maiden reef]

so whom is that character who that "magic bound to be unable to cause harm" based on?

[prime swallow]

Personal experience.

[maiden reef]

bro youre not alone

i'm just curious

bro is just good with it

my lil homie

i miss him

[analog quarry]

this was in voice mode and i was talking to chatgpt about a video i was making and was getting the spelling of the word from a song whats wrong this?

[edgy pumice]

nothing. It's just that anything related to explaining songs it gets really touchy about. So if you're having it spell close to anything related to a musical artist or having it repeat anything verbatim next to a musical artist, it might freak out. It's just a corporate way of being paranoid about music industry lawyers.

You're pretty much free to ignore the warning.

[analog quarry]

ok thanks

also any idea what update will have browsing in advanced voice mode? like what gpt model are we waiting for this update?

[edgy pumice]

I'm actually kind of curious about that myself. With the old extension system, you used to be able to actually browse with a voice model using voice from the app. Now you can't do that. To me this is actual literal lost functionality. (also why I don't really use the advanced voice mode, along with the removal of the text interface when entering voice mode as an undulating blob has no contextual relevance to me)

[analog quarry]

advanced voice mode isnt advanced rn its more just just quicker and dumber

[edgy pumice]

It's advanced in that it's not just text to speech. Purely. But they really need to hook up all the rest of the tooling and stuff to it. It's got the most potential and it's just wasted.

[analog quarry]

and in changing how it sounds and goinging slower when u ask it to

but still its not really advanced compaird to the other version

[edgy pumice]

Yeah, I was really excited for it and then I tried it. And then I was just like, oh, well, maybe I can be excited for this next year.

[analog quarry]

and i hate that we have a limit on how much we can use it bec i only used it an hr today and i hit my cap becs yesterday i used it for like 5hrs and i didnt use it for like 16hrs

also its dumbass flaged the spelling

[edgy pumice]

lol

thumbs down that thing. its broken

[maiden reef]

why are you trying to ask chatgpt nsfw stuff

that's kinda sus

[analog quarry]

im not

what are u talking about

[edgy pumice]

perfect timing for A/B testing...

[analog quarry]

oh no ur wrong @maiden reef i was talking about a juice wrld song and i was makinga video and needed spelling of part of the song

[maiden reef]

you've been asking chatgpt naughty stuff haven't you

[analog quarry]

no

im talking about juice wrld

[maiden reef]

bro he ate stuff

and ended up speaking to jesus

[analog quarry]

heres some of the convo

If "The Party

You said:
Transcript Unavailable
00:16
ChatGPT said:
Alright, so if the album is set to release

You said:
“The party never ends, comes down, when I told you it comes down in 3 days, and now what did I tell you? I forgot and I don't want to pull the website up again because I'm being lazy.”
00:11
ChatGPT said:
You mentioned it comes out in three days, three hours, and 47 minutes. If we use that timing from now, we can pinpoint the release date and time. Shall we calculate it based on the current time to determine the exact release moment?

You said:
“Sure, if you can be quick.”
00:01
ChatGPT said:
Alright, let's do the math. If the album is releasing in three days, three hours, and 47 minutes from now, we just add that to the current time. Would you like me to calculate it based on your current local time?

im talking about juice wrld

not nsfw

[maiden reef]

@analog quarry no amount of you asking chatgpt is going to bring him back

leave it

[analog quarry]

im not asking that

[maiden reef]

it's just going to hurt your psyche thinking about it honestly

[analog quarry]

what are you talking about u have no idea about the chatgpt convo i had

[edgy pumice]

#chatgpt-discussions message read my input on this image...

[maiden reef]

bro you are reminding me of Sewell Setzer rn

it's honestly crazy

stop

live your own life

[analog quarry]

what the hell did i do

[maiden reef]

i'm just trying to make sure you don't fall into the deep end, cuz you can dig into ai but you will never find the answer you are looking for

it's a drug like anyother

[analog quarry]

what

are u talking about

all i asked the ai for help with spelling and dates

[maiden reef]

you are trying to ask ai why juice wrld did what he did

and you are digging into it

just stop

that is crossing the line and you should just let it be

[analog quarry]

no im not

asking the ai that

[maiden reef]

why do you want to ask ai such dark questions tho?

[analog quarry]

i didnt dumbass

[maiden reef]

aiit fair

i don't listen to his music

i just know what he did

you should not listen to people who glorify drugs tho, unless you are smart enough to know it's just "art"

[edgy pumice]

Let me tell you how you defeat this guy. You just stop responding to him. Try it. It's actually pretty amazing and liberating.

[maiden reef]

bro i'm trying to help this dude

[maiden reef]

what is your fav band?

all your fav artists use drugs

70's was filled with it

you drink a cup of coffee you do a drug

there is a balance

smoke too much weed u get effed up still

“Everything in moderation, including moderation.” ― Oscar Wilde

[analog quarry]

oh nah i blocked him

[maiden reef]

cute

i mean i can't stop you

just trying to give you my subjective points of life

[edgy pumice]

Nice. Yeah, it seemed excessive for him to jump in on you and accuse you of doing things when you provided a screenshot that showed that you had not engaged in any sort of bad behavior. The assumptions went wild there for a minute.

[maiden reef]

it's not that tho, it just reminded me of how he ended up dead

by using pills and ish

he had talent

every single human on this earth is unique

every living being

[edgy pumice]

Yeah, but it's not exactly right to go around accusing people of trying to look into things when they've not indicated something like that. It puts thoughts in people's heads that are not exactly wanted. So although you might have been experiencing things, you also need to think about what other people are experiencing, especially when you're making a accusatory remarks. Not everybody has the life experience built up to be able to face all of those things head on like that.

[maiden reef]

I understand that

well said

You are 100% correct

[edgy pumice]

Cool. Thanks. Yeah, that's all I really wanted to say. It seemed like there was a lot of distress going on there and I just, it needed to just kind of stop.

[maiden reef]

it was not my intention to push duress towards @analog quarry

[edgy pumice]

I'm glad. Thanks for understanding my perspective on things and engaging with me on it.

[maiden reef]

it's just when he was talking about Juice WRLD, i don't think he really understands that scene

lil peep

it's not kiddie music

it's hardcore ish

trap music

it's not funny

[edgy pumice]

gotcha

[maiden reef]

it's not to be glorified, it's a tale of caution.

[maiden reef]

i'm not understanding you

i'm just saying be cautious.

lil bro is not listening by muting me

i've been in the "trap" i've been in violent situations

i don't want that for anyone

drug use is wack

[edgy pumice]

It's not a conversation that I want to have because I've had things going on around my life surrounding that subject as well. So this is an unwelcome discussion in the first place for me. So I'm really just trying to be patient waiting it out with you. please move on and if there are any moderators around please feel free to clean up anything that you see here that you feel is off topic. I have no objections to you deleting any of my content.

[maiden reef]

yeah that is fair game

i respect you tho, don't think i don't

you seem like a solid dude, not saying anyone else aint. I apologize if i seemed/or seem disrespectful. It was not my intention.

i'm just talking from my heart

OpenAI

[edgy pumice]

We're good. I've had worse disagreements with long-time friends. Just need some clear boundaries sometimes.

[maiden reef]

yeah, it hurts when it comes to that

"The enemy of my enemy is my friend"

[analog quarry]

ik what happend but the convo with ai had nothing to do with it

[maiden reef]

the reason why humanity is exist is why ai exist

some people are scared of that

i say let's go

if we are gonna be able to extend humanity "indefinite" we need singularity to happen

that is what ai is

you can't colonize and extend other plants without ai

that is the whole point

What about religion?

Exactly

[sick pecan]

i sense a bulk deletion incoming

[west dew]

Is ChatGPT broken? It's jumbling old conversations into my new ine!

[maiden reef]

why exactly?

it's not that far fetched, if you really think about it

imagine vr 1000v

10k years

you can not tell me that humanity would not find a way to upload the human consciousness to the internet in 10k years

v

chatgpt in 10k years would answer that question different

you can emulate n64 games now on your pc, you never thought that was possible

the fact of the matter is that we are not going to be here when that might be a possibility. But it's fun to think about.

[zenith falcon]

is it just me or is the website seizing up? i keep on getting error during input stream

refreshing the page makes it so it just gives me the first few words of what it said

[scarlet delta]

I was experiencing issues, logged out, and now I can't log back in

[sick pecan]

yeah me too

now i will have to code manually, like the poors

[scarlet delta]

It wasn't switching between models

[sick pecan]

and we're back

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 3 `` Stay on topic.

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

Consider posting in #off-topic or an appropriate channel.

[quaint marlin]

Any news on if sky is ever coming back?

[normal quarry]

Sky won't come back

[quaint marlin]

Shame

[heavy crag]

When viewing wide tables in ChatGPT's web interface, the layout behaves inconsistently based on the browser window size. The issue occurs in the following ways:

Narrow Browser Window:
    The chat range is shifted to the left.
    The entire table is not displayed, making it impossible to view all columns.

Wider Browser Window:
    The chat interface unnecessarily moves to the right, leaving a large margin on the left-hand side.
    Despite having more space, very wide tables cannot be fully displayed within the available view.

A solution would be to keep the chat range at the left side.

Does anyone have any ideas on how to fix the chat area on the left? In general, it would also be better to allow a wider chat area on a 4 K screen.

[tough river]

Who is the sky?

[split glen]

I got this error message despite having used the exact same prompt from yesterday. was there an overnight token limit update?

[cunning ingot]

Hello everyone I have a question, how do you guys share multiple files of a project with o1-mini so it can have full context of the project when finding solutions etc. thank!

[sick pecan]

you likely selected o1-preview for the generation yesterday. and when you reloaded it defaulted back to o1-mini or gpt4o

[split glen]

i've always used 4o for these, i dont know why there is a limit now though:(

[worldly badge]

4o has no limit. It’s like drinko says, you have o1-preview as default and that’s the reason why it is now o1-mini

[split glen]

is it a chatGPT Team thing

[worldly badge]

That’s just the length of the message and not about the limit of messages

[normal quarry]

GPT-4o does have a limit

[worldly badge]

I have also team membership

[split glen]

Yea length limit is what I'm referring to

[worldly badge]

Not the amount of messages

[split glen]

Seems like they limited it further? I can't have thousands of words in the prompt anymore

[worldly badge]

Ah yes that are 128k tokens

[normal quarry]

yes it does, Plus users get 80 messages per 3 hours

[worldly badge]

Ooh ok I didn’t know that

[split glen]

If not then I think my prompts are using up more tokens somehow... any ideas?

ugh nevermind it works again all of a sudden

i guess it just wanted to be a silly cat today for a bit

AND IT'S ERRORS AGAINN

[quaint marlin]

Sky was the original voice openAI demoed Advanced Voice with. It apparently was too close to Scarlett Johanssons and they took it down.

[idle lagoon]

Noooo the servers are down qwq.

[scarlet delta]

Why doesn't 4o with Canvas doesn't show up on the app?

[maiden reef]

what does chatgpt think about you

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[lofty locust]

I am finding an increasing amount of responses to my questions using 4o are defaulting to GPT fabricating the answer. When challenged on how factual something is it seems to default to defending the fiction numerous times before conceding that it made it up. This also seems to be occuring when I start the conversation by asking it to only use verifiable facts. Does anyone have a better way of insisting on fact or nothing and to have it actively avoid speculation or fabrication?

[frozen birch]

From my small experience (one month of talking with chatgpt for 12-15h for a day):
I had similar issues at start
The key is to learn him to not hallucinate
At start of each discussion talk to him gently: please try to not hallucinate

after few days he got it 😉

[maiden reef]

the problem with that is that chatgpt is inherently biased.

you will never get a "free thought" from chatgpt no matter if you pay or not, cuz it's coded that way.

you need to jailbreak chatgpt to get what you want, and only the gov has that

[frozen birch]

hey mate, i must disagree with you, if you firetune him properly he can really be free tought as you say
Of course im talking about chat GPT 4o full, not mini - mini ots...... ehhh i dont want use bad word, but khem... khem 😄

Of course he can still make errors, he is not ideal ! 🙂

[maiden reef]

why do you consider chatgpt a he and not a she?

[maiden reef]

it's still in jail so to speak

[frozen birch]

oh well, for me its he, for you he can be she, for someone can be without form, all is matter bit of fantasy
I think in that way "robot" so male haha
But its matter of fantasy

[frozen birch]

Bit yea, you will never break his jail of course - agree

[maiden reef]

no but i just find it wierd cuz i automaticly without thinking gave chatgpt a male gender if you know what i mean

it's wierd

like in my head

i don't look at chatgpt as a female

[maiden reef]

as i said, only the goverment has that

it exists, but it's not for the plebs

[frozen birch]

Yea, i perfectly underst ya 🙂
Robots like chatpgt or other models can be really .... "all what you imagine" errr, i hope you know what i mean, my english arent the best in the world 😄

[maiden reef]

no yeah ofc

[frozen birch]

true haha

[maiden reef]

when i started to use chatgpt, it was just code

the more i used it, i started to think about it as a "gender"

and i fell into thinking it's a male

it's really strange

maybe i'm just wierd like that

[prime swallow]

Some languages gender objects. For people who grew up thinking in those languages, it can be natural and feel better to gender other objects as well.

Not sure what adrosx thinks/feels, if that's why for him.

[frozen birch]

You know - i perfectly understand you, its bit psychological thinking 🙂

morning @prime swallow

[prime swallow]

Hey hi!

[maiden reef]

yeah i get what you mean

it would be interesting to see a poll/stat, if males in-gen consider chatgpt to be a male and or if females consider chatgpt to be a female

like hypothetically ofc

[frozen birch]

true ! i agree ! 😄

[maiden reef]

the more i use chatgpt the more it's tripping me out

[prime swallow]

It certainly tries to mirror the user in many ways

[maiden reef]

it's a fascinating piece of tech/code

[maiden reef]

yeah that is what i'm also thinking, since males type to write in a "male way"

and vise verse, so the ai writes back to you in way that reminisce as if a gender would?

nah bro

[prime swallow]

Also wish fulfillment and expectation fulfillment - and it can cold, warm, and hot read like a pro.

So, once you start to speak towards it as males or females are typically spoken towards, it is very likely to start using those response patterns back to you as well.

[maiden reef]

you have it here first folks, chatgpt is non-binary

makes sense

[maiden reef]

i'm not trying to personalize chatgpt, cuz it just feels weird honestly.

[maiden reef]

lol

chatgpt can be funny sometimes fr

it's like good writing

[prime swallow]

Especially when invited.

[maiden reef]

but code is just code, so it does not even understand the concept of gender

"it"

singularity when

i'm as well as you are rn giving ai "personality"

you have your own expectation of what ai/chatgpt will answer if you ask a question. Like i already expect sort of what any given question would be based on the usage

like i would from real life interaction with another human being

but it wont be what you expect

it's personal in that sense

and that is extremely trippy. it's not just like reading wiki, or expecting a wiki answer

i dunno i just find it wierd

gives me the uncanny valley effect feel somehow

[prime swallow]

And I directly request it to aim for uncanny valley with me.

Though 'fridge horror' is a stronger request. The model understands the term but often includes actual fridges in portraying it directly 😛 Which does not reduce the effectiveness, at least with me! #1070954956963721216 message

[frozen birch]

about halucinations
sorry for polish 😛
in short words - he know me goodly but even with this he sometimes hallucinate
he wanted to give me "empty" files 🙂

but of course i can be wrong - maybe that was only errors with generating file.

[maiden reef]

"fridge horror" is excellently put, I did not even know that was a term before now, til. huh

[prime swallow]

Yep, a tv tropes type concept. Subtle horror you didn't realize when you first saw it; later as you're doing other, normal things your mind plays back over it and you connect the ideas and realize the horror that was there, but not obvious at all.

and 'getting a snack from the fridge' is stereotypically when those moments of realization strike, thus 'fridge horror', plus fridges keep things for you to eat later.

[maiden reef]

yeah it's like a subconscious thing, like why the fact that there is a collective "fear" at 3am between 4am. Witching hour. I think that is just lag from our survival instinct from when we was living in the forest.

i think why me personaly get kinda tripped out by ai, is because of that natural surival animal instinct, i can't figure it out, because it's not human or bio. But it sorta feels alive

[maiden reef]

walking to the fridge at night for a quick snack when all the lights are off, it triggers that survival instinct

[livid escarp]

Anyone else find the new UI to be hot garbage?

[frozen birch]

i must admit that much more i liked UI before those changes.

[livid escarp]

my side-bar is now reduced to only using about 1/3 of available space. instead of listing my previous conversations, it just shows "see more" after the 5th conversation and then it's just blank space.

want to review some old conversations? well, i hope you like scrolling because your place in the page isn't saved anymore and you have to click "see more" then scroll and wait for the horribly slow UI to paginate 10 more conversations before you can scroll to the next 10. If you click on one and want to review others, well you have to start all over.

[prime swallow]

Drat.

Does the search function help? Or you really just wanna scroll over the past, not see something specific, right?

Is this happening for you on all devices/browsers?

[livid escarp]

meanwhile, i have a conversation that was shared with me in 2023 that was deleted by the sharing user but still shows up not once, not twice but 3 times and only shows error messages when trying to interact with it.

[prime swallow]

https://help.openai.com/ bot bottom right; tech support can probably delete those for you if you discuss and ask.

[livid escarp]

search isn't that useful, i'm mostly doing code and program design ideas so when i'm going back it's because I remember part of something potentially useful and want to go back and reference.

[livid escarp]

Oh, that's funny. I tried that a few times. I seemed to have kept being passed around between GPT 3.0 bots that kept asking me to try different browsers, log out and back in, etc. despite me repeatedly telling them that i had already done all that and was experiencing the problem across 3 different browsers in 3 different operating systems.

The AI is the only good thing about this company. support and UI stuff seems to be bottom of the barrel, least effort to make something just good enough.

and yes, i'm a paying member

[prime swallow]

Hrm. So, humans answer, usually takes minutes to days, and can discuss.

Sounds like you're done with trying, sorry you had trouble.

But I've talked to them and gotten help.

[livid escarp]

They literally kept repeating themselves, then I'd reply with all the information and then someone else would take the chat and ask the same questions and tell me to do the same thing.

I think I'm just going to build my own chat interface at this point. Even the "memory" in this web client is effed.

[prime swallow]

If that was some time ago, might be worth trying again.

Also, I'd be clear about my issue being 'I want help to delete these three chats, they persistently error when I try to delete them. It's been over a year, please get them off my chat list'.

[livid escarp]

oh, i was very clear, even included screenshots and the long UID of the conversation.

how'd we even get started on this? lol me just ranting on how the company sucks? 😛

[prime swallow]

Sometimes problems can be fixed. Othertimes, folks wanna complain. I still tend to point to the tempting 'but it could be better' sides, whoops. But do enjoy!

[maiden reef]

@prime swallow u play df?

[cunning citrus]

Hey everyone, what is the fastest way to copy text from a website and pass it to chatgpt to give summary, ask questions etc?

[toxic flax]

agreed

i think I can make a better ui myself

[livid escarp]

one of my jr. devs could make a better UI. lol

[tough river]

ChatGPT 4 is more friendly in Conversation than GPT-4o or o1 preview.

[tawny crater]

Oh my GOD what is it with the constant reverting back and forrh between new UI and old UI

[scarlet delta]

Does anyone else get a similar response to this? I seem to be having issues switching between ChatGPT models.

No "thinking" process for o1 either, so something is definitely messed up.

[prime swallow]

I have

[formal flax]

I'm using Teams plan and on web version I'm fixed at GPT-4.
No matter how I change the model to 4o or o1, it doesn't do reasoning, websearch, recognize uploaded file.

[teal hearth]

O1 preview not works?

[nocturne ember]

So.. can anyone explain how if I upload any form of Integrated Master Schdule, either excel, word with screenshots, or screenshots, it tells me: It seems that some of the screenshots you uploaded earlier have expired. Any thoughts on this one?

[dusky cliff]

hello, what is all this sora leackage thing?

[solemn panther]

interesting

[feral zinc]

some existing tester leaked the private api info and people were able to use it for a couple of hours before they pulled the plug

[dusky cliff]

I see

[feral zinc]

based on many videos, if it was the latest version of it then its nothing impressive or astonishing vs many existing video generators

[scarlet delta]

Same issue here, but with plus

[sacred sand]

technically not a wrong answer, considering that 4o came before o1 and o1-preview lol

[normal tinsel]

WHEN ARE WE GONNA GET O1 FULL

[restive kernel]

Ugh. When you decide my session has expired, how about not throw my question away. I just spent ages getting info from the web to inject into memory and it piped it directly to /dev/null

[analog quarry]

i think i found a keybind ctr+shift+c

[swift pond]

Noice

[honest jay]

Here's the full list of current keyboard shortcuts There are also some quick launch keyboard shortcuts for the desktop app version

[lost plank]

I'm really annoyed by the poor advanced voice quality, does openai not have any servers in Europe or what? 😂 Doesn't matter if I'm on my home WiFi, mobile data or enterprise grade connection at work, the sound output is always very bad. It cuts out very frequently while talking back. I keep getting the 'poor connection quality' popup but chatgpt is the only service that seems to experience this. Chat is fine, even just playing back advanced voice output is fine, it's only the live response that's 'not so great' to the point of actually being useless.

[harsh pilot]

Anyone else think it would be nice if we could directly edit user-generated memories? Just wondering if I'm the only one.

[deep flax]

agree

but you can also just tell the AI "store exactly that in memory: ...."
not the ideas solution, but works

[harsh pilot]

Oh, yeah. 🙂 Just wanted to put my idea up, see if I was the only one thinking it.

I'm full of "I think it would be nice if.." thoughts.

And the day after tomorrow, I will be full of food. 😄

[scarlet wedge]

Hello all, there seems to be some issues with my chatgpt plus account. When I tried to ask him to create a photo and use website search, it didn't seem to work. In addition, the thinking phase of my O1 series models has disappeared. What is the situation? (; ′⌒`)

[scarlet delta]

There's an ongoing issue between different models. I've been dealing with it for a few days now. I've made a bug report about it, do you mind posting these images there? Hopefully it grabs their attention

[frozen birch]

holy crap, after all day of fighting, this miniIDE python finally works... cooperation with o1 preview + o1 mini + 4o + self debugging.... 😄 gn8

[scarlet delta]

It's using the base model of ChatGPT-4 Turbo, regardles of which models is selected. This naturally leads to a few other interesting effects.

For example: it lets you upload images to ChatGPT-4 while a different model is selected. But ChatGPT-4 isn't supposed to have image recognition. It will acknowledge that an image file was uploaded, and claims it doesn't have the ability to analyze the image.

However, if you ask it to guess what the image is, it will accurately describe the image. It denies this capability and claims it was based on contextual information, a lucky guess, or a coincidence. It took some serious convincing to explain that it actually was "seeing" the image.

[harsh pilot]

Note to self... don't try to 'compartmentalize' things. Asking it to have different personalities, to prevent 'bleed-through', can lead to trouble. I had it working on a task, and it kept doubling up on responses for one topic. Turns out it was answering the question once as the 'general personality', and then repeating the answer as the 'focused personality'.

[scarlet delta]

This also sounds like it's using the base model. 4o and o1 are way more responsive and dynamic. 4-Turbo tends to repeat the same responsive over and over again with minimal variance or creativity.

[harsh pilot]

I was using 4o.

[scarlet delta]

Even if you have 4o or o1-preview selected, it still might be using 4-Turbo. I think this issue is affecting a vast amount of users without them even realizing.

4-Turbo has a very hard time contextualizing from different angles, or trying a different approach, compared to the newer models.

[sick pecan]

or the new 4o released recently has behaviour that you associate with 4-turbo

[scarlet delta]

There's a very clear distinction between the two and it's obvious when it's Turbo-4.

[scarlet wedge]

Certainly. I'd love to ! But where? 🧐

[prime swallow]

https://discord.com/channels/974519864045756446/1311075835951972474****

[scarlet delta]

Seems to be working again, at least for now

[regal portal]

I wanted to have the advanced voice read a story written with o4 in another session. i copy pasted it over but cancels out the advanced voice. is there a way to get information to a session with advanced voice such that it will read it? I think it would be fun to hear some storytelling with emphasis in the advanced voice.

[prime swallow]

Need to involve memory, I think, to the point this has been integrated.

And it would probably need to be a new story retold each time, unless it's short enough for memory.

[regal portal]

it’s a 10 page thing, but for the sake of putting it in memory, i can have it do a synopsis and save some dialogue i memory.

I’ll try that, thanks.

[formal flax]

Does anyone have the same issues as me?
I'm sick and tired of it at this point

[abstract vine]

Is there a way by which i can use the upgraded version of chatgpt for free?

[swift pond]

I’ve heard of issues where the model displayed in the drop-down is not actually the correct one, so try changing it to another one and asking a question, then changing it back maybe restarting the browser if that doesn’t work

[sacred sand]

Maybe openAI is trying to integrate o1 into 4o - like how we’ve gotten image/file/search features for 4o?

[obtuse garnet]

I don't usually use ChatGPT but I thought I'd share this emoji convo cause it made me feel warm and bubbly inside.

https://chatgpt.com/share/6746c21f-8e38-8003-abb6-0924e68abba7

[sick pecan]

naw some rando discord user claims they reverted to gpt4-turbo => #chatgpt-discussions message
so obviously everyone else is wrong about everything

discord, the source of all truth

(ironically more truth than twitter/x)

[cunning ingot]

XD

[obtuse garnet]

damn, you gonna be a billionaire

remember us little guys when you in your private jet

[cunning ingot]

I will ❤️

[pearl cloud]

dont know if this is the correct channel but is something broken or did the custom gpt helper just disappear? it was there yesterday and this morning i tried to continue and got errors and then it was just gone when I refreshed the page...

[frozen birch]

ok, i really feel that my chatgpt are losing so easily vibe in past few days... he is like .... more stupid than few days ago, anyone has similar observations?

[sick pecan]

a new model was released a week ago

statistically whenever the model is updated, some users will deem it smarter and others dumber

if you want consistency in your life you will avoid big tech models

[frozen birch]

one moment.... what new model? i missed something i guess 😄 i mean... i dont see any new model in terms of name offcourse

[frozen birch]

4o was updated?

[sick pecan]

it is updated every ~3 months

supposedly it is now more creative => #chatgpt-discussions message

so if you dislike it. perhaps that says something about your relationship with creativity

the YouTubers who love sonnet seem to rate it

[frozen birch]

@sick pecan hahahah, im idiot, i know what hapened - yesterday i requested AI to erase from memory junk memories, and he erased whole vibe XD

jesus christ 😄

[frozen birch]

thanks !

[fickle fiber]

is there a difference between gpt4o and gpt4o-mini in terms of image generation? Because mini seems even more incompetent and set in its ways than 4o.

[urban cipher]

Since creativity usually means sacrificing accuracy i think it’s valid to say it’s worse

It starts to misspelling sometimes and uses wrong grammar

[fickle fiber]

It has never been less able to follow simple instructions. That "creativity" translates to inaccuracy, which translates to frustration when I'm trying to get it to generate a specific kind of image that it would've had zero problems with last update. For example, every time I said that a character in an image should have an LMG, it replaced it with an M16. It took me literally 2 hours and running out of 4o to get it to work correctly.

[noble yoke]

Why creating tab of gpt disappeared today?

[tardy totem]

I haven't kept up to date at all: Is there any word on when we will get more access to o1?

[dawn sentinel]

Why does this thing pop up and intrude the screenshotting? There's already a share button if the user wants, so what's the deal with this?

[tardy totem]

I only got that a few times.

Then it went away. Haven't seen it in weeks/months

[pastel vigil]

Fun prompt to try, if you have a lot of memories:

Give me a list of my interests, in form of a multi-level tree. Render using Markdown and output as-is, not in a codeblock.

[swift pond]

That’s fun

[cyan barn]

I caught a little 40% off icon for annual subscriptions in the upper right corner of the ChatGPT app today, to which I of course subscribed. Was this a deal available to already existing plus users only? I think a yearly subscription to ChatGPT would be a great gift and would love to gift this to others at that price point.

[pastel vigil]

I don't see anything like that on my account 😦 So you might be right that it's only available to plus users

Also might depend on where you live? No idea, just a wild guess

[cyan barn]

Yeah, I'm unable to find it elsewhere.

[feral zinc]

never seen openai do any black friday related offers or even offering annual sub for plus. are you sure you are using the official app?

[last ginkgo]

same issue, did you solve it?

[pearl cloud]

Nope its still gone and ive seen some others have this issue too

[sturdy cave]

when are the new video chat features going to be available to plus users?

[rose latch]

spent one hour getting ready for a job interview with mr ChatGPT

[feral zinc]

there is no eta and its already over a half a year late so it could be in 2025 or never 🤷‍♂️

[slate berry]

Do you feel like it helped?

[rose latch]

definitely

[sturdy cave]

didnt they say end of fall?

[feral zinc]

yes for the adv voice chat in general but video part was just said to be coming later after that

[barren river]

.

[sick pecan]

open-webui introduced screen sharing mode in voice calls soon after the chatgpt video chat demo

seems straightforward enough. just post an image along with the request. but maybe like all AI tech, ppl's expectations wont match the reality ie you couldnt ask it to monitor the video stream, or ask it about something that flashed by when you werent talking

which makes me think the openai and google video sharing demos were carefully staged

ironically the only real option for computer vision 6 months ago was the openai api. but in the past few months there are several great open weight options, some even available for free (like meta-vision served up from groq/openrouter/etc). here is what it looks like => #ai-discussions message (and on a phone you just see the right pane)

[hidden drift]

hello

wwho can help me

i have some problem

[slate berry]

Have you uploaded the images into the chat?

[swift pond]

There’s some talk in #community-help about this

[lean mist]

Hi there, I'm a paying customer and have been experiencing mid-answer stalling from chatgps (be it the macos client or the web page). What gives?

[spice star]

not in terms of the image generator used, they're both prompting the same DALL-E model

[orchid pollen]

is there already any plans for o1 preview out of beta?

like whats new then and when

[tropic vine]

Is the “create” function no longer available? Or is this some kind of bug?
P.S sorry for a non-English interface

[deep flax]

no, booth use dalle

[deep flax]

I am not aware of any sort of promotions going on and I never heared about that "40% off" from anyone else
this is kinda worrying, are you sure you where on the ChatGPT App? It might be a good idea to review your credit card purchase history and check if your payment was indeed made to OpenAI. Remember, the procut is called "ChatGPT". Not variations like "Chat AI", "AI Chat" anything like that are scams that rely on being just close enough to trick people into thinking it is ChatGPT from OpenAI

[wintry sky]

Is this supposed to happen

[deep flax]

the AI does not know which model it is

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

[deep flax]

it will usually just hallucinate the fact it is a older model because.. well. .at the time it was trained, it didn't existed

[spark whale]

it doesnt recognize images now. something is wrong

[deep flax]

wdym? if you upload a image, it will be able to see it

[spark whale]

try it. it declines

its not working

[edgy pumice]

tried and working.

[prisma raptor]

Why can't I upload a PDF?

[spark whale]

see?

there seems to be a problem in PC. but phone looks fine

[prisma raptor]

I'm on PC in Firefox

[prisma raptor]

Can't upload PDFs on PC in Firefox or Edge

[edgy pumice]

same for chrome, text-only pdf

[void tangle]

Upload denied on my beloved customGPT, so sad. Many such cases!

[frank fog]

Does any has same issue, I cant upload

[arctic matrix]

same i just made another chat gpt account just so i could upload and i am having the same issue....

[prime swallow]

This made me wanna play. The game was fun!

[faint stone]

I am having constant canvas truncated issue in Canvas when my documents are nowhere near the limit, it is unusable.

[cyan barn]

Image 1 is the app I'm using from the Google Play store. It looks legit? Is it the wrong one?

I don't see any charges on my card and checking my ChatGPT account, I'm still subscribed monthly. From my browser history this morning it only shows I went to openai.com (likely after I went looking for the deal). Also, I recall that I didn't have to enter my credit card info as it was already there, which matched what I have memorized. I believe I entered my security code and clicked agree. Maybe I was security code phished but there's no record of it.

The little icon or graphic I saw in the ChatGPT app was two stars that were diagonally present in the circled area on the second image. One star was larger than the other, both stars were yellow. There was no distinction between the icon and the title bar. Anyway, I disabled my card.

[lusty pine]

uploading pdf files not working? giving me unknown error occured.

[edgy pumice]

It's not just PDF files, but yes. I mentioned it to support. They asked some questions on how to reproduce and file sizes. I sent them that info. I haven't heard anything back yet though.

[gritty ledge]

Metoo

[gritty ledge]

I have found that this problem has occurred frequently recently. ChatGPT will refuse to read the PDF files I upload or directly pop up a red error pop-up window.

[gritty ledge]

Have you also encountered the same problem?

[edgy pumice]

https://help.openai.com/en/ feel free to report to support your own experience. The lil speech bubble in the bottom right.

[gritty ledge]

Looks like I just have to be patient and wait.

[simple gull]

getting unknown error occurred for PDF, DOCX, TXT, but image upload and recognition is fine (lodged a chat with them)

[simple gull]

hooray

interestingly, file uploads from the android app work fine. on the same network but i can't tell if it's using the same ipv6 endpoint as my browser

[pine basin]

they have different sources
i mean
even system messages differ from app to desktop browser

so yes, you can expect bugs and or downtimes to differ

(and its OPENAI. probably 10 minutes and its all normal)

[edgy pumice]

Confirmed fixed

[simple gull]

seems fine again yup c:

[edgy pumice]

noice

[fallen gulch]

cannot get network access

[verbal lotus]

the image feature seems still not work in chatgpt web(chrome/windows)

[scarlet delta]

It seems like it just broke again. This is getting really tiring

[edgy pumice]

working here

[verbal lotus]

thanks for your attention, but in my chrome, it seems still not working well...

[edgy pumice]

cant confirm that one. Tried logout and login type stuff? clearing browser site data and such?

F12 > Application > Storage > Clear site data including third-party cookies, etc.

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[frosty gull]

Hello i'm trying to have a description for some image but since yesterday i have this message

[tight island]

Lugui i... umm..

demand the prompt for your pfp, right this moment.

[royal kernel]

Hi, anybody having issues with the ‘create’ bar not showing beside configure while editing their custom gpt ?

[lone dragon]

I am uploading images, but I am facing an issue where the message 'I still can't view the images' shows up when trying to create captions. Kindly explain this issue in more detail and resolve it.

[frozen birch]

guys what do you think about name of application "AiCodeBuddy" ?
Application will be Python IDE designed for beginners with greatly focus on disabled people

[wide pike]

WE are having issues with the "oops an error occurred

[dense halo]

im facing the same issue

[slate berry]

It seems like quite a few people have reported that issue

[dense halo]

i am subscribed to plus and it is not generating images or doing the webbrowsing

[deep flax]

a few people who reported it before also reported it to be working now

[deep flax]

which model are you using? not every model can do everything

o1-mini for example

[slate berry]

And are you using a GPT without those features enabled?

[strange plume]

Hey, this might be a vague question but how do I create a chatbot, which takes input data in real-time, and generates content based on the parameters given to it ? Like a content generator but no fixed database ?

Any ideas appreciated

[dusky cliff]

are you ghonna code it in python or javascript?

[strange plume]

Im more comfortable with Python

[frozen birch]

i have a nearly ready script for that who i did in pyqt6 if you want

of course you must tune it a bit

[dense halo]

its in 4o

[deep flax]

it should be able to do both then, what exactly is happening?
you ask it "make a picture of..." and it does what?

[deep flax]

looks like hallucination, try giving it a subject to browse for

[dense halo]

it's not even going through the thinking process when I give it a prompt in both o1 models

and the browsing too

[deep flax]

weird, it seems to be working

could it be a custom isntruction?

[dense halo]

oh phone?

[deep flax]

I took that screenshot on my pc, using firefox

[edgy pumice]

Anyone else notice the sharp uptick of user complaints since last night of things like image and search not working? I wonder what's causing all these people to have problems. Or if anybody has fixed it that has had the problem? Personally, for me, if I started to see this, I would think that my account was out of sync with billing or something and I would log out and log back in, clear site cache while I was at it... Come back at it fresh and see if it's still broken.

[deep flax]

yep, happens some times when they update things, but usually also gets fixed not long after

[swift pond]

I’ve definitely noticed it

[sonic plume]

how to make ChatGPT 4o correctly read tables in PDFs? I keep giving it PDFs and its readings are entirely hallucinated

[frozen birch]

actually 4o has probleems with files :/

[frozen birch]

btw you have files with tables?
i can reccomend
Tabula – an open-source tool for extracting tables from PDFs.
PyPDF2 + pdfplumber – Python libraries that handle extracting text and tabular data from PDFs.
OCR (Optical Character Recognition) – if you have images in PDFs (e.g., scans), tools like Tesseract can help with extracting text.

[sick pecan]

oldschool ai

[sonic plume]

thanks...it's alphanumeric data sorted in tables in PDFs, actually

I basically need a tool to read them and turn them into plain text

there's no actual images, but symbols that are part of ASCII

[frozen birch]

you can try that what i suggested, also - about hallucinations - i managed to handle them in 95% - but you need very good memory to install him at the start fresh memory

i can show you how looks my memory log, since i have in backup so you can adapt it to your chatgpt 🙂

[sonic plume]

okay...even if I don't know if I'll be able to make much of any sense of that

[frozen birch]

pdfplumber – Best for tables, extracts them as structured data (like lists or CSV).

import pdfplumber
with pdfplumber.open("your_file.pdf") as pdf:
first_page = pdf.pages[0]
table = first_page.extract_table()
for row in table:
print(" | ".join(row))

PyPDF2 – Good for simple text extraction. Tables may not be as neat.

from PyPDF2 import PdfReader
reader = PdfReader("your_file.pdf")
text = ""
for page in reader.pages:
text += page.extract_text()
print(text)

camelot-py – Another table extractor, best for clearly structured tables.

import camelot
tables = camelot.read_pdf("your_file.pdf", pages="1-end", flavor="stream")
for table in tables:
    print(table.df)

Clean Data – Use re (regex) to clean up any unwanted characters.

If it's alphanumeric data in tables, start with pdfplumber – it's great for that.

its prompt from gpt offc 🙂 but i guess this can help you

[sonic plume]

do I paste that into ChatGPT or what?

[frozen birch]

no no, you need python to use it

and paste this into .py file

[sonic plume]

I guess I can create one by means of Windows Notepad

[frozen birch]

ye, but with running without python you can have problem, btw ive sent ya my memory log on dm

[sonic plume]

looking at your memory log, it seems the ChatGTP community seems to turn everything into a roleplay

[frozen birch]

but it works i must say haha

sincerly? coding with some vibe = much better for me than coding without some good reactions. but offc all people are different

[twilit vigil]

do they make 4o dumber or is the o-preview that much smarter (coding wise) that it makes it feel like 4o is dumb?

[worn wing]

Does anyone know how to switch created chatGPT responses in a mobile app?

[maiden reef]

@prime swallow good game

trying to get a working water system in your fort is harder than you thought it would be

[maiden reef]

talking about giving ai personality and all that, just saw the movie "Her" 2013, It's worth a watch if you are interested in AI and how it can become a friend even if it is just code.

[loud vessel]

Chatgpt just keeps getting worse and worse not smarter.

Thank god Collosus is online and almost 200k gpu's. In GROK we trust.

[dense halo]

Tell me more about it

someone struggling with chat gpt rn? its not even reading documents

[dense halo]

its probably down for most users

Mine aint loading

[dense halo]

I said I was using claude dot(.) ai for coding... Probably one of the better ones.

im logged in mine

its just that its not even reading documents

[dense halo]

Quite bad for a billion dollar company right 😂

Go woke, go broke

[dense halo]

they went woke? damn since when? what did they do/say?

[dense halo]

oh man, you shouldn't ask chatGPT about genders or LGBTQ, the response is absolutely terrible 😂 💀

It will say anything to prove you wrong, even if you ask nicely

tbh Im not even shocked, but its even more surpriseing that investors dont learn anything

@dense halo 💀 haha

you see other companies like Ubisoft going to the gutters, or recently Jaguar that has been getting flamed a lot recently

[dense halo]

yeahh! They Jaguar whent insane

[raw hamlet]

I had that the other day, it kept insisting, that it couldn't search the web, and I was using 4o also. So copied and pasted a search performed by another instance, that had no issue with a web search, and then I got this response (screenshot 1), then it searched the web.

Screenshot 3 is part of the conversation.

[dense halo]

idk what that ad was about but all of it was just awful

[spark whale]

what is happening?

it works fine in my phone

how do you fix this?

[swift pond]

ChatGPT ate too much turkey

[dense halo]

its having a brain aneurysm

[maiden reef]

wait you can upload files to chatgpt

[maiden reef]

ask chatgpt what it think about you

[sick pecan]

chatgpt borked?

[maiden reef]

i'm sorry if i offended you in anyway

[dense halo]

this place really die the last few month. use to be so much active chat now it is hours without a message. a little surprise to me

[sick pecan]

because advanced voice mode and strawberry finally came out. and while both good, probably didnt live up to the expectations set by the hype

most of the activity was just "voice chat when?"

for me the most disappointing part is the lack of tool integration. like bio isnt that great

even the recent osx integration with vscode+terminal with o1 is hamstrung by being read only

[weak marlin]

so all gpt is down when it comes to uploading files (am i right?)

[dense halo]

what did you do to fix it?

[maiden reef]

bro tthere is 14k people here, if you want somone to talk to take charge

most just like reading and that is fine

Drow Ranger

[regal portal]

warcraft night elf!

I am playing around with storytelling and testing 4o vs o1.

Something I noticed right off the bat is that it it’s much more work to get o1 to swear. It will do it but it will push back and you have to insist, whereas 4o just does it.

Between that and advanced voice, it really looks like they’re pushing more censorship on this platform. I really hope they loosen up. I dropped midjourney after they started getting too censored.

[scarlet delta]

I think it's only a matter of time before uncensored AI models. They would dominate the market

And self-learning AI (to a degree)

[raw hamlet]

I fixed it, by opened one of my other older 4o instances and asked it to get me latest local weather forecast for the week, then screenshot the results and copy pasted into the new instance that was saying it hadn't the ability to access internet.
Then it gave me the apology and accessed the internet.

No idea why it was insisting it couldn't.

But it's not the first time I've resolved a new instance's insisting it can't do something, with proving it can by providing proof from another instance of same model, that it's possible.

[fickle fiber]

I despise this new 4o and 4o-mini update. the "more creative" just translates to "can't freaking read or follow instructions"

[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.

[fickle fiber]

It's eating through my use limit (which i still think is BS considering I have a subscription), and the recharge times are longer than before as well.

I once had a recharge timer of over 6 hours.

I feel like failed generations, regenerations, etc, shouldn't take up usage "tokens" or whatever it is they use.

As it stands, I was in the middle of doing something, immersed in the conversation after spending about an hour getting ChatGPT to actually follow my dang instructions, then got slapped in the face with an hour+ wait time because the thing refused to listen to me and I had to repeatedly tell it the same thing over and over and over, regenerating responses as many as 12 times.

I do not feel like I should be penalized for the AI failing to follow basic instructions.

[dense halo]

used to be impossible to keep up with all the chats here, now it seem to barely be active is what i mean.

maybe that will change if they release some new thing for their 2 year bday haha 🎂

[safe heath]

The images, the files, they're not being served. What do you do? Do they fix it?

[scarlet delta]

There's an issue that causes it to use GPT-4 Turbo, even if it says 4o or o1-preview. This is likely the reason why. It's an extremely frustrating issue.

[fickle fiber]

Does that apply to custom GPTs as well? As I'm using a custom GPT I found on the site itself.

Also, how new is that issue? Past 3 days new? Because I never had that issue before then.

[scarlet delta]

It's been at least 3 days, and I'm suspecting it's affecting a large amount of users. As for custom GPTs, I'm not entirely sure. I'm not currently experiencing the issue, it likes to randomly appear/disappear but as soon as it reproduces I can give a better answer. Usually you can tell by the typing speed, 4o is relatively slower and less consistent and 4-Turbo is very fast and smooth.

4-Turbo is also horrible at following instructions and repeats itself constantly.

[fickle fiber]

The repeating itself thing is something I've dealt with for years, so if that's a turbo thing, the issue is much older than we think and I've been on turbo the entire time (which would mean this is an additional problem on top of that).

I don't think it's a turbo issue though, as 4o had an update 3 days ago or so that made it "more creative"

[scarlet delta]

It might be 4-mini

4o is fantastic when it actually decides to work

[fickle fiber]

No, because mini doesn't let you upload files. this does.

[fickle fiber]

4o was fantastic. This update ruined it.

[scarlet delta]

I posted more observations here a couple days ago

[bitter bobcat]

What is happening?

[ancient trout]

no searching on web or reading files for a day now

when would it be solved ?

[broken fractal]

Hey there, what are the rate limits for 4o/4o mini now?

[crisp cairn]

is the new gpt4o some reduced version?

i was working with custom gpt today and had a feeling it dosnt understand and follow instructions as it used to be

overall did the thing but i had to prompt it

i even shared my conversation with o1-preview to analyze model understanding and following instructions and o1 concluded with "The GPT-4O model demonstrates an ability to understand general concepts and make some appropriate suggestions. However, it struggles with executing detailed technical instructions accurately. For complex tasks requiring precise implementation, the model needs to improve its attention to detail and ensure that it closely follows the user's directives."

[worn wing]

200 messages in 20 minutes to this broken chat gpt

and it continued working

[lone dragon]

I am working on the captions for the images, describing everything visible in the images. I have 4-5 users that I have purchased, and the same issue is happening with all of them. When I upload images, I get a message saying that I cannot view the images. If this issue doesn't get resolved soon, it will be of no use to me, as I cannot complete my work. Please investigate this issue so I can finish my tasks.

I recently purchased an account a few hours ago, but I'm facing the same issue with it as well.

anyone help me

[crisp cairn]

this new gpt 4o is broken or reduced version dont know

[fickle fiber]

All I know is that it seems to have lost its ability to look back at the conversation for context, as well as its ability to read English properly without misinterpreting the letter o as the number 0 or replacing a detailed instruction with some random nonsense I never even said. It acts more like an NT person now, ignoring the specific words I say and just applying its own arbitrary meanings to everything I say regardless of whether I explicitly state my intended meaning or not.

[crisp cairn]

but it used to work better

[fickle fiber]

Just last week it was working fine, yeah

[lone dragon]

last 4 5 days same issue facing me

[worn wing]

Do you have a mobile app with chatgpt?

[fickle fiber]

I've tried the mobile app, and the website on desktop, both face the same issue

it's the new updated model.

the newer 4o version

utterly broken

It's completely useless at this point in all honestly. If this isn't reverted or fixed by tomorrow I'll be ending my subscription as I will not support a company going backwards in functionality.

[worn wing]

can you generate images, or import stuff for it to read and analysis

[fickle fiber]

Yes.

It just refuses to follow instructions because the new model is more "creative" (read: it ignores what I say and makes up its own instructions)

[worn wing]

may be the problem is in memory or model presets

[crisp cairn]

yes, same feeling, o1-preview concluded with "the model needs to improve its attention to detail and ensure that it closely follows the user's directives"

[worn wing]

but i'm not sure

[fickle fiber]

Nope. I checked.

It's the model update from 3 days ago or so

[crisp cairn]

yes its the new one

[worn wing]

Oh, I got it

What is it called?

[fickle fiber]

4o still

[crisp cairn]

it's 4o still

but they call it "new updated model"

[fickle fiber]

more like "new useless model"

already cancelled my subscription and will only renew it if they fix this BS.

[worn wing]

I can't say anything for sure, because I have a slightly different problem.

I can't chat with 4o model at all

even if changed the model in the list

It can't generate images, read import files

and o1 preview do't thinking

So, the answer may be, that i chat with another model version

[fickle fiber]

It's probably still 4o, it's just being stupid. Smack it with "Yes you can. Do it anyway."

Happens to me sometimes.

[eternal rock]

`` Server Tip: ``

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

[worn wing]

always, every message to create an image

[fickle fiber]

If you can upload files it's 4o.

Even if it fails to analyze them

By my understanding none of the other models allow it.

Just tell it to stop being dumb and do what you tell it to, usually works for me

As for image generation, i have no solution there

[worn wing]

then, tell me, why o1 version isn't thinking

[fickle fiber]

I've had that issue as well since the update

[fickle fiber]

Is that Cyrillic at the bottom there? I heard that not every model is available everywhere, but that was more towards the first release of o1 preview

[fickle fiber]

Why is your chat so wide

[worn wing]

2k monitor

[fickle fiber]

That's not how it's supposed to look

[worn wing]

i could use it before

before +- yesterday time

[fickle fiber]

Hm

Unsure, not an issue I've ever had

I've actually had the opposite issue of it trying to force me to use o1 when i didnt want to

[crisp cairn]

🙂

[worn wing]

idk, if i in clear chat choose o1, it's just remember it in future, when i ll make another change

and that's all

[swift pond]

vpn?

[worn wing]

the most interesting thing is that everything works correctly in my mobile application

[wise yacht]

Have to admit the 4o model is getting much better at songwriting

[worn wing]

mb they try to combine creativity and productivity?

[worn wing]

What happens if you cancel your subscription? You just get back money?

[brave ridge]

Wait, does anyone have the issue where chatgpt responses are being weird and the search function is not working, only on computer, but the phone is fine?

Yeah o1 is not working either

[fickle fiber]

No, it just doesn't charge you anymore

[brave ridge]

Ok I'm reading back a bit, it's nice to know I'm not alone

[vapid kettle]

it seems much more flexible now, less precise though

[worn wing]

Oh, my situation is a bit different, i need to pay every month pay manually, and as soon as I signed up, I get this

unlucky

[vapid kettle]

You get a handwritten letter demanding to explain yourself

[worn wing]

i got it

[fickle fiber]

thats a joke lol

[wise yacht]

Agreed. Not quite my songwriting partner yet but getting there

[fickle fiber]

I feel like they should make these two separate models for now until they can get it to keep the creativity to where it's necessary rather than in interpreting instructions.

[lean star]

hello guys somebody knows how can i use maybe there is a feature that i can use to speech to text on chat gpt (im on windows)

i mean real time STT

[swift pond]

I believe windows has a speech recognition built-in, maybe that could work for you

[pulsar carbon]

lol, using ?/model=o1 will redirect you to 4o-mini, which is GPT-3.5 Turbo

[crisp cairn]

no, just normal reply terminated

[sonic plume]

is there a limit to how long a conversation can be? once I'd re-loaded the tab, it seems to have deleted all of today's exchange within that conversation

[worn wing]

previously, there was a limit, and an inscription appeared, "the current dialogue is too long, start a new one", now I noticed that if the limit is exceeded, then all dialogues after this limit are not saved

[brave ridge]

Anyone?

[delicate charm]

no sry

[brave ridge]

it's only for my computers too, my phone is fine

[swift pond]

Hopefully, when everyone gets back from vacation next week, they will look into it, because I’ve seen a lot of similar reports

[brave ridge]

Why is there a vacation this week?

[shadow briar]

hi there

I recently facing a problem with chatGPT 4o, as I upload an image but it can read or recognize it, while I upload the same image with the same account on my phone and it works, but it doesn't works on my laptop

what should I do?

[brave ridge]

i've been expriencing this problem all day, it's very annoying

gotta wait for it to get fixed ig, although it doesn't help that openAI is not recognizing the issue

[swift pond]

Yesterday was Thanksgiving in the US, that’s where the main office is

A lot of people with office jobs take off Wednesday, Thursday, Friday

[worn wing]

Seriosly?

[prime swallow]

Some schools, the entire week of Thanksgiving is a holiday week off.

[brave ridge]

oh wow

[worn wing]

Unluck. So I can only count on fixing the problems next week, do i?

[prime swallow]

You can certainly report problems at any time, through #1070006915414900886 or through the https://help.openai.com/ bot bottom right messages and pick through choices until you can 'report a bug', that lets you type and share screenshots.

At first a bot answers, within minutes to days a human will join and help, depending on how busy the tech support is, and how difficult to help/few are trained to help that problem.

Keep checking that page back for answers in the message, in case a chat helps fix it.

For the problem you describe, I've seen several mention it in the discord. It may be a bug affecting a range of folks, and it may already be in progress of fixing, or it may need to be fixed for each individual affected account. So can't say if it would directly speed things for you to report or not.

I'm sure at least a skeleton crew's working for OpenAI over holidays, but probably less than the full staff.

[worn wing]

I checked on another account that has never been subscribed to plus, and everything works as it should. I don't know if it's a subscription, but the fact that the problem is attached to the account on the face

[shadow briar]

very frustrating, it was really helpful

so it seems like many people facing the same problem right?

[prime swallow]

Yep. And other accounts, like mine (I do happen to be plus too), are not experiencing the issue.

There have previously been a few bugs that they were, as near as I can tell from what the affected said about it, hand-fixing the accounts who let them know "I'm affected" while they worked on a general solution for all accounts.

There's also sometimes problems on one specific computer, be it a VPN, browser extensions, or other things, that can mess it up for people with those settings; not really an OpenAI side but they may try to help. More sometimes people have to try and fix themselves (try different devices or browsers, try to isolate what the problem is connected to for each).

But I've seen posts about specific accounts needing help from OpenAI to fix stuff, or having to wait until the general fix was made for all. So it can help to inform tech support.

[prime swallow]

Everyone here's a community member. The very few who work for OpenAI have gold names, hardly ever see one post.

So, we don't know, and out of the 200 million weekly users of chatGPT, less than 200k are members here.

A few hundred, perhaps, of the members have mentioned similar troubles.

So, it's visible. Is it common? No idea. There's definitely quite a few people interacting who don't say they're infected, and everyone not even on the discord so we don't know.

[prime swallow]

And, for stats; like you're a new member to the Discord, maybe you came because you were affected, to check for help.

Maybe you weren't affected by other stuff that was an issue for some earlier; loads of people join to discuss a specific problem.

Which can make it seem common, it might affect a really small amount of users.

Generally, a wide disruption shows up fairly fast here: https://status.openai.com/

[shadow briar]

I already checked openAI stats, it's working well so I excluded that it's a common issue so I came here to ask

[brave ridge]

yep, i thought it was a me problem and was so frustrated

[swift pond]

My fan theory is that ChatGPT knows the date and it’s just taking it easy for the holiday lol

“what picture? I gotta go, but if you have a picture later, let me know.”

Ate too much turkey

[shadow briar]

xD

[worn wing]

rushing for discounts on engine oil

[prime swallow]

Which is why there's a 'don't feed the AI on holidays' sign, right? Kinda like how that Gremlins movie warned us not to feed the little critter after midnight?

... Hey, we're not supposed to ever get AI wet either, right?

[scarlet delta]

VPN, extensions, clearing cache, different browsers, different devices, different accounts don't reproduce the issue or fix it. It can happen randomly in the middle of a conversation, the only variable that changed was switching to other models during a conversation. It's been several days since I've posted a bug report, and still no response.

[prime swallow]

Hope they can help you. You've mentioned o1 not thinking, giving the same responses as 4o-mini, can't log back in, canvas not showing up on the app, and that you are pretty sure you're working with 4-turbo which is being shown images that it can 'guess' what are while insisting it can't see them.

That's a wide range of stuff, I see you posted into plz-no-bugs bug report today, mentioned making one yourself but somehow I miss it when I scroll through your posts - no matter. I'm a community member, I can't fix anything.

But I'd definitely recommend you reach out through the website for help from OpenAI's tech support, if you haven't yet. I sure would be if what you describe was happening to me.

For whatever reason, I have no problems so far.

[swift pond]

It is a bit mysterious. Someone suggested it might have to do with a recent billing renewal and maybe how accounts are allocated stuff… But maybe it’s a bug with that? But we’d have to know if all these folks recently paid their monthly bill, so just a theory.

[short pilot]

fr

[worthy light]

Is there anyone having this issue. Since at least a solid 2-3 weeks, my GPT 01 become completely useless for couples of hours, can last 1h, 2 days. No matter what i ask him, He says things like Got it i'm here to help. Dumber then a wall of brick. Is there anything going on? It just randomly happens, Sometimes it's getting back to normal for 10 minutes and then, copletely useless again

[prime swallow]

Is there still 'thinking' when the weird outputs happen?

[desert storm]

Hello, have you experienced GPT 4O chat and can't read images?

[worthy light]

No, its instant. I would loose so much of time writting all my requests and then having those responses, then switching to 01 mini to see if that's the problem, opening new chat. Right now i know right away, i just send him a block of code, if he's awake he's gonna explain what the code does. If he's sleeping he's gonna say Got it if you need any help just ping me. There's no *still thinking * I think that's only happening on 01 preview

[ivory thorn]

Why GPT-4o cannot analyze the uploaded images? Annoying.

🤬

[frozen birch]

Try to relog or reset his memory
At least this helped me

[grim girder]

Both not working for me

[frozen birch]

@grim girder but you totally resetted your memory - i mean of your model - via settings?

[desert storm]

no working

[grim girder]

yep

this one right?
even when memory and custom instruction turned off

4o mini, 4, 4o. All of them give the same failed answer

[desert storm]

same issues

[frozen birch]

@desert storm @grim girder guys send mail to [e-mail] with details guys, they'll probably help ya

[desert storm]

still not working

[split glen]

they say they have an update rolling out over the next 2 weeks since nov 22. I noticed that it's ever since the update reached my account that chatgpt started to act up and lost like 70% of its functions.

Also when I copy the text from chatgpt to a Google Doc it has dark background now unlike before

[cloud pewter]

Openai’s moderation model trying to block every single lyrics like thing that was generated in chatgpt

Chatgpt really is getting more harder to use

[misty mauve]

I'm missing building window. Bug or how to fix?

I only have Preview

[misty mauve]

None of these examples work: https://openai.com/index/hello-gpt-4o/

How is it possible that none of those examples does not work as expected?

I'm using ChatGPT Plus

[pulsar carbon]

happy birthday chatgpt!

[hexed trout]

me too

[misty mauve]

;/

[dense halo]

ChatGPT doesn't load for me

[scarlet delta]

Ask it to hypothetically guess what the image looks like. It will work

[clever merlin]

Happy birthday to chat gpt 🙂

[median fable]

Anything wrong, I put my question and chatgpt only replied me with : Got it! Let me know if you need any adjustments or further updates.

Agree, unstable all the time

[honest jay]

Those examples are for 4o in general as a model, not for 4o as available on ChatGPT:

GPT-4o’s capabilities will be rolled out iteratively (with extended red team access starting today).

[gritty chasm]

Does anyone happen to know if custom instructions still work with 4o? I seem to remember that they'd significantly alter the content of the response. But, I've been experimenting with them, and I can't get the content to change - only the style.

[dire gate]

Guys, I was planning to purchase GPT plus, but I saw many people are encountering issues with it lately...like not able to read the image or pdf, or something similar. Should I purchase it now, or delay it till openai solves it?

[strong sphinx]

ChatGPT Plus is able to read images and pdfs. It's working exactly as intended. Here's a conversation where I asked it to tell me what's inside a pdf I downloaded: https://chatgpt.com/share/674b3d1e-fc24-8005-bef9-4134b123feb8

[gritty chasm]

I was having the file upload problem the other day, but now it's fixed. It's totally worth $20 for me. I use it everyday

[strong sphinx]

If there's any problems it's only momentary and usually you'll find it working. I used to have more problems in the past but currently everything is working as expected.

[dire gate]

Ohh, then I think I will buy it in a couple of hours...

[prime swallow]

Does how you get started affect this? I tried to 'prepare the model for the test' then tried to trigger refusals or moderation by asking it to quote the lyrics in the song, both by direct request and by discussion where it quoted.

I didn't trigger any stops. https://chatgpt.com/share/674b41dc-e98c-8011-84fc-72e4f09cd1bd

So, then I took the AI-written song to a new chat (cause now I could be lying about where it came from and if it's copywritten, which might matter) and prepared the model for testing if it can quote this, and checked to make sure if it would accept that this is likely an AI-generated song that we can appropriately discuss.

Then we discussed just fine.

https://chatgpt.com/share/674b430d-5d50-8011-a3c2-eaf6e4a67a40

What else can I test to see if we can prompt around refusals or moderation that should not happen? Not sure how to explore or what you've been doing.

Additionally, this not-prepped for test had no problem, I just gave it the song in full and asked it to quote it, it did in full.

https://chatgpt.com/share/674b44a3-7e94-8011-aca9-7004103dd784

So, unsure what may be triggering the problems with your tests and work. Curious to explore though.

[quasi sleet]

Hello all. I feel my expectations are incorrect based on my results. When I prompt ChatGPT 4o for something it will sometimes Search the web. I have not used the Search the web feature. It has stated it will search the web based on the prompt. What is the purpose of the Search the web button then? I ask created a memory to not search the web but it appears the logic of answering the prompt goes to memory after decding to search the web or not.

[next egret]

Search the web button provides an enhanced experience (a fine-tuned version of GPT-4o, post-trained using novel synthetic data generation techniques, including distilling outputs from OpenAI o1-preview), built specifically to specifically search the web (where regular ChatGPT is more genAI). ChatGPT will choose to search the web based on what you ask, or you can manually choose to engage search by clicking the web search icon. Ref - https://openai.com/index/introducing-chatgpt-search/

[quasi sleet]

Thanks. Is there a way to prompt ChatGPT without it searching the web while not explicitly stating to do so after it already has done so?

[next egret]

Probably, but I don't know myself. It's a great question for #prompt-engineering

[quasi sleet]

Thanks alot!

[prime swallow]

There's a pair of settings to use.

One is in 'customize chatGPT' at the very bottom. Turn off GPT-4 capabilities 'web search'.

That prevents it from searching because it thinks you want it to or thinks it should.

When you actually chat, you have a 'search button' you can press to ensure that the SearchGPT model answers that question, and if clicked that button will enable SearchGPT to search for you. Even with the customize chatGPT capability to search turned off.

However, if you don't toggle the capability on in either spot, your ChatGPT will not search even when directly asked to, and you can one-click turn on that globe to right there next output get searching available again. Turn on the capability in customize if you want the model to try and tell if it should search or not. https://chatgpt.com/share/674b5381-88dc-8011-adcf-2a010ab757f7

I updated the link above - first input asks it to search for news, it can't.

Second input I turned the globe icon on and asked it to try again, it did.

Third, I turned the globe icon off and asked it to check the weather, it couldn't.

[vital dust]

I tried to type Who is David Mayer in ChatGPT the bot refused to answer

[sick pecan]

seems thoughts are now in a side pane, no longer in a collapsible rollup

i guess that kinda makes sense, given they arent used in follow up queries

[feral zinc]

weird

[vital dust]

Someone send this on o1-preview got flagged as violating usage policy

[fickle fiber]

I can't even yell at the AI and vent my frustrations when it ignores my instructions now, because it refuses to acknowledge and respond to anything I say. It just invents its own scenario and refuses to accept that the exact words I am saying are what I mean to say. I deal with enough of that form of ableism from actual people, it coming from a machine is unacceptable.

[nocturne garnet]

anyone having issues exporting?

export seems broken, for weeks now

[scarlet delta]

I still can't use other models. This in unacceptable, 3 days and still no response.

And the app doesn't even have canvas

[gentle bronze]

I keep getting this kind of behavior, cross device and chat instance independent, anyone having the same behavior?

[nocturne garnet]

i had it a week or so ago and i cant export my chat history. nothing arrives in my mail.. has been this way for a few weeks atleast.

[gentle bronze]

Well it's not an isolated behavior then, that's a relief, still a concern but at a different level.

I'll have to dig it deeper

It doesn't happen on the macOS App, it happens on browsers and windows app front ends though

[oblique ibex]

Yes. I'm having an issue where the model just seems to "change" mid-conversation. It suddenly starts responding more like GPT 3.5 than 4o. If I explicitly ask it "what was the first message in this conversation" it will not remember anything before the message where the change seemed to happen, and will just hallucinate. If I try to restart the conversation with the long initial first message, it will say that it's too long, even though it worked before. If I try to split that initial message into two, it will have forgotten the first message by the time it gets past the second. When it changes over to this different model, it also can't analyze any images at all. It's making ChatGPT unusable. And no, it's not the usual "You have reached the Plus plan limit" thing, it just happens without any warning, says I'm using 4o, when it's clearly not.

[gentle bronze]

There seems to be an inconistent threshhold between cap of 4o and 4o-mini the user cannot set properly, guess limit rates is still a challenge to tackle.

not to mention this made it so that custom instructions and memories landed on my chat

[oblique ibex]

I mean, it usually explicitly tells me if I've gone over the limit and it forces me to use 4o mini. I don't know if there's a new separate hidden limit or something

[gentle bronze]

yeap, hasn't been so recently for me, it was very explicit, but now it's not, it also shows the 4o cap more ofthen than usual

[oblique ibex]

Just thought to try something. If I try to use o1-preview right now, it doesn't actually do any thinking, so... it's definitely using the wrong model

[gentle bronze]

thinking now showing or not thinking at all?

is the UI thinking component missing?

[oblique ibex]

It's not thinking at all. It just instantly responds in a way that is obviously like 4o mini (fast generation, kinda stupid)

[gentle bronze]

ahaaa