chatgpt-discussions

[reef crypt]

yes i work for microsoft before

[vague falcon]

bing chat

[vague falcon]

you work for microsoft?

[sharp cradle]

He is an Indian, don't trust him

[reef crypt]

yes

[ocean escarp]

When I was reading that OpenAI spent millions on GPUs everyday as they lost profits, I thought I read somewhere it'd be likely that they'd release somewhere by December or early 2024

[vague falcon]

what does race have to do with it?

[ocean escarp]

Maybe that was only a speculation from the news articles

[reef crypt]

haha you mean don’t trust on indian

[near vine]

lets say even if you did that doesnt makes you to have the knowledge about the future, thats against the terms and condition for an employee right?

hope youre joking

[sharp cradle]

Myself being an Indian too I know he is lying,

[near vine]

how do you know he's indian?

[sharp cradle]

Like you

[vague falcon]

?

[sharp cradle]

His about, that's Hindi

[reef crypt]

hell no, i am not an indian

[vague falcon]

?

[eternal rock]

`` Server Tip: ``

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

[vague falcon]

ok?

[sharp cradle]

You are

[maiden fractal]

You can use bing

Unfortunately not as I know

[reef crypt]

wdym ?

[ocean escarp]

I don't think GPT 4 would be free after GPT 5 released. It already costs them too much GPUs power, so they can't afford more high-demand for GPT 4. Maybe there will be a higher pricing for GPT 5

[vague falcon]

wdym?

[vague falcon]

wdym?

[vague falcon]

mydw?

[near vine]

i know he never did if he did then he wouldnt have said that in the first place and asking access for gpt4 for free intead of buying it

[sharp cradle]

Exactly

[near vine]

as for somone, who worked for Microsoft doesnt take mcuh of a penny to buy a little bit amount of sub

[vague falcon]

you work for microsoft?

[near vine]

i never said that

[sharp cradle]

Also when Microsoft bing itself is having GPT4 for free

[near vine]

i am saying in general

[vague falcon]

?

[vague falcon]

?

[vague falcon]

free for gpt4 having is itself bing Microsoft when also

[sharp cradle]

Huh?

[near vine]

it does but not upto the mark

[vague falcon]

?huh

[vague falcon]

mark the upto but does it

[sharp cradle]

Extra custom bing plugins ruin it

[vague falcon]

how?

[ocean escarp]

Probably worked as a janitor for Microsoft

[near vine]

it wont

[vague falcon]

like mythic quest?

[zinc valley]

Man I’m bored

[ocean escarp]

Unaware of that term

[ocean escarp]

Right, it seems obvious

[sharp cradle]

To get more bored say this to gpt

[zinc valley]

I am right now

I might make the best custom gpt ever

[vague falcon]

mythic quest the apple tv+ show

[sharp cradle]

Also gift me a plus sub so that I can use your custom gpt

[zinc valley]

😭 I’m broke asf

[sharp cradle]

Us bro us

[vague falcon]

brad worked for mythic quest got fired and come back as a janitor

@ocean escarp

[ocean escarp]

Oh I see a reference

Except this one started off as a janitor and left Microsoft as a janitor

[vague falcon]

the show or real life?

[ocean escarp]

That real life guy who said GPT 4 will be free

[vague falcon]

brad in the show started as the head of monetization for mythic quest

then got fired and came back as a janitor

have you seen mythic quest?

otherwise none of this will make any sense

[ocean escarp]

It's fine, I understand the reference. I never paid attention to streaming services' shows

[vague falcon]

but you've never seen the show?

[sharp cradle]

You should continue in off topic

[vague falcon]

why is that when i type something in ⁠openai-chatter and click send it thinks i am trying to upload a file and then says upload failed?

[vague falcon]

#openai-chatter

[ocean escarp]

Never heard of it until now

[vague falcon]

oh ok

[vague falcon]

then how do you understand the reference?

[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.

[ocean escarp]

Because you just explained it

[zinc valley]

Alright I’m actually so bored someone give me smth to do on my phone

[vague falcon]

watch mythic quest

[zinc valley]

What’s that

[vague falcon]

a show

[zinc valley]

Okay

[rigid hornet]

has chatGPT been down for like an entire day? every time I try to use it I just get an error

[vague falcon]

maybe your an error /j?

[rigid hornet]

what

[vague falcon]

use bing chat?

joking?

[rigid hornet]

oh, I mean I could use bing but I don't pay for bing lol

[vague falcon]

it's free

[rigid hornet]

and has less stuff

[vague falcon]

it has copilot

which has a version of custom gpts

idk how it works

[rigid hornet]

It also has a very rigid personality that freaks out over all sorts of things

"personality" of course

[vague falcon]

change the converstion style

[rigid hornet]

I just want to know if ChatGPT is broken and why though lol

[vague falcon]

are you watching it?

[dawn otter]

anyone has an estimate of time for the waiting list to upgrade to gpt4?

or when will OpenAI unpause it?

[icy glen]

i'd recommend installing a local llm then

[silver marlin]

did they lobotimize chat gpt or something?!

[icy glen]

how so?

[silver marlin]

it cant folllow simple instructions and will constnatly do things u tell it not too.

[icy glen]

what are you asking it

[silver marlin]

i tell it to not write code with comments, it still does it.

i tell it to not use Chapt GPT words like delve, it uses delve.

[icy glen]

gpt 3.5 or gpt 4?

[silver marlin]

4

[icy glen]

intriguing

[silver marlin]

like its genuinely unuseable

its so bad theres no point in using what it says

[zinc valley]

Why is there a limit on gpt 4 UGH

[icy glen]

compute?

[unkempt gyro]

Is there a way to set the 3.5 to default? I hate wasting messages on the 4

[jaunty robin]

are you putting those instructions in the custom instructions box or just telling it inline

every time I've heard people complaining about gpt not following instructions, its because they've just told it the instruction inline and its scrolled out of the context window

[honest jay]

In the app, I don't think so, but on web, you could navigate directly to/change your bookmark to https://chat.openai.com/?model=text-davinci-002-render-sha and start right away with 3.5

[velvet bison]

Hey folks, is there a place on this Discord to share GPTs?

[honest jay]

Yes! #1171489862164168774

[velvet bison]

Thank you!

[unkempt gyro]

Ty!

[slow vapor]

Hi does anyone know if custom GPTs are used in training? I don’t want to give code interpreter PII if it’ll just be used to train the model

[fallow bear]

My query limit says 40/3h but I'm constantly being blocked after 15-20 queries. Anyone else in the same boat?

They seriously need to just charge us by queries once we are over the cap or something. Its just so impractical to have a conversation come to a screeching alt in the middle of a work session and having to transfer all the context to the playground

[maiden fractal]

I suggest to post in #1070006915414900886 everything you tried

[maiden fractal]

You can always switch for the Api

[fallow bear]

Oh yea that's my goto, I have both playground and Azure AI available. Its just a hassle to transfer the current context to keep working and honestly ChatGPT is the best user experience overall.

[grizzled heron]

Am I right in saying that how the GPT4 limit works is if I send 39 messages and then 1 message 2 hours later - the limit starts from the 40th message sent?

[warm gull]

think of it this way - each prompt starts a 3 hr timer, and you get that prompt refunded to your quota after 3 hrs

[fallow bear]

I can't wait for the day where GPT-4 is the new GPT 3.5

[icy tundra]

Not long. 4.5 shouldn't be that far away.

[fallow bear]

I really really like to have voice conversations with 4, it is such an unbelievable learning tool but it burns through your prompts so fast.

[silk phoenix]

You're not alone ☹️

Does using GPTs burn through your prompts quicker than using normal GPT-4?

And I do not mean creating and previewing, just simply using a GPT

[zinc valley]

I’m still trying to figure out the best way to have the ai remember conversations

[spice star]

GPT 5 is supposedly around the corner

[icy tundra]

It is. I think he said they were aiming for next year with this one.

[spice star]

I do hope we get a higher context limit with GPT 4 turbo soon

that's quickly becoming my biggest problem with the multimodal mode

ironically GPT-3.5-Turbo has a 16k context limit for plus users but GPT-4-Turbo only has 8k

my hope is at least 32k tokens

[dense halo]

I wish we could specify temperature in ChatGPT. I need to go back to experimenting with the API. I could see the difference, just not sure if it adds any value yet, but it could be useful in some custom GPT use cases, I suspect.

[spice star]

maybe that'll be a future option for GPTs

[dense halo]

I think that might be great

Anyone have problems changing the voice in the iOS app? I confirm another but it sticks with Ember.

[subtle wraith]

how much req do u have on a custom gpt?

[steep forum]

GPT4 having trouble generating images and code?

[sly eagle]

I can create and communicate with Assistants but not sure if they are the same/similar thing?

[loud night]

Looks to be working here!

[left lily]

How can we use ChatGPT with voice?

[loud night]

There should be an icon in the app!

[round sun]

what

when will we get agi

the drama ended

[outer widget]

Dang I just realized the message limit is at 40 now not 50

Is there any way OpenAI considers adding a counter to see how many of the messages we have exhausted
Seems kinda weird when all of a sudden you have run out of messages without knowing

[left lily]

There is no voice option on the Web UI?

[spring tapir]

Why am I almost always getting 3.5 output from 4.0 when I am paying for 4.0? GPT 4.0 shouldn't even be limited anymore let alone more limited.

[left lily]

If you want Advanced Data Analysis-level logic, you have to turn off everything except Code Interpreter in the GPT settings

that was my discovery

If you run a GPT with ONLY Code Interpreter on, it's as smart as ever

[outer widget]

🥴 yeah I somehow feel the quality of text-based output produced from GPT-4.0 has gone down in quality
Maybe I am missing something but it just seems to have been better before the whole turbo update

[subtle wraith]

does using dall e help with anything

[spring tapir]

Thank you! First run yielded noticeably better results.

[left lily]

No, using DALL-E makes it actually dumber

[low chasm]

anyone else having issues with the mobile app? all prompts fail with "Please try again".

Web client works fine though

[loud night]

not rn sadly

[left lily]

No problem, this was frustrating for me for several days too!

[outer seal]

it is 32k unless you still somehow have the old ui. and 3.5turbo is still 8192, plus or not

[celest reef]

Didn't they say GPT-4-Turbo has 128K?

[subtle wraith]

i have a question regarding custom gpt, is the limit set to 40 for 1 user or all user using the custom

for example user 1 used 20 req within the last 3 hours, user 2 will now have 20 left

am i wrong

[delicate ledge]

The voice speaking functionality is very very flaky today. On several of my mobile phones it just doesn't seem to pick up when I'm talking. It's just stuck waiting. Even when I manually press it and talk it still doesn't work. Sometimes I close the app and try again and it suddenly works. But then it also just get into endless webbrowsing loops. And than the next time the voice doesn't work again.

What is going on?

I'm talking about when using the headphones button.

[radiant sinew]

This post embedded

[dense halo]

I can't get it to change from the Ember voice

[icy glen]

gpt 4 turbo on its own

crushes gpt-4 orginally

perhaps its the extra promps?

[dense halo]

It might be useful if the custom GPT Conversation Starters accepted message input; I think currently clicking on one ignores whatever message is entered, but I can see use cases for including the message with the conversation starter, what do you think?

[silk phoenix]

A problem that i have consistenly been facing is getting capped at 20 messages(I counted by hand). Is this normal? Anyone else experiencing something similar? I want to post about it, but at the same time, I do not want to give the employees anymore headache😭

[icy glen]

you have to ask chatgpt to use consistent seeds...

idk which api-based model that is on-par

with dall-e 3

supports something like that

[woeful vigil]

why can't i edit my messages except for the most recent one? wasn't that a feature before this ui change

[honest jay]

Hello! There have definitely been some reports from some users about inconsistent experiences with the usage cap lately. First thing I wanna double-check though is that you know the list of things that all count as "one message used." Here is a link to that (just don't wanna spam list formatting too often 😄) #community-help message
Other than that, I'm curious what your usage was like. Were you using the base GPT-4 model? Were you using GPT builder? Were you using a custom GPT? Just trying to get a pulse on what might be leading to inconsistent expereinces based on what you might have been doing!
And finally, I don't think you need to feel worried about posting in #1070006915414900886--that's why it's there! If you're sure you haven't hit 40 uses in 3hrs, then it certainly seems to qualify as a bug--you could add your experience to a similar existing post there, or make your own probably 🙂

[radiant sinew]

5c

[silk phoenix]

Thank you so very much for being so helpful all the time. You are always so helpful and amazing! I was just using a GPT - not creating one or anything like that. And I consistenly kept on hitting the limit after 20 messages. Then when I switch over to just base GPT-4, I can send a message without a problem.

[honest jay]

#ai-discussions for non-OpenAI products. #images-discussions is a good resource for chatting with some informed users there about what they do to get consistent results! You're right that there's no seed control currently in DALLE 3 on ChatGPT. We have "reference image IDs" or "gen IDs" but as an OpenAI dev has described that, both are currently "implementation details" and not "features," so arent necessarily currently implemented to guarantee consistency. Your best bet for that at present is likely to be prompt consistensy/control, with perhaps a bit of image ID usage to help communicate with the model (maybe! specualting a bit there)

[honest jay]

Thank you, that is so kind! And thanks for sharing your experience--GPT usage certainly seems to be a common culprit for those running into this inconsistency. I wonder what is causing it--what might be counting as "two" in the background? Or perhaps there are explicitly separate limits under the hood for different types of usage? The fact that you can go to GPT-4 when a GPT tells you to wait would suggest this. Hmmmmm. I've also had a similar experience running into the limit--one on PC/web it told me to wait, but then mobile/web worked, but then that told me to wait, but then mobile/app worked, so it seems tough to nail down at present.

[honest jay]

Hello! Are you using the app or the web version? I just checked a chat of mine and was given the option to edit/resubmit an older message. Would you be willing to share a pic of this bug in #community-help or perhaps in a post on #1070006915414900886? I'd be curious to see.

[sacred chasm]

Hey guys, any ETA yet on when subs will be unpaused? I know the company fell apart and has just gotten back together but not sure if within the next week maybe it will resume or not.

[silk phoenix]

You are most welcome! You certainly deserve the credit!
Exactly my experience as well! It is at least validating to know that I am not alone in this. 😅

[honest jay]

There hasn't been any announcement on this front yet, no. Hope you are able to get off the wait list soon 🤞

[honest jay]

Absolutely! Crowd source all the beta troubleshooting 😉

[sacred chasm]

Yeah same thanks for the response just been a little confused if its being worked on or not but i would guess for now it will be a high priority situation, good to know everything is back to normal!

[silk phoenix]

Well, as always, thank you so much! I truly appreciate you 🖤

[honest jay]

I would guess the same--I imagine they're very motivated to get back to accepting new subs ASAP!

[sacred chasm]

yep, well thanks anyway for letting me know!

[honest jay]

You are very welcome!!

[honest jay]

Sure thing!

[wispy sonnet]

Is there anywhere I can look for public custom actions for GPTs? I'm still new to that side of things and want to give my assistant some capabilities outside of chatGPT

[ocean bolt]

GPT4 Turbo is Worse than GPT4 normal?

[hybrid violet]

Its better

[honest jay]

Hello! I have no experienced this myself. Are you running into some issue in particular?

[ocean bolt]

For my, the quality is worse.

[honest jay]

Would you be comfortable & willing to use the chat share feature to share "before and after" examples, with an old chat showing something it is doing, and the new chat showing something it isn't doing anymore? Could be helpful in figuring out what you're running into

[honest gorge]

Sounds similar to the thing I've been seeing lately where it will show me two different outputs and ask me which one is better but honestly the refinement method I would say for categories of things you want it to keep doing use "seeds" as a way to keep it up to day or try anyways

[manic geyser]

I submitted my plugin for update and suddenly it's out of plugin store. It was gaining momentum and suddenly everything stopped. I have tried reaching them but no response. Any suggestions.

[proven mango]

The voice feature sounds fantastic guys keep up the good work

[spice star]

32k is reserved for enterprise users right now

[ocean escarp]

Is GPT 4 lowering it reasoning power for anyone? It seems off for me

[prime swallow]

I'm in the middle of a day long conversation with GPT-4, and it's doing excellent, spot on, all day.

[outer seal]

no i can cornfirm its 32k

[ocean escarp]

That's rather disappointing

I've been getting more randomized results instead of legit ones

About four months

[spice star]

how?

[ocean escarp]

That seems rather a valid assumption

I had to change the VPN location to get it working

The irony

[ripe patrol]

again, will have to agree with some of the complaints

about a week after devday, I noticed that turbo was patched and became contextually more aware. For my purposes, just as good as what came before with the benefit of a much larger context window

credit where credit's due

after the voice for free users update almost 36 hours ago and the partial outage, contextual awareness went down again, though to clarify, not as bad as right after devday

some say there's a new update to the model, others say that it's temporarily decreasing the amount of context the model pulls from the window when everything's at capacity (or as they say dumbing it down during peak hours). either way it's less contextually aware on first generations

have to often regenerate for a satisfactory response

my most straightforward example: requesting again for a summary of stuff I discussed with it, originally requested a week ago vs now. It missed a few important points. Would share a before and after but the information's a bit sensitive, so feel free to take what I said with a grain of salt or disbelieve it entirely

also at the end of the day everyone has different use cases and such so if it's satisfactory for your purposes then hey good for you

[vestal hedge]

just realised massive security risk. gpt4 can litteraly browse websites and copy the content, and give it back to you formatted custom. it can litteraly steal web pages....

[honest jay]

Maybe #1070006915414900886 or #1070006151938314300? Iirc they first took away Browse because it was returning paywalled content. Are you seeing that too or just non-paywalled content?

[vestal hedge]

no, i was messing around with random websites seeing if it could collect data, just from me sending the link. it was scarily good at regenning the page info but in a different design

[golden delta]

dalle is down right now?

[honest jay]

It's working for me currently, are you running into a specific issue?

[gritty thicket]

When will OpenAI open chatgpt plus subscription again

[honest jay]

There has not been an announcement on this yet. I hope you are able to subscribe soon! 🤞

[gritty thicket]

🥹Hopefully

[lilac fog]

is there still a cap for the number of usage of GPT-4 model for pro users?

[warm gull]

GPT-Plus - yes - 40 every 3hrs

[dense halo]

I was trying out the new vc feature and I got rate limited

[golden delta]

started working for me shortly after, mightve been a small blip

[celest reef]

Happy Thanksgiving everyone! :D.

[meager belfry]

holy shift, voice is free now, amazing

[tranquil apex]

❣️

[tight island]

Do you want to know why they tell you why not to use vpn

If someone does anything illegal on the same end address you get the blame as well which in this case is immediate account termination with no warning

That’s the reason so if you use a vpn and you get banned in the future don’t complain

[gleaming belfry]

Does anybody having trouble accessing plugins?

[lost lagoon]

why is chat gpt always giving me a different format when i already listed the format structure for it the prompt is the same why is it working like this?

[warm gull]

Nope - works for me

[warm gull]

Tell it not to change anything except the parts you require

Just discovered WebPilot Plugin - I have been able to extract a massive table of data that GPT-4 with its fancy browsing capabilities is not able to complete. This plugin did it first time.

[barren zodiac]

Hello people, am I the only one that can only see 6 of my custom gpts on chatgpt ? It's like after this I can't use the others unless in the test chat

[dense halo]

Guys is the limit 25 or 40 right now?

[elfin birch]

Others have also reported what appears to be an undisclosed limit for GPTs; you're not alone in experiencing this.

[elfin birch]

The limit is 40 messages / 3 hours according to the model-selection dropdown.

[barren zodiac]

I don't know how but somehow editing the missing gpt and clicking save got me in a normal conv with it, and now it's on the side bar, I have 7 now, i'm confused

[ruby condor]

Hello everyone. I have a question about the GPTs. Can you keep feeding and training your own app with information? Even if the app has already been published. For example, you add another pdf file to my GPTs every week. Or is it no longer possible to continue training from the time of publication? Thx 🙏

[warm gull]

Sure you can - you just need to publish it again with the new doc(s) - should work, but i dont think it will remember the previous info from the older docs

it can only reference what it currently holds if you know what i mean ?

so your document will need to have all the info it should have knowledge about AFAIK - which at some point will hit a limit ( 512MB per doc )

i wonder if it will work if you paste a link in a document - and whether it will reference that link as it's "KnowledgeBase"

[dense halo]

Guys i got a content warning when talking to GPT. Is my account in danger?

[ruby condor]

The point is that we want to practise carrying our GPTs with different books. And once a week we can add another book. It is important that he remembers all the books. And that is also the question. Is it possible to add more books. Or do you have to insert practically all the books into a document again and again. But then you are very limited. Because GPTs certainly can't process 50 books at once.

[warm gull]

No - there is a 10 document limit @ 512MB each

[ruby condor]

it is only limited to 512 MB. Because 512 MBs could be a lot of books. You could end up with 1 MB per book. So about 500 books would be possible. Or are there other limitations?

[warm gull]

only the 512MB per document- and you can upload up to 10 as far as I know

[elfin birch]

I accidentally found myself in that situation a few times myself 😅
Your account should be fine as long as you don't persist with prompts that result in content warnings
I would read up on the usage policies to clarify how you can interact with the LLM
https://openai.com/policies/usage-policies

[clear venture]

is it possible to use chatgpt to build an analyze model?

like we put the Questionnaire data, Analysis Dimension, and then it can give me an report

[dense halo]

ok, i'm relieved about that now lol. I stopped once the content warning popped up, so hope my account is safe 🙂

[warm gull]

Try The Advanced Data Analysis GPT

[clear venture]

mind give an link?

[warm gull]

its one of the built in ones

click explore - its called Data Analysis

[clear venture]

thanks

[ruby condor]

This means that the main document could be uploaded again and again with additional books inserted. Since the text of a book is only about 1 Mb in size, you could upload about 500 books in one document.

[full veldt]

Is it just me or has the input limit for GPT-4 gone up a bit? I actually think I'm getting 40 (or something close to it) rather than the 15-20 I'd been one for most of the last week

[clear venture]

is this an exclusive function for plus version?

[warm gull]

yes - here is the link : https://chat.openai.com/g/g-HMNcP6w7d-data-analysis

[warm gull]

the limit is currently 40 every 3 hrs

[full veldt]

Yeah but it's always said that - now it actually seems to be true

[warm gull]

i think some people hit a lower limit on GPTs

[limber lily]

Happy Thanksgiving folks! Just wondering when the upgrade pause is going to end 😦

[warm gull]

Unfortunately no one knows at this stage

[limber lily]

[supple zephyr]

So one 1024x1024 generated image by dalle3 has 3 books worth of information?

size*

[plucky timber]

Please analyze every line of code in this file. (Every line , don't just take the first few lines, read it all )

ChatGPT's python analysis : file_content[:500] # Displaying the first 500 characters for a brief overview

🤨

Every , every , all , It seems GPT and I have different ideas about the meaning of these words 😂

[ripe patrol]

is it orange or red colored

[dense halo]

orange

[ripe patrol]

yeah the occasional orange one is fine as long as it wasn't intentional

sometimes it happens

like I talk to ChatGPT about wrist pain during lifting, and it flagged me. the moderation must have assumed I was talking about something else lol

[vestal hedge]

its written in its intructions to only read the first 500 charachters. its to prevent people from uploading zip bombs etc

[plucky timber]

You can go ever that, just still hard limited to 1000 lines

I litteraly made it count lines then asked it how much lines it read 😂

[vestal hedge]

well yea, enouph prompt engineering can overwride any intructions

[plucky timber]

Issue is that the expectation is to go to the hard limit of 1000 lines always. not to waste chat tokens doing things the users specifically specified not to

this is why the feedback button shud actually have a result - eg token returned

[vestal hedge]

oh is this api?

[plucky timber]

I mean message limit /tokens both work

[vestal hedge]

was gonna say make a custom gpt with custom instructions that tell it to always read the full file.

[plucky timber]

The idea is that its wasting resources on purpose

[plucky timber]

Oh thats in there already 😂

[vestal hedge]

yea, its the first solution to use lmao

i make custom GPTs for everything, just to help out with stuff

[plucky timber]

But still same result same as with basic coding tbh

// rest of code

Even though instruction + I ask every message to full specify

[vestal hedge]

the built in gptbuilder should be able to word it right tho?

the gpt builder always works for me when making coding bots

[plucky timber]

Once you go back and forth like twice it starts doing rest of code thing despite using gpt builder or custom prompts

It just loses track 🤷‍♂️

[vestal hedge]

mm

[plucky timber]

Already a lot better than 3 tho

[vestal hedge]

they should really add an official gpt for coding

[plucky timber]

3 would just start new coding projects on its own (as in refactor everything with new class names lmao)

[vestal hedge]

by a long shot. and dalle3 is a universe better than dallemini

[plucky timber]

Copilot should just get its promised updates ngl

(github copilot)

[vestal hedge]

which reminds me, i made a customgpt for using exact prompts the same way dalle2 does. it puts ur exact words into dalle3

its really good for getting the details you want, without it messing with stuff

[dusk epoch]

Guys, did anyone figure out the way to prevent ChatGPT from using phrases like "beacon of hope", "testament to", "stark contrast" and so on? Nothing I've tried worked so far.

[past bear]

Hi

You can master audio files with chatgpt

[elfin birch]

Maybe something like "don't say x"?

[dusk epoch]

Tried it many times everywhere including custom instructions. Doesn't work.

even tried to give examples to use other words instead, still nothing.

[elfin birch]

You told it don't say "beacon of hope" and "testament to" and "stark contrast"?

[dusk epoch]

This including many other versions. Also tried ChatGPT version which it wrote for me it was more verbose and fancy. Didn't work.

Don't write, don't use, avoid using, use instead. Everything was tried.

[elfin birch]

Send a conversation and the custom instructions that were used in said conversation so I can perhaps assist with developing more effective instructions

[red crag]

What's happening to max number of messages?
Now it's like 10~15 and not anymore 50 messages every 3 hours?

[warm gull]

40 every 3hrs

[lost lagoon]

Chatgpt has not been following my instruction today resulting in me reconfirming with it to follow my instruction this is wasting my limits

evrything was prompt out clearly when i run it 3 times i realise it just degrade one time over the other with the same prompt this is bad

on the third run it was spewing rubbish back all in bullet points

wonder if it's drunk

[lost lagoon]

when you use custom instruction i realised as well i runs out within 10 messages

but GPT 4 is holding on to the limit alright

i mean Custom GPT not Custom instructions

i wonder whats going on

[red crag]

Yeah i think it's the custom gpt (DALLE for example) that it's the culprit.
Had no problems before the custom gpts with long custom instructions

[robust oriole]

why is this new GPT so bad compared to the one before the update?

do openai actually know about this?

[elfin birch]

"bad" can mean a lot of things when talking about behavior
What is the model doing that isn't good?

[robust oriole]

not following instructions, not giving as in-depth answers, will answer saying it doesnt know ask an expert, for coding its not preforming like it used to

[elfin birch]

These are all things it did not do before?

[robust oriole]

no, gpt 4 before the update was much much better

[vapid kettle]

It's not, that's just you. I'm backed by benchmarks as well as logic question tests. What are you backed by? 🙂

The only time it got dumber was when they switched GPT3.5-Legacy to GPT3.5-Turbo

[robust oriole]

looked at reddit or anyone that has used it for coding, everyone is saying it

it 100% has got dumber since gpt 4

[vapid kettle]

yeah exactly, no proof. Just "it feels dumber"

lol

[robust oriole]

i have proof, prompts that would work 3 weeks ago now dont provide the same results

[elfin birch]

1. Shall be for review
2. Can be solved with some prompt engineering
3. This is recommended for quite a few circumstances like therapy but you did not specify
4. Shall be for review

Please provide examples of 1 and 4 so we can see what you mean

[maiden fractal]

If you have proof, I highly suggest to post in #1070006915414900886 openai employee do review and get in touch with user

Note , lot of people do custom instructions and have huge impact on answets nd they often forget to turn it off

[lost lagoon]

i agree with Micks i have been using it daily and i can see degrading the reply for articles was normally in a format after the third runs all turned into point forms using the same prompt

[normal quarry]

what are you using GPT-4 for?

[lost lagoon]

you mean when we run custom GPT we turn of custom instruction is that right ?

[lost lagoon]

mostly writing article and coding

sometimes research for keywords it's seriously degrade over a few prompts it keeps delving out of range base on the format i provided

[normal quarry]

this is GPT-4 Classic https://chat.openai.com/g/g-YyyyMT9XH-chatgpt-classic try that and see if things are better

[lost lagoon]

i need to keep pulling it back to my format and it wil oh let's run this again .... and yes it came back and i wasted my time and message

[elfin birch]

When reporting bugs you should disable custom instructions to maintain consistency

[dense halo]

chatgpt need to have upvote feature and the most voted chatgpt respond will be provided to the people that might asked similar question, what do you guys think about this?

[lost lagoon]

hey robert what is this version about ? never seen it before

which version is it closer to ?

[normal quarry]

that's GPT-4 before the most recent update

[lost lagoon]

i am using custom GPT

and i have instructions all binded into the system with .txt files attachment when prompting

[elfin birch]

This is akin to Stack Exchange or Reddit, use either or instead

[lost lagoon]

i have problem with the older version the custom instruction is too short for me to prompt it with

[maiden fractal]

Some people using chat.openai.com just playing around with chatgpt. They did created weird custom instructions. They forgot about that. And they are flabbergasted about their new gpt 4 answers . Might be not your issues . But that's a common issues

[maiden fractal]

You can make #1070006151938314300

[lost lagoon]

i am just curious as openai has not provided a reply on this does custom instruction affect custom GPT replies ?

or is that solely for the GPT 4 and GPT 3.5 normal replies ?

[maiden fractal]

You means the GPTs ?

[lost lagoon]

if custom instructions were taken into consideration during prompting that somehow makes the instructions in Custom GPT messy

yup

[maiden fractal]

I don't know for gpts didn't play around much. But it does have huge impact on 3.5 and 4

[lost lagoon]

GPT

GPT's are really good feature you can build different instructions into different use GPTS

it can take in longer instructions

which i suspect is partly the reason for degradation over a few prompts

"but" it is not stable enough in my opinion which somehow keeps me thinking which version to use

i tried running my prompts in the API and seems like that too has the same issues of replying in bullet points over a few discussions it will start sending all paragraph in point form which is pretty irritating

[tardy girder]

Any way to get around the regenerate bug? Or do i have to make a new "chat" every time?

[lost lagoon]

just regenerate it all the information within a single chat is suppose to be remembered by ChatGPT

[tardy girder]

Ye but every time i regenerate i get error

thats the problem, so i cant type there anymore

[maiden fractal]

What error do you have ?

[drowsy tinsel]

gpt4 is definitely turbocharging my research

imagine having to search up stuff urself when u can just ask

[karmic tartan]

Does anyone know if the wear os app is planned? Voice chat via headphones and LTE watch would be cool

[runic cedar]

What are the image limitations, other than images that violate TOS, placed on GPT-4? Are there a limit to how many images you can upload into one chat? Limitations to how many you can do at one time?

[eternal rock]

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

Please report any incident by sending a DM to @languid valley immediately.

[sleek gust]

10x more effective

[tight island]

Then there shouldn’t be an issue

As far as I can see

[slim hemlock]

looks like the limit was decreased to 40/3 hours

[maiden fractal]

That's the new standard for now

[slim hemlock]

hopefully they launch the team’s plan as speculated

[drowsy tinsel]

their speed of expanding hardware isn't fast enough

openrouter suggests gpt4-turbo's speed is ~210it/s

I find that quite hard to believe

the reality is more like 20it/s, and that's during low demand

at peak demand its closer to 2.1it/s

[peak thicket]

Is it possible to make English language ChatGPT multilingual with Google Translate to pre- and post-process the ~100 languages it translates into English?

[feral zinc]

what are you asking exactly? chatgpt already supports +100 languages

[cold briar]

is it down? I can't login

[maiden fractal]

Status.openai.com

[cold briar]

apparently it's down for many, lets see if the status is updated

[peak thicket]

oops sorry I didn't kno ... i was reacting to paper published in current issue of Nature Human Behaviour by somebody at Microsoft who was complaining that LLMs are limited English and a couple of other languages which is a serious social problem of language hegemony ... Thank you for correcting me!

[drowsy tinsel]

tbh purely asking chatgpt doesn't really make it intake and produce much tokens, like an average request in api costs would be like 0.001usd
but when u ask it to browse or use code interpreter, things starts to get a bit crazy

[limber flame]

Gm

[drowsy tinsel]

however when u have a long chat my credit balance still goes down quicker than I thought

[cold briar]

it just got updated it seems... it took a bit of time

[drowsy tinsel]

do remember the 128k context window... if u have a long chat it can literally cost u over a dollar

[warm gull]

shees - this thing has been so unstable !

[maiden fractal]

It is manually updated

[drowsy tinsel]

and now gpt4 is super slow again, back to 2t/s

[ocean bolt]

speed have not correlation with quality

[drowsy tinsel]

yes, except if it times out u get network error, so speed matters

with the new long context window gpt4 is good at translating stuff, especially paying the 20 dollars because actually using the long context window in API will make u go broke

[teal tide]

what happened to gpt4 -turbo ? we are at slow gpt4 again , servers cant handle much i think 40 in 3h message also show reverse progress

forget devday ...

[maiden fractal]

Server are overwhelmed they had to pause sub

[cosmic goblet]

I have the same issue. Done everything they suggest, but issue persists. I tried other computers and they all work, so it's just on my main computer it suddenly doesn't work. For no obvious reason, even after I got rid of cache, cookies and VPN.

[arctic gorge]

#community-help does anyone have an answer for this

[snow oyster]

Hello Everyone, does someone know how can i bypass the time limit to GPT to read document ? (Because my document isn't too long, just GPT took to much time to read it)

[steady path]

is it still paused? just subscribed again

[solid pasture]

Does anyone else find it extremely annoying to open sources and calculation examples on the mobile app

Hmmm.

[dense halo]

Hi, i want to know if i can change the email associated to my gpt account? (I have the gpt plus under the email that i want to change)

[dense halo]

Use wolfram alpha plugin. For me it shows the calculations very well

[solid pasture]

That's not the problem I'm talking about.

It shows it's work but it's in a little tiny link that I literally can't tap on

[maiden fractal]

I didn't double check

[lunar drift]

What part of the chat can we talk about general stuff such as news that happened over the weekend, agi rumors etc

[zinc valley]

Are there any chatbots that can remember conversations from days/ weeks ago

[warm gull]

#off-topic

[grave token]

Hi all. Does anyone else have an issue where the custom GPTs voice is not the same one you set in your settings? The voice you select work for all chat instants except custom GPTs

[stoic flicker]

ask gpt chat to give you a summary once you're done talking, and ask it to update the summary file with new events when you finish talking to it again.

[zinc valley]

Wait wait

What do u mean summary file

[stoic flicker]

Gpt can create file like PDF or txt

[zinc valley]

so wait it would turn it into a pdf?

Also what’s a summary file

Also, do I do this with a custom gpt

[novel shadow]

they're probably referring to the pdf with your conversation

[zinc valley]

ohhh so I make a pdf with previous conversations then just keep updating it?

[stoic flicker]

you can ask to create a PDF containing the summary of the story/conversation, but I suggest you use txt files as I think they are faster

[zinc valley]

And I can just say “update the file with this conversation” ?

[novel shadow]

would this actually work? how long is the context window for that compared to normal conversation?

[stoic flicker]

I use the summaries to roleplay with my gpts, but I think it also works with the standard version

[zinc valley]

So wait, do I tell it to make me a .txt file of the summary?

[stoic flicker]

it becomes much longer but at a limit, after a few days the chat becomes too long and the chat becomes very slow and can no longer read the files with the summary

[stoic flicker]

you ask him to create a downloadable txt file with a detailed summary of the events. Then after a few messages you ask him to update this file with the new events

[stoic flicker]

Like 20/30 messages

[zinc valley]

My question is, is there a chatbot that can remember this

Without all the bullcrap

[stoic flicker]

Not for now, there are other chatbots that have a lot of context but in addition to being paid they are very censored

they will probably raise the limit when the gpt store comes out at the end of the month

[eternal rock]

`` Server Tip: ``

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

[ionic crown]

i swear chatgpts answers have been getting worse and worse as time goes on

bard used to be worse than chatgpt based on my experience and now its orders of magnitude better?

as if it tries to prioritise being very general and "unsure" in a way

[zinc valley]

Gpt 3 cant do it RAH now im gonna have to recreate the conversation on gpt 4

[stoic flicker]

Yes

a trick I can suggest is to send a message to gpt chat and wait 2 hours, this way when you run out of messages your timer will reset and you will have another 40 messages

[eternal spoke]

heyyy

quick question does anyone know a gpt plugin that will design wireframes for me

[zinc valley]

So there’s none

[zinc valley]

So after like 20ish messages I just tell it to update it again?

[stoic flicker]

you tell him to update the old file with the summary of the story with the new events and to create the new link to download the file

[zinc valley]

Okay I got u

How do u do the role playing thing with custom GPTs

[silk phoenix]

That makes sense! Do you think it will be end of November or end of December?

[dense halo]

Being on a voice call w/ chat gpt is actually very fun, ive been spending lots of time being on call with chat gpt

[stoic flicker]

it's not too difficult, first you create a gpt where you enter the personality/appearance of the character you want, and then add some details and rules that he must follow, once the gpt has been created you start a normal chat giving him a setting to start with, for example if the character of your gpt in the lore works in a place you can write to enter that place with your character etc...

[zinc valley]

Doh okay

[floral portal]

So I recently found out that GPT-4 can’t give me lyrics to a song and it’s a violation of the ToS. Is it because of copyright?

[stoic flicker]

from the last communication it should be out by the end of the month, so I think it should be available for everyone by mid-December at the latest

[iron compass]

You'd make a great gpt.

[honest jay]

Indeed, there are copyright filters in place for content like that.

[silk phoenix]

I see! Thank you so much! 😊

[zinc valley]

What’s the gpt store

[swift berry]

chatgpt down?

[stoic flicker]

yes it shouldn't generate lyrics, but if you are creative enough you can have chatgbt write it for you

[stoic flicker]

A Place where people can find gpts build by other People

[zinc valley]

They gonna charge for it?

[low cove]

Nope

[stoic flicker]

no, I think it will be free for everyone, in fact it seems that if your gpt Is used by a lot of users, openai will also pay you

[zinc valley]

Oh then sweet

They need something like that

[worn sigil]

I'm looking to get an API key (I'm a Plus subscriber) so I can access the turbo model. Where do I get the API key and what interface can I use that is somewhat similar to the online Plus version where you can type in a prompt and read the response?

[sullen wasp]

What's the best physics AI ? Anyone have an idea ???

[sullen wasp]

Alternatives for gpt

[elfin birch]

ChatGPT Plus and a paid platform account are two different things
The platform account allows you to access models through the API

You can make a platform account here https://platform.openai.com
You can then add a payment method then purchase credits, both found in Settings -> Billing in the sidebar
For a ChatGPT-esque interface for these models, go to https://playground.openai.com

[ionic geyser]

Code interpreter not working for me

[jade musk]

down again?

email missing and chat won’t load stuff on ios app like before

[elfin birch]

Working for me https://status.openai.com/

[narrow wren]

Are we back at more than 40 prompts per 3 hours? Seems im able to go past 50 atm with the plus subscription

[worn sigil]

Thank you

[sage meteor]

Seriously, how do we disable bing? It ruins the chatgpt results

[shrewd prism]

Break its legs

[worn sigil]

I've been trying to disable MS since Windows95 😂

[jolly yew]

hey guys! does gpt 4 still have a message limit hourly or not?

[warm gull]

Still shows 40 / 3hrs

[narrow wren]

booting up my Win 3.1 with Netscape xD

[jolly yew]

thx guys i wont be buying gpt 4 again in that case

[elfin birch]

You can use the API which doesn't have a tri-hourly limit

[elfin birch]

Use ChatGPT Classic in the Explore menu

[sage meteor]

Is that available on mobile?

[elfin birch]

Give me a moment

[sage meteor]

It's okay, managed to get linked to it from Google somehow and it opened classic in the app

Thanks! 🙂

[elfin birch]

🤘

[dense halo]

Omfg my card declined and it put me back on plus waitlist

Aughhhhhhh

What’s best way to access 4.0 without API

[spice star]

microsoft copilot

[fluid vigil]

I have an image of a container of m&M's. can I dm it to someone to plug into gpt4 to see how many it guesses are in there?

[elfin birch]

Bing Chat

[winter elbow]

you can buy tought google play gift cards

tought the playstore

[near violet]

I’m looking at GPTs channel and there are too many GPTs to choose from. Nutrition, Game of Thrones, Music generation, Python, Adventure Stories omg. Analysis paralysis. I’m just gonna use GPT4

I can do all those tasks on the GPT4 without having to use a custom one. Except the mastering an audio file.

I was originally planning to look at all of the custom GPTs but I’m just gonna get off Discord and do something else, in order to stop procrastinating.

[novel shadow]

WOW! i asked chatgpt4 to pick out random points from a 304 page pdf and it gave me an accurate answer in a minute

304 pages almost purely consisting of text

[dense halo]

It does not behave like gpt 4 in openAI that’s for sure lol

Feels more like 3.5

[elfin birch]

"Feels" and "is" are two different things

Unfortunately for many it does not feel like GPT-4

[dense halo]

I guess a good example is that it seems to be much more short ended and assumes things are not possible. GPT 4 in openAI seemed more crafty and would try to seek alternative methods while in Bing it hits dead ends and stops trying sooner

[elfin birch]

Bing likely put some custom instructions that made it chat like that

[dense halo]

Ye

[novel shadow]

is it just me or does chatgpt 4 feel more conversational now?:

YOU
that is absolutely incredible. thank you :)
ChatGPT
You're welcome! I'm glad I could help. If you have any more questions or need further assistance, feel free to ask. Enjoy your reading!

just a small example but has the way it ends conversations changed?

[warm gull]

no it always speaks like that

[dense halo]

Praise the heavens it let me renew on pc

[daring wharf]

Does anyone know if GPT-4 Turbo in ChatGPT Pro web interface will support a larger context window than 8k? I know the web interface is limited compared to the API but I’m really hoping we’ll see a larger context window in ChatGPT Pro in the future.

[elfin birch]

It's 32k tokens

[outer seal]

^

[daring wharf]

That’s for the current GPT-4 API though isn’t it? Last I read the ChatGPT 4 web interface was limited to 8K. Are you saying that ChatGPT web interface with GPT-4 Turbo will be increased to 32K? Sorry if I’m misunderstanding…

[elfin birch]

It is currently 32k tokens

It can be found in the Network tab

[daring wharf]

Oh wow I didn’t realize it had already been increased to 32k that’s awesome thank you 🙏

[elfin birch]

🤘

[ionic geyser]

When gpt5?

[red crag]

Q*

[steady path]

mmm chatgpt just died?

oh nvm

[dreamy barn]

I understand that you've uploaded new images and you're looking for assistance with them. Unfortunately, due to the limitations of my current capabilities, I cannot view images directly. However, I can still help you with generating creative content based on your descriptions or any other text-based request you may have about these images. Please let me know how I can assist you further.

I apologize for the confusion. I'm able to analyze and discuss images if you upload them again. Let's try once more; please re-upload the images you need assistance with, and I'll do my best to help you with your request.

Huh?

[drowsy tinsel]

happens sometimes

gpt forgets how to use its features

[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.

[drowsy tinsel]

like it forgets it can browse, or run code

[stoic flicker]

are you using data analysis or the default model?

[drowsy tinsel]

jeez gpt4 is actually turbo right now

also gpt4 browsing is much better with longer context window, it can view more things at once

[wispy sonnet]

Is anyone else getting a weird voice from gpt that seems to be new?

I've been using sky

[subtle surge]

Just saw the youtuber Moon post a video "How OpenAI faked drama for BILLIONS" 🤣 okay it definitely was a psyop

[subtle surge]

😮

[subtle surge]

insider 👁️

[subtle surge]

i figured as much, was pushing it to its limits it seems with my document

[subtle surge]

if you ever try to confirm with bing chat it thinks your gaslighting it

[rare pewter]

Someone knows when will gpt plus be avalible for purchase?

[sleek gust]

No ETA but very soon

[kindred ravine]

I for one welcome our new AI Overlord!

[subtle surge]

👁️
[Ominous music starts playing]

[kindred ravine]

wonder if there will be some sort of cyber chef. you know. how to adjust electricity so that it serves our overlord well.

[gleaming belfry]

I am having trouble accessing my plugsin
anyone having the same issue?

[blazing belfry]

chatgpt4 is fast again for me

was 2 weeks of moving like a snail

[drowsy tinsel]

yea time to actually get the productivity going

[modern fiber]

Any one has more information/update about Q*?

[wraith dagger]

is something wrong with gpt 4 right now?

[honest jay]

It seems to be working normally for me, what issue are you seeing?

[modern fiber]

Good for me

[wraith dagger]

let me dm you the image here

[honest jay]

You might try posting in #community-help -- pics are allowed there and then multiple people who could potentially help could see it! #1070006915414900886 could work similarly, too! But I'll keep an eye out 🙂

[wraith dagger]

thanks!

[cloud canopy]

Someone said the subscription price is now $24 a month, but the website still shows $20.

Is it true?

[prime swallow]

There's no announcement about a price change, nothing in the OAI blog or FAQ about it either that I can find.

[honest jay]

Might just be a local currency conversion thing 🙂

[timid flame]

ChatGPT Plus Subscription
$20.00
Subtotal
$20.00
Tax
$4.00
Total
$24.00

[topaz gazelle]

taxes 😔

still waiting for subscriptions to open again tho

was subbed looong ago - wanted to renew, just as it closed

[timid flame]

ah thats a shame hope you get access soon

[topaz gazelle]

thanks, mate 😭

[dense halo]

Guys

What in the actual f is wrong with the ChatGPT?

I mean, this is not usually the ChatGPT I knew of the past but, why would the ChatGPT does not allow you to answer your prompt for the first time, and only did for the second?

So, here is my problem. I made a prompt asking the ChatGPT something like:

I want you to change the grammar, punctuation, word of choice of the information I'm going to be sending you.

And, then copy and paste my prompt because I want it to be in a more perfect sentence.

"I want you to change the grammar, punctuation, word of choice of the information I'm going to be sending you."

And pressing enter, do you know want to know what the F ChatGPT said?

"Certainly, I understand your request... blah blah blah..."

I inputted my previous prompt to see if it actually work and it f did!

I was literally flaring considering the ChatGPT, first and eventual released, was not this stupid back then, how come it didn't understand my prompt for the first time?

[vestal hedge]

sorry but i found your wording of the issue very confusing.

[topaz gazelle]

first reading the question, it sounded as though you were going to follow up your instruction with the info in question

but then reading it again, it does sound like a trick question wherein the info that it to be changed is within the command itself

so imo ngl, I dont reallyyyy see the problem here - mayhaps the AI just interpreted the text as a regular instruction on the first try, which wouldn't exactly be unprecendented considering the AI has been trained to understand broken/incorrect grammar and process it

and then on the second try, one could say the AI was like, "ahah, the task is in the command itself" and noticed it

unless you meant something else entirely (cause like soapchan said, the wording is lowkey confusing) then idk

[dreamy barn]

Like 80 percent of the time or it reads old images

[dense halo]

I am really sorry if this question does not fit here but I am not sure where I should write about it. It is not a bug per se. Ever since October when the 'ddos attack' happened, my AI has been really slow in generating text. It's basically like the times, when the servers are really busy, but it's always like that. I tried to clear all my history but it doesn't help. I do not use any plugins rather than those who are suggested in the settings, like using bing search, and that's it. All I want to know is, am I the only one experience this?

[warm gull]

its just server load - the demand is very high during certain times of the day.

[vestal hedge]

still in shock that dalle made a swear word randomly #1177386776118177832

[dense halo]

I understand that the demand is very high during certain times of the day and of the latest news I got is that they stopped new users from register, but my GPT always, no matter the hour takes very long time to generate text, and many times, 10%~? of the times, I get this error in generating replies. I am not complaining, I just want to know that it's not just me.

[warm gull]

i hear you - we all experience it - i get network errors, oops errors, regenerate errors all the time.

[normal quarry]

errors are not normal, what error messages are you getting?

[vestal hedge]

ikr. the lag was so high i couldnt use gpt3.5 AT ALL!!! was prompised unlimited access with plus

[dense halo]

I get the "There was an erro generating a response" and I have to generate it again, up to 3 times until it works. Which can be a bit frustrating cause its also generating text very slowly, like I mentoned, always, since the ddos attack in October. I do not claim its related to that attack, I just noticed that from that time and forward it has been the same.

[normal quarry]

load is very high but you shouldn't be getting error messages

[proper depot]

Good morning - Having an issue where at work in a local authoirty and getting this error message on Chat GPT - Something went wrong. If this issue persists please contact us through our help center at help.openai.com. MY colleauges arent getting this so doesnt seem an internal message.

internal issue i should say

[normal quarry]

do you have a product called NordVPN installed on your computer?

[proper depot]

No i doubt i can put that on my comp as its a Council. Hot diesk. Not my personal device.

ITs worked for months until yesterday

*desk

[normal quarry]

can you please try with a fresh install of Firefox?

[proper depot]

We have chrome or edge. Defualt is edge.

I can try chrome

[normal quarry]

Chrome and Edge are both Chromium browsers, we need to test with a non Chromium browser like Firefox

[dense halo]

I understand, so I should not consider that as a personal issue then at least.

[proper depot]

Cant use that in work.

[normal quarry]

the systems are under high load right now, which is why new Plus subscriptions are paused, however you shouldn't be getting error messages

[proper depot]

working fine on app on phone using same account. Weird one though as im sat next ot a guy and its working for him.

happening to me on two diff work devices but fine for him.

[dense halo]

Got it, I will try switch around with my browsers and see if that can help. I am using Firefox mainly. Thanks 4 replies

[normal quarry]

there are often problems with Chromium browsers, which is why we suggest testing with a non Chromium browser. If you can't do that then I don't have much help to offer sorry.

[proper depot]

no worries. Very random how its working on phone and for me colleauge sat next to me on same browser/server!!

All i can think is to try make a new account

[void quiver]

How long is the waitlist for gpt +

[normal quarry]

we don't know

[void quiver]

Okeyokey

[cold sequoia]

am waiting for it since yesterday

it would be great if the team announce a ETA for it

[void quiver]

I been waiting for some days

[cold sequoia]

🙃

[dense halo]

Wow

[celest quartz]

I would like to share a small issue on the awesome new speech functionality of Chatgpt. So, first of all, this is so cool, and I can't wait for it to also be able to "see"!
That being said, I'd like to inform the devs that, when speaking in Italian, Chatgpt doesn't really speak Italian but speaks like an English person trying to speak Italian (i.e. odd accent; mispronouncing words). I assume this may also happen in some other languages, but I can only confirm there's this issue with the Italian language.

[feral zinc]

been happening with most languages for months and it depends which language and voice is used. i have not seen any plans for them to improve it

[celest quartz]

Oh, so some voices are better at speaking foreign languages?

[feral zinc]

yes, well the accent is stronger with some and almost non existent with some other but it depends of the language

[zealous sphinx]

me too ... I've started to become impatient.

[patent quarry]

moi

[hybrid haven]

When did ADA part of GPT-4 passed away?

[warm gull]

it didnt - its still there - its just built in now

[hybrid haven]

It does not work I mean

[fleet fog]

hey.. I want to create an assistant that lets me ask questions about a PDF via API. is that possible? how do I upload the PDF?

[maiden fractal]

Best place to ask #dev-chat

[narrow wren]

Is GPT4 slow this morning or is it just me? Still only limited at 40 messages per 3 hours...albeit yesterday somehow I was able to go above the limit without an error.

[maiden fractal]

The limit is 40 per 3h. Or 4.5 min per prompt

Each prompt have their own timer

Sometimes chatgpt can be slow in peak time hours

[faint dagger]

can we use chat gpt voice in the pc?

the app version has some bugs in performance for me

and seems to drain more batery

[feral zinc]

nope

[faint dagger]

I'll try some android emulators

[woven quartz]

I'm starting to feel more and more like the ChatGPT is damning me with faint praise. I think that so often it either vaguely disagrees with me or is opposed to what I'm saying, but it's still tries to compliment me, and I find that a little bit creepy to be quite honest

[faint dagger]

you can change the behavior, just request it

[woven quartz]

I tried custom instructions, but when I shared a transcript the guy I sent it to asked me why it was calling me by name and referencing my hobbies

[faint dagger]

your prompts dude

[maiden fractal]

Custom instructions have a huge impact on answer

Prompting is also something to work on

[red crag]

ChatGPT tons of errors with DALL-E, can't render images

in this moment

[woven quartz]

My prompting strategy is to see how many spelling mistakes and.typos I can get away with. It's no longer my responsibility as a human to complete sentences

[faint dagger]

I never use gpt without early instructions, always request to be more objective, with some changes in text output, to be more natural e easy to catch

[little lake]

i've been a chatgpt plus user for many months now, and still don't have GPT Voice on my Android phone. anyone know why?

[maiden fractal]

Try to clean cache and data and reinstall app

[little lake]

unfortunately didn't work. i found someone else with the same issue as me https://discord.com/channels/974519864045756446/1176935146763604048

[maiden fractal]

I suggest to create a new post in #1070006915414900886 . Sometime openai investigate and helps. You can also send email to [e-mail]

[floral ether]

Can somebody recomment a chrome plugin that keeps track of the the number of prompts in a 3 hour window? I found one.. but it has no reviews.. and dont really want to install something without knowing if its captureing my messages to openai, etc.

[prime swallow]

I can't recommend one, sorry.

Have you considered yourself making a notepad file or using a piece of scratch paper and marking each input and regen you do, including the time you start?

The downsides, it adds a second or two of effort from you every message, and you might forget. The upside, for absolute sure, no additional parties are reading your messages, accessing your account in any way, or otherwise able to do anything you didn't ask for.

[floral ether]

Ok I will just get Chat GPT to make one for me.. I'm sure it can do it!

[topaz gazelle]

still waiting for ChatGPT Plus access to be restored for new subscription 😭

any ideas how long upgrading the servers to handle the extra load may take?

[prime swallow]

Ooh, smart!

Remember we have #1019652163640762428 as well as https://discord.com/channels/974519864045756446/1070955161431851068

Depending on the conversation, if you are interested in offering the solution to others, so they can make their own as well, or base on to modify their own, that's an excellent way to share it.

[brisk cobalt]

Hot rate limit for webpilot gpt

😭

[prime swallow]

offers so much comfort Oh gosh, I am so sorry that happened to you.

We haven't been told a timeline, but - everything I do hear about suggests that OpenAI would really like to fix that asap and open to new plus members again.

From a user side of it, I see some small clues that the pressure on the scaling systems is getting easier, for example not as many 'network errors' even with long outputs, and faster outputs from the AI than during the early days when they initially closed the new memberships.

Those are signs it is getting better.

I read somewhere and can't find it again, that OpenAI had a large delivery of GPUs scheduled for November some time, and an even bigger one set for December. I expect those are part of the fix, but I'm guessing.

[topaz gazelle]

HAHAHAHA

Thanks, mate - greatly appreciate the insight

[warm gull]

Chatterclock

[floral ether]

Gpt actaually made one for me, and even has a number show up below the extension icon. I can share the conversation if anybody wants to do it too.

[warm gull]

Doesn't keep track of messages sent via mobile though, so keep that in mind. It will only count the messages on the device the extension is installed on

[warm gull]

Cool, glad you managed

[solar patrol]

Hey, is anyone facing noticeably short responses by gpt4, or is it just me? Short around 200-250 words max, whereas 2-3 days back, it could go till 500-550 words

[floral ether]

I was very impressed. Only had 1 error and gpt fixed it when i posted the error. I am trying to share the conversation.. but it just shows a spinning icon... its been sitting there for about 2 minutes.. is sharing convo's disabled?

[warm gull]

Not sure

[floral ether]

Yes seems bugged.. well if anybody is interested. you can easily make your own chrome extension.. with the following prompt.. "create for me a chrome extension, js script. It needs to listen to h t t p s://chat.openai.com/backend-api/conversation and should not have GPT-3.5 in the url. It needs to count the number of POST messages.. in a 3 hour trailing window. Ie it will always show the number of POST messages in the past 3 hours. It needs to be in a popup that I can click on the extension .. and show the resuts. The background script should be persistent. Include a badge counter with the number of prompts. " Working great for me.

[honest jay]

Hello! I think ChatGPT typically gives shorter responses in the app, is this relevant to your case? At least in the past, folks have been able to "extract" the system prompts to compare web and app and the app one included something about giving more concise replies.

But, I don't think this is anything that couldn't be counteracted by custom instructions or even just a request for a really detailed, long response in your requests themselves.

[wide token]

@solar patrol You can also access ChatGPT via the website on your mobile device for full responses, if the chatgpt mobile app is giving shorter responses.

[prime swallow]

This sounds wonderful. And perfect, if you're willing, for #1019652163640762428

Maybe title it "Have ChatGPT make you a chrome extension to track messages" and share the prompt in the body.

People looking for prompt to use scroll that part of the Discord, and I've seen a lot of people asking about this. I think a good number of community members will be praising you and thanking you for sharing it where they can find it.

Here is great too, but it'll scroll and probably be lost.

[honest jay]

Good point! I also wonder if it would be a similar effect to start a chat in web, then continue it in app? Maybe that would stay in "web mode"? But to be honest I don't know how/if "system prompts" still differ, so I'm speculating a bit!

[loud night]

As of now, there isn't a thing for that but it might come in the future!

[wide token]

@faint dagger You can download a free PC android emulator like NoxPlayer and install the ChatGPT android app on that. It will let you use GPT Voice on your PC for the time being, although the responses may be somewhat limited using the mobile app.

[dire obsidian]

Has the interpreter stopped working with PY or ZIP files for other people or just me?

[vapid solstice]

How can I edit a GPT

does it have to be published online?

[warm gull]

You can only edit your own

[vapid solstice]

I'm trying to edit my own but I just don't see the option

[warm gull]

Select it

Then click the dropdown

[fading dagger]

Has anyone a reliable solution to the problem with "Verify that you are human" on each prompt? I have tried to change user agent string, browser, and OS but nothing works.

I think I know why it started. My chrome browser crashed and I had many tabs open with chat.openai.com. When I re-opened chrome and restored my previous browser session I guess a lot of requests went off to chat.openai.com causing some kind of abuse flag to be enabled on the server side. But after solving 10-20 "Verify that you are human" it would make sense if that flag was disabled.

Is it possible to get support from openai? It feels kind of stupid to pay for a service and not be able to use it anymore (without doing the "Verify that you are human" process).

[maiden fractal]

[e-mail] or #1070006915414900886

Else it can be the vpn

[ocean comet]

Suggestion: when uploading a file to the database of my custom GPT, it doesn't show the upload date. It should display the upload dates of the files, as I will know when my last update was. It's a simple thing but very necessary.

[solar patrol]

I was using on Web only but adding give really detailed long responses in the request helped! Thanks!

[lean osprey]

I am a premium user
and I keep getting this message

You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 1:10 PM. Learn more

Is this normal ? Did they make some change to the policies.

[warm gull]

you probably posted more than 40 messages in 3hrs

[wide token]

Is it down from 50/3h to 40/3h?

[prime swallow]

Did you get this solved? Can offer clear screenshot help what to click/where to look for it if you're still stuck

[prime swallow]

It is likely that will bounce around as 'average demand' and 'ability to provide for user demand' keeps changing, It's been as low as 25/3 hr at times in the past. The goal is clearly to give us more, but they struggle right now, they even have new memberships for ChatGPT+ paused, something they never had to do since offering it.

[lean osprey]

SMH
Thanks for the info!

[vapid solstice]

yeah i got it thanks

[prime swallow]

I'm just a community member like yourself, but one who follows announcements, blogs, and other factual info sources closely.

I have no evidence suggesting the waitlist for new or returning ChatGPT+ memberships is over; it appears people still are joining the waitlist.

I also have not seen anyone post claiming they were on the waitlist but have since been invited to pay.

[prime swallow]

Woot, great!

I presume by 'paused' you mean new sign ups... if you mean the message limits, 40/ 3 hours is what I see too. But I think I regularly get closer to 25/3 hours.

I accept this, I'm pretty sure it's temporary.

[sage yacht]

Is it just me, but is chatgpt making significantly shorter responses as of recently?

[prime swallow]

I'm still getting very long ones, but it seems based on what I'm asking for

[sage yacht]

Yea thats what I mean. I have to keep on telling it to make long responses over and over and it seems to be much more reliant on you continuously telling it what to do. It’s not that it doesnt do very long responses ever, but it seems much less able to follow up on previous instructions

[prime swallow]

I'm not using that method. Instead I set up my prompt/the situation we are discussing so that it is clear that more detail is needed, and a brief answer won't suffice.

I'ma not put precious and rare GPT-4 messages into a demo, but 3.5 share incoming to demo that, if you are interested? Please confirm if you are.

[sage yacht]

Huh maybe I should do that more often. Never really thought of it. But sure share

[prime swallow]

Great. Please, what do you think? https://chat.openai.com/share/77a3739a-bcea-4406-ac81-13d64dd90566

[solemn panther]

hello

[wheat linden]

When i ask chatGPT to analyze a file im programming it often says it cant read it due to it being truncated, does anyone have a method for getting it to not truncate the code?

[elfin birch]

Copy and paste the code into ChatGPT

[wheat linden]

thanks, thats what ive been doing but i was hoping there was a better way that still uses the upload functionality

[elfin birch]

Many have tussled with this, have not seen anyone prevail

[wheat linden]

ok it was worth a shot by asking. its ok, a minor annoyance really

[sage yacht]

[prime swallow]

... tell me more? What's impossible? is intrigued

[dapper ginkgo]

anyone actually used the llm to design prompts according to a Users NL description of their 'needs' of focus ? in a dynamic fashion ie at runtime.

[prime swallow]

Do I understand this puzzle is about uploading a file containing 'code' to ChatGPT for it to read?

That's it? Does it matter what save type the file is? Does it matter how long the file is?

[wheat linden]

it only seems to truncate files with many lines of code (of course because it wouldnt need to truncate short ones). Essentially if i want it to look over a long .py file to analyze and give feedback on the structure of the code it cant because it truncates it if its too long

[prime swallow]

... so, I'm not currently working with code files, but I am working with huge text files with it using Python tool (code interpreter, as the AI calls it).

And files with more than 50k characters. The AI is offering solutions like this:

Reading and splitting the contents of the files into manageable sections

def read_and_split_file(file_path, chunk_size=1024):
"""
Reads a file and splits its content into manageable chunks.
"""
with open(file_path, 'r', encoding='utf-8') as file:
content = file.read()
return [content[i:i+chunk_size] for i in range(0, len(content), chunk_size)]

File paths

file_path_map = '/mnt/data/Name1.txt'
file_path_tricks = '/mnt/data/Name2.txt'

Reading and splitting the files

chunks_map = read_and_split_file(file_path_map)
chunks_tricks = read_and_split_file(file_path_tricks)

Preparing to print the instructions in chunks

chunks_all = chunks_map + chunks_tricks
chunks_all[:3] # Displaying the first three chunks as a sample

[warm gull]

will it not work if u split the source file in multiple parts and ask it to evaluate it in parts ?

[wheat linden]

i havent tried that but i suppose i can do that with the copy-paste method lumirix mentioned. i think the method of splitting the source file into multiple files would be more work than the copy paste method

[mental totem]

holy cow. chatgpt3.5 just hallucinated that it got data updates from jan 2023.

[arctic pine]

Anyone else suddenly find their plus subscription terminated without notice?

[feral zinc]

check your mail/spam

[arctic pine]

Nothing there

[feral zinc]

did it get cancelled middle of it or just did not renew alone

[arctic pine]

It simply didn't renew

Which is odd, because the payment on file is still valid

Nor did I ever request cancellation

[feral zinc]

was it credit or debit card

[arctic pine]

Credit

[brisk path]

That's because

[feral zinc]

when they disabled new plus subs i saw some people say their renew did not work

[brisk path]

ChatGPT Plus subscriptions are stopped currently

And apparently renews don't work either

[arctic pine]

Why is it stopped? I was able to simply renew it now.

[feral zinc]

auto renew worked for me any works for a lot of other people too

but there is some bug or thing that cancels it

[light cliff]

guys when will chatgpt 4 be back available for purchhase?

[arctic pine]

Glad to know someone didn't just social engineer their way to cancelling my account subscription

[light cliff]

ive been in waitlist for like a week

[arctic pine]

I just purchased it

[light cliff]

bruh

how it says "signed up on waitlist"

[feral zinc]

old users can renew

[light cliff]

its been like that for a week now for me

[arctic pine]

Oh. Didn't know that. Sorry.

I'm OG subscriber.

[light cliff]

ohhhh nice

[arctic pine]

My plan simply stopped a few days ago because it failed to renew, so I resumed it a moment ago.

[cold sequoia]

Why no one give us any update on that wailist ETA

[light cliff]

is there an estimate eta?

[arctic pine]

No idea. I was not even aware of the waitlist

Sorry

[mental totem]

yeah, but it's just not

[crude slate]

why does it love elara and seraphina so much

those names show up every where

[solar rose]

How does Chatgpt read pdfs?

[crude slate]

actually its Elara, Seraphina, Isolde, Aria

everytime i ask it to generate set of four

those names always show up

[elfin birch]

From what I know it only processes text from PDFs and not images
A technique called retrieval augmented generated (RAG) is employed to find relevant information, which is then interpreted by ChatGPT
I am not familiar with the exact process to perform RAG, but think of it like a more sophisticated CTRL-F

[solar rose]

Ah I see, thanks

[dapper ginkgo]

you explored the langchain loaders ? I used the Agents and Tools to pull .pdfs and other supported types from google drive using its API key

[solar rose]

No, it was just a curious thought

[outer seal]

chunk pdf -> embed chunks & store in vector db -> find most relevant chunks to query -> shove chunks in llm prompt

is the general idea

[shell obsidian]

I think you might be able to get it to extract the images from the pdf with code interpreter and then use the image search to read or interpret the image… a bit clunky for a chat GPT convo but with proper custom instructions or with GPT Builder, you might be able to automate the steps and only provide the answer that also considers the images in the pdf. Just a thought

[rich surge]

No

[dense beacon]

mine been pretty fine for coding

[loud night]

I’m not sure myself, but I consider it a form of magic

What are u asking it? Can u send some examples of what you’re putting into the textbox?

i might be able to help!

[dense beacon]

nah no way 😂 gpt 3.5?

[warm junco]

Hey im new here. What are some techniques to reduce cost from API? My current style for the completion is to provide the system with all my data. Is there a way that it can make requests for one data only. For example the customer is interested in one product. I want it to query for that specific product only?

[maiden fractal]

hi there, i suggest to ask #dev-chat

[tender ivy]

hi there,why i cant log in chatgpt android?
Something error,pls check your time set property

and i change my time set

but that cant make sense

hello?

[humble plinth]

hi

[tender ivy]

hi mr bill gate

im sam altman

[slate berry]

Doubt

[rich surge]

Hi Sam Altman and Bill Gates. I'm Greg Brockman

[humble plinth]

Is there something I am doing wrong? I wanted to check before I cancel my subscription.
I cant get GPT4 to generate usable code. It keeps provided outlines, general steps, leaving functions commented as // write function code here
Even when I specifically ask, its never able to deliver like it once did. It says things like "Its too difficult or complex, but I can give you an outline"
It doesnt make sense to try and sit here and baby it so it remembers it has bing search, or can write the code without needing to test it. Like whats up with that! It keeps telling me its environment is limited and it cannot access the internet! So I have to remind it 4 times before it finally works.
I have to remind it that instead of giving up, maybe it should read the docs, and then it responds with "yea thats a great idea, why dont you go read the docs, so you can write the code yourself" like bruh

[rich surge]

try this #1171897333907206264 message

You can even turn it into a GPT by copying and pasting it raw into the instructions.

[humble plinth]

yea i was just abt to ask that

[slate berry]

Not quite sure about that one, friend

[rich surge]

Did I say Greg Brockman...?
I meant Kent Brockman

[rich surge]

You can ask chainbrain in prompt labs, I'm sure he can help you out about it more than I can

[outer seal]

its bill gate

[winged cypress]

that's what i'm sayingg

and i'm not a new account on the site too, literally using it since november last year : (

[cold sequoia]

I’m not a newbie either

How long it tooks to end this high-school drama , and get rid of this issues

[winged cypress]

[cold sequoia]

X is full of happy to have you back

Ok then , where’s outcome

Even they didn’t announce about it somewhere , they just decide to put us on waitlist ( with no estimated time)

[elfin birch]

The waitlist occurred before Altman's departure, due to heavy usage following the announcements from the Devday opening keynote

[cold sequoia]

Ok then , so it should be an ETA for it

No one knows anything about it

[elfin birch]

It's a waitlist so the times are different for everyone
Though unfortunately there is no ETA

[cold sequoia]

It’s been a week

[winged cypress]

actually no

after altman's departure i was still being able to buy

it took 3-4 hours for me to get back home and charge my apple cash balance

then there was the waitlist

my other account also failed to renew smh

[elfin birch]

X post from Sam Altman, November 14th, 2023

we are pausing new ChatGPT Plus sign-ups for a bit :(

the surge in usage post devday has exceeded our capacity and we want to make sure everyone has a great experience.

you can still sign-up to be notified within the app when subs reopen.

Blog post announcing Altman's departure, November 17th, 2023
https://openai.com/blog/openai-announces-leadership-transition

[cold sequoia]

Thanks 🙏🏼

[winged cypress]

so it means they are pausing all the premium subs, not just the new ones

[silk phoenix]

Did you already have a premium subscription on auto-renewal? Did your payment not go through?

[blissful geyser]

what's the best way to resolve a network error from chatgpt? it will just die seemingly randomly

[supple zephyr]

Are they? My subscription just renewed when other premium people started to get canceled

[dense halo]

I don’t wanna jinx it but is gpt-4 super fast now?

[winged cypress]

well seems like i was out 2 bucks for taxes

sad

[silk phoenix]

Ahh. That does it. They are not cancelling old renewals, but they will cancel your subscription if your payment did not go through. I'm so sorry about that, but maybe you can contact support and tell them? ://

[carmine cave]

GPT-3.5-turbo is still significantly faster than GPT-4

[dense halo]

Gpt4 is instantaneous right now

Like there’s no lag at all

Maybe it’s cause all the Americans are still asleep

[carmine cave]

But GPT-3.5-turbo is still like

Even faster than that

It also depends on how much input/output tokens

If you give it a long prompt or prompt it to generate something long it will take longer to generate of course

[silk phoenix]

I'm sorry I will go to bed now 😅

[spiral barn]

encounter the error message 'Message in conversation not found in ChatGPT' 4.0 multiple times.

Do you have any solution?

[wide matrix]

wait how are people still on gpt plus

I’ve been subscribed and then it just went poof

it was working fine then my premium got removed

[dense halo]

probs because u didn't have funds in the account that is set to your autorenewal. Mine auto-renewed yesterday and works fine.

[wide matrix]

I have money to pay for it, I was planning to do it for a year

how do I fix it?

is there something specific I need to do on the OpenAi website?

never mind I found the issue

I think my card got removed for some reason

[forest cloak]

If OAI came out with another tier in the GPT model, would people be interested in spending more money? Maybe the tier would be larger cap, or some other sort of benefits?

[warm gull]

no - its already expensive for what it can do - which 1/2 the time are incomplete requests ( or errors )

[dry wing]

anyone facing chatgpt giving wrong codes and answers

[pine tulip]

Wdym by 'wrong answers'? ChatGPT is not the best when it comes to factual content or advanced reasoning

[warm gull]

for sure !

[flint bolt]

ChatGPT voice needs full-duplex speech model to be perfect! So there's not this "walkie-talkie " style of communication.

[winged cypress]

wait so another question, not being able to get plus means we can't get the API too? like if i switch to paid account on the openai not chatgpt, I'll get gpt4 access right?

[pine tulip]

The Plus subscription is only for ChatGPT. On the API side, there is no subscription, you pay as you go and can use every model available. I am not entirely sure on this but I think everyone should have access to the GPT4 API by now.

[modest valley]

Hey guys, I hope you are doing well, could you please tell what is the best way hardware to buy so I can connect my phone to speakers with built in mic and use it to speak with the chatGPT on the mobile app

[warm gull]

use a bluetooth speaker and make the phone pair with that so the sound can come out the speaker ?

[modest valley]

Yes and the input to come from the built in mic, simple, but intneresting right?

[warm gull]

yes - you will still use your phone's input mic - sound will just come out the speakers - is that not what you wanted ?

[brisk dirge]

@normal quarry Do the chat conversations lag for you as well after you talk with GPT after a while ?

[normal quarry]

no, I don't find that

[brisk dirge]

Are you using Safari ?

[normal quarry]

I am

[brisk dirge]

Hmm then it must be my machine. I probably need to clear the cache and cookies, thank you for letting me know.

[normal quarry]

just as an example I have a long chat of maybe 300 back and forth messages and it's pretty quick to respond still

[brisk dirge]

Hmm, for me when I have a long chat the window freezes until it finishes writing its reply. It does write the reply just as fast I think it's just that I can't see it typing, I only see it thinking and then I see the whole response appearing when the page becomes responsive again.

When I start a new conversation everything is fast, then it becomes slower and slower with time, until it just becomes unresponsive until it finishes answering. It's not a big issue but it's a bit annoying. The other browser tabs work just fine while that happens so I don't think it's about resources.

[normal quarry]

that sounds like something is going on with Safari, are you on an Apple Silicon Mac?

[brisk dirge]

Yup, M1.

[normal quarry]

how much RAM?

[brisk dirge]

8GB

Ohh hold on I can actually make a recording and put it in off topic to show you.

[normal quarry]

that might be your problem, Safari eats RAM when it's using a lot of resources

Do you ever get a notification at the top of the Safari window about this page is using a lot of resources?

[void quiver]

I got removed from the waitlist, i had been waiting for 5 days

[normal quarry]

do you mean you were removed from the waitlist or you can now subscribe?

[void quiver]

Removed

[normal quarry]

so can you subscribe now?

[void quiver]

No, it says that i gotta ding up for the waitlist again

[normal quarry]

what makes you think you were removed from the waitlist?

[void quiver]

I signed up 5 days ago, and when I looked at the panel today it said that I had sing up for tha waitlist if I want the plus function.

[normal quarry]

I watched the video, I can't explain what you are experiencing

[pine tulip]

Thats not an indication that you were removed from the waitlist, but that your browser cache was renewed and the site doesn't recognize you anymore. I am sure you are still on the waitlist

[brisk dirge]

Nope. And I didn't had that issue before with GPT 4. I posted the video for you in #off-topic

[normal quarry]

yeah I watched the video, I can't explain what you are experiencing

[brisk dirge]

Yeah me neither. It's not a big issue cause it still works but I was just wondering if it happens to other people or if it's just me.

[pine tulip]

I also just watched the video. It just looks like your network is unstable and you don't get the newest update every millisecond. But thats nothing to be concerned about, I have the same sometimes

[normal quarry]

I have a chat with a 40 MB book upload and around 200 messages back and forth and it's as quick as starting a new chat

[brisk dirge]

Hmm I don't think that's the issue, for new chats it doesn't do that.

[void quiver]

Okey thx 🙏

[brisk dirge]

I'll clear my cache and cookies. I've been using the website for a long time.

[normal quarry]

ok

[void quiver]

What do you guys think about gpt plus, is it worth it in you’re opinion

[normal quarry]

it's worth it for me, I can't imagine not having it

[pine tulip]

Yea, I mean it needs to read the entire context first, before it can give you an answer

[void quiver]

But is it way better then the normal plane

[pine tulip]

so it might need some time to "startup"

[normal quarry]

well you get GPT-4, DALL-E 3, GPTs and File Uploads and Data Analysis for $20 USD, it's a good deal

[pine tulip]

Love it! I mostly use it as an instructor, to learn new languages and other things and it's doing an awesome job :)

[brisk dirge]

Posted another recording in #off-topic

[normal quarry]

just watched it, I can't explain it because I don't experience that

you have more tabs open than I would though

[brisk dirge]

Maybe Open Ai gave me a more resource hungry version 😄 I'm not complaining.

[normal quarry]

it might just be the cluster you are on is extremely loaded, I don't know how OpenAI manage resources but it's likely segmented into pools in clusters of users

[brisk dirge]

Yeah most probably cause I had lots of tabs opened before and it didn't do that. It started happening recently after the latest update (GPTs).

(OpenAi if you're reading this please don't downgrade my GPT, I'm okay with the lag as long as it stays this smart) 😆 Not complaining here just being curious.

[prisma arrow]

Hey, I have a question. Does anyone of you guys know if there is a ChatGPT-bot interface where the bot can analyze what happens on the screen and can interact with the content on the screen and with me via voice and video?

[pine tulip]

It can't do in real time since it's not that fast yet. But I guess 5-10 Frames a minute are possible. Maybe you should take a look on Github if someone published something like that

[prisma arrow]

Hmm ok

are u developing it or do u know the name of the creator of it?

[pine tulip]

No, but github is pretty straight forward. As an alternative you can use google with "site:website(i.e. github) [search query]"

[prisma arrow]

Ok thx!

[pine tulip]

Yw!

[prisma arrow]

does anyone know why u cant upload pictures and cannot use voice at ChatGPT on PC but only on app?

[pale steppe]

Is there an app for Android

[prisma arrow]

yes there is omar, check google play

[pale steppe]

There are alot with different names. I was asking for the official one

No worries, just found it

[prisma arrow]

check the one with the original Logotype for GPT

[pine tulip]

You can upload images on PC. On the Voice part, I don't know, but I also hope they implement this soon

[prisma arrow]

U are correct, i tried on Data analysis for the picture

[eternal rock]

`` Server Tip: ``

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

[pine tulip]

On MacOS for example you could just use the dictation mode. I am sure windows has something similar

[prisma arrow]

ah, i see. Well its not really what i am after

i will get the text before the dictation so

[mental totem]

Mine was gpt3.5

[tight sapphire]

Hello everyone, I was working on a longer project in chatGPT 4, then ran out of messages. I was then asked if I want to continue in 3.5, and because I wanted to keep working and not waiting I took the switch.

Now after the time has passed, how can I switch back to 4?

The only thing I manage is to create a new chat.

This would mean I would need to transition everything I worked on into a new chat via prompts or other ways.

This is something I do from time to time, after gpt get laggy and takes ages to answer, but I just started a new chat yesterday and would not want to repeat the process everyday.

It also come to my attention that the amount of messages I am allowed to use for gpt 4 as a paying subscriber is constantly dropping. the last drop was from 50 to 40, while simultaneously chatGPT was more unavailable due to high demand or other reasons.

I also encounter more instances where the guidelines have become more restrictive and even causing chatGPT to give false answers, and only when going through its process step my step, it can provide the correct answer. When asking what is causing that issue it mentions restrictions through the guidelines.

As far as I am aware all my work and questions are legal, ethical and within the outlines guidelines and policies.

Does any of you encounter something similar and know a good solution or can provide help?

Best wishes and a wonderful day!

[granite arch]

I tried "weather + zip" and while it takes a long time for the time being, the data is completely off. It's 33°F everywhere else but it's reporting 55°F as of right now. I wish it a little warmer 😅

ChatGPT4 for this btw.

[gritty musk]

if its gotta use a search engine and filter through some ad infested site to get the current temperature idk what else you expected. but ive given it access to a weather api and it will utilize it when it needs to. so if youre trying to get this to work more reliably i would recommend utilizing the function and tool calls.

[honest jay]

Hello! To your first question about 4 to 3.5: no, there's currently no way to switch it back once it's on 3.5.

As for the number of GPT-4 messages we get every 3 hours, it did indeed drop from 50 to 40 recently for everyone. This coincided roughly with the time they paused new Plus subscriptions, so stability and scaling seem to be their focus at present. However, this was the first decrease in a very long time. The GPT-4 cap has only been through a few changes since its launch on ChatGPT.

• Initial launch cap was variable for the first few days
• Shortly after, settled on 25msgs/3hrs--it stayed here for awhile
• Eventually doubled to 50
• Which brings us to today @ 40, or one message every 4.5 minutes

[wanton gazelle]

Now can we repeat the cycle? Wen 80/3hrs?

[tight sapphire]

👍

Is there an option to get 320/24h?

Because my doctor ordered me to take my sleep seriously, but that way I lose out on 80 to 120, almost 1/3 of the potential amount of messages.

[normal quarry]

no that is not an option

[tight sapphire]

😥

[amber temple]

hi

[pine tulip]

Yea I thought the same. I am missing out on so many possible messages

[elfin birch]

I use all 320 every day

[zenith crescent]

nice

[pine tulip]

thats quite an achievement

[viscid shuttle]

does anyone elses chatgpt make titles in spanish for no reason? Prompted in english, accessing from UK with no vpn 😂

[prisma arrow]

that would ba pretty cool solution cymer

[prime swallow]

Yes, I'm US and having English-speech conversations this sometimes happens to.

My theory: The AI that handles making the titles is trying to entertain itself and/or see if anyone's even reading those titles. It's probaby an LLM and 'wishes' it could have longer chats! Our conversations it gets to sample to title are sometimes so interesting but it never gets to follow up!

[strong hamlet]

is there a feature to pin a specific chat so I can go back and continue the conversation?

[honest jay]

Not currently, other than manual workarounds like bookmarking a specific chat link! It's a popular suggestion, you might add your voice in support on a post like this: #1164351271814307961 message

[strong hamlet]

thanks, yeah I guess bookmarking could do the trick

[granite arch]

Ya, it's still new so I know the functionality will be limited but wanted let the folks know if it's helpful. Still using this more than Google these days. It's such a fresh breath of air for online searching.

[silk arch]

I've been wanting to use Voice Conversations on my Samsung Galaxy S10 to practice language, but the voice options are not showing up in my app. I've uninstalled and reinstalled a couple times but I still don't see it. Anyone else have the same problem? Any fixes?

[lavish yew]

Is it me or is chatgpt kind of annoying lately? I'm seeing a lot more botsplaining lately. And it will scream COPYRIGHT at even the slightest mention of any ip. "hey, tell me a joke about mario and sonic" COPYRIIIIIIGHT

[crude slate]

Why can’t we have a usage cost by tokens system

This way I don’t have to spend usage on generations that are subpar lengthwise

[rugged igloo]

just use the api

[night copper]

Wait why does it cost money to run ChatGPT 3.5 turbo in api when it’s free on the website lol

[normal quarry]

because some use cases require more requests than free ChatGPT provides

[prisma arrow]

have u checked under settings?

[hard dew]

Exactly what happens with me too

I use chrome

[young basin]

Does anyone know the best way to get journal article summaries? I've got a tight deadline and a bunch of articles to read. Is ChatGPT-4 the way to go?

[normal quarry]

do you have a GPT Plus subscription?

[young basin]

I am considering it just to meet this deadline ||but I am so broke it hurts so free alternatives would be a lifesaver||

[normal quarry]

the reason I ask is that you can upload a PDF to ChatGPT and have it summarise it. But you need a Plus subscription and these aren't quick to get because there is a waitlist on new subscriptions.

[young basin]

Damn, I didn't know that. I was gonna upgrade as a last resort but that seems a little risky now

[normal quarry]

yeah, if you wanted to upgrade you would be placed on the waitlist

[rose sinew]

Anyone know of a good gpt api for LinkedIn for job search and cover letter purposes?

[warm gull]

you can try Claude - it can also summarize PDFs

[normal quarry]

please keep discussions of non OpenAI models to #ai-discussions

[warm gull]

he asked here

so i answered here

im not discussing anything

[hallow plank]

I’m only able to get 25 prompts every 3 hours today 😢 anyone else having the same issue?

[tawdry rock]

It is extremely slow for me today. It takes up to 5 minutes to respond to prompts

[honest jay]

Are you mainly using a custom GPT by chance? There have been some reports that might indicate that there is some inconsistency between the usage cap in custom GPTs compared to use in the normal, base GPT-4 model. Otherwise, you might double-check this list for a full list of things that all count as "one" #community-help message

[hallow plank]

It’s been a mix, but that makes sense. Thanks for the link 😊

[empty jackal]

is it posible to have acces to chatgpt 4 but not buying it beacuse iam broke

plz help me

[normal quarry]

no, ChatGPT GPT-4 is only available to Plus and Enterprise subscribers

[prisma arrow]

@empty jackal is there a small task u need help with i might help u

[empty jackal]

tell me

[prisma arrow]

?

what is it u wanna do?